Ceb Talent Analytics Quarterly q1 2017

Talent Analytics
Quarterly
Q1 2017
In This Issue:
What We Heard from

You in 2016
Aligning Talent Analytics

to CHROs’ Strategic
Priorities
Starting Simple at EY
Talent Analytics
Innovations: Separating
Fact from Hype
Beginner’s Guide to Text

Analytics for HR
The Advantages of
Machine Learning in
Candidate Selection
(and Pitfalls to Avoid)
Drawing Bold Conclusions

at AstraZeneca
Lessons from Marketing:

Communicating Data-
Centric Insight
Contents Talent Analytics
Quarterly
Q1 2017
Q1 2017
CEB Corporate Leadership
Council™
Editors
Brian Kropp
Daniel Metz
4 What Are Your 2017 Talent Analytics Priorities? Authors
Infographic Danielle Douglas
Matt Dudek
Caitlin Dutkiewicz
6 What We Heard from You in 2016 Blakeley Hartfelder
Omar Hoda
Jessica Knight
Andrea Kropp
Susannah Schools
9 Aligning Talent Analytics to CHROs’
Marianne Stengel
Strategic Priorities
Content Publishing
Solutions
11 Starting Simple at EY Graphic Designer
An Interview with Blair Hopkins Stacey Phipps
Editor
Melanie Meaders
14 Talent Analytics Innovations: Separating
Fact from Hype
16 Beginner’s Guide to Text Analytics for HR
22 The Advantages of Machine Learning

in Candidate Selection (and Pitfalls to Avoid)
25 Drawing Bold Conclusions at AstraZeneca

An Interview with Christian Cormack
28 Lessons from Marketing:

Communicating Data-Centric Insight
31 In the News
2 Talent Analytics Quarterly © 2017 CEB. All©rights

2017 CEB.
reserved.
All rights
CEB170353GD
reserved.
Letter from
the Editor
It’s an exciting time to be involved in talent analytics; in many ways, the field
is exploding. The ranks of talent analytics professional groups are swelling:
three-quarters of large organizations are increasing their investments in talent
analytics this year, and CEOs and boards are giving increasing attention to
analytics’ potential to improve talent management’s impact.
For the first two issues of Talent Analytics Quarterly, we’ve approached these
changes from a variety of perspectives. In our first issue, we focused on
foundations—interviews with talent analytics leaders who were building their
functions and creating their organizations’ talent analytics strategies, advice on
how to build foundational data management capabilities, and hiring guidance
for building a successful talent analytics team. In our second issue, we focused
on the future, with specific features on how to set priorities for 2017, key
trends and influences on talent analytics organizations, and how to manage
the emerging trend of organizations working with both chief data officers and
talent analytics groups.
In our third issue of Talent Analytics Quarterly—and the first in 2017—we dive
deep. In the pages that follow, readers will find an excellent primer from CEB’s
Talent Management Labs on using text analytics, an article on the use of
machine learning in candidate selection, an early look at major trends in talent
analytics technologies and innovations, and a look at how we can draw lessons
from marketing analytics on how to communicate data-centric insight to
clients—an important and perennial issue for talent analytics leaders as well.
And of course, we continue our well-received tradition of bringing you

interviews from your peers, featuring conversations with Christian Cormack
and Blair Hopkins, talent analytics leaders at AstraZeneca and EY, respectively.
As always, we welcome your feedback and suggestions for how we can

improve this publication for you and your teams. Please direct any comments
or questions to [email protected].
Sincerely,
Daniel Metz
Research Leader
CEB Corporate Leadership CouncilTM
Q1 2017 3
What Are Your 2017
Talent Analytics
Priorities?
Just
1 in 10
talent analytics leaders believes his
or her organization effectively uses
talent data to inform decisions.
4 Talent Analytics Quarterly © 2017 CEB. All rights reserved.

Most plan to increase investments in talent analytics for 3 reasons:
2.
In response to 3.
1.
sponsorship or To address
To address key
demand from specific business
talent risks
senior HR challenges
leaders
Where are they investing in 2017?
Functional
Specific Projects Maturity Data Quality
Strategic workforce Staff costs are the Inadequate data
planning and attrition biggest expenditure quality inhibits over
projects are top for talent analytics 50% of organizations
priorities for budgets from using talent
2017. in 2017. data.
What projects are they working on in 2017?
65.5% 61.1%
60.2% 58.4% 55.8% 52.2%
Strategic Attrition/ Engagement Diversity Talent Leadership

Workforce Retention and Attraction/ and
Planning/ Inclusion Acquisition Succession
Talent Planning
Forecasting
Access our latest insights on our Talent Analytics Portal

today to support your 2017 initiatives.
Q1 2017 5
What We Heard
from You in 2016
By Marianne Stengel
A
s we built our support for heads of • The idea of sharing imperfect data to help leaders
talent analytics and their teams over realize they need to hold data owners accountable for
data quality came up a few times in our discussions. For
the past six months, we spoke to many
example, one organization took this a step further by
of you around the globe. We heard four primary deciding to formally measure and report data quality. The
challenges—and a few solutions—from you. organization established a data quality score, published it
in a quarterly dashboard, and pushed accountability for it
1. Engage users in improving data quality.
to data owners (HRBPs or business leaders).
Data quality came up in nearly every conversation. For Creating transparency on data quality weak spots across
54% of organizations participating in our recent survey, it the organization pushed data owners to improve it. The
is the top barrier to talent analytics effectiveness, and it talent analytics team further supported data owners by
often appears insurmountable. However, we heard a few compiling a data and metrics dictionary.
encouraging stories of how far you can get with less-than-
perfect data and how you can partner with talent analytics
clients to improve it:
• One organization built a simple attrition model starting Takeaway
with publicly available data from consultants on the
possible cost savings from reducing turnover. Combining Don’t let poor data quality hold you back. Use
this information with its own incomplete and unreliable the data you have anyway to get a reaction from
data on turnover, the organization built a forecast model data owners, and share a few simple steps with
and let business leaders tinker with different assumptions. them to get them started on improving it. See
a replay of our event, Three Critical Obstacles
While imperfect, this model showed leaders that even to Creating Talent Analytics Capability, on the
with turnover rates remaining the same, current diversity CEB Corporate Leadership Council™ website
hiring strategies were not going to achieve the desired for more tips.
goal. And more importantly, it started a conversation
about what needed to be done to improve data quality
over time.

2. Take ownership of data privacy. • Ownership of employee data privacy is increasingly
shifting from IT to HR, which is looked to as an
Talent analytics leaders are generally well aware of their advocate for employees’ interests. This requires regular
responsibilities to protect employees’ information from meetings with Legal and a good understanding of the
unauthorized access and to respect employees’ wishes organization’s information needs. Some talent analytics
concerning how their data will be used. In the past, data teams designate one team member as data security and
privacy was seen as a predominantly European issue privacy lead to make decisions about how talent data
(due to work council involvement and now the EU GDPR may be used.
requirements), but it is now firmly in the spotlight worldwide.
We have heard of a few strategies companies have taken

to reassure employees and protect themselves from legal
or reputational risk:
Takeaway
• Several talent analytics leaders have made data privacy Keep data privacy top of mind. After viewing
the centerpiece of their strategy, above even technology the replay of Three Critical Obstacles to Creating
or team skills development. For one company, this takes Talent Analytics Capability, consider who on
the shape of a formal commitment to only use employee your team could take the lead on data security
data for purposes that employees would expect and privacy and how they should evaluate risks
(e.g., analyzing diversity drivers but not predicting against potential data use benefits.
individual attrition likelihood). This builds employees’
trust in talent analytics and increases the team’s access
to employee data.
• Building trust in talent analytics is particularly important. 3. Focus on HR staff’s analytics skills to drive talent
One talent analytics leader described to us how he analytics impact.
tries to build the image of talent analytics as a neutral
Most talent analytics leaders report frustration with their
consultant; for example, he involves universities—which
peers in the wider HR function when it comes to analytics
are only interested in knowledge—in projects, and he
understanding or skills. Too often, HRBPs struggle to make
posts the results of each project for everyone to review.
Q1 2017 7
sense of analytics insights and 4. Align talent analytics
therefore don’t use them with with business needs and
their line leaders to initiate capability to drive impact.
change. We have heard many The team asks the
Many organizations mentioned
different strategies to improve requestor to consider the
HRBPs’ or specialists’ skills and the need to prove the impact
comfort level with analytics:
questions: What do you of talent analytics. This is
think the answer is going easier to do in cases such
• While many organizations as attrition projects where
push e-learnings or webinars, to be? If I tell you the
there are direct links to cost
the most effective training answer, what will you do savings. It is harder, however,
focuses on in-person, practical
about or with it? Over time, to show how concepts such
exercises. Two organiza-
as engagement or leadership
tions described to us how this teaches requestors to
make the business more
they organized traditional, ask better questions. profitable—or at least a better
classroom-based training
place to work.
sessions in which HRBPs work
on real-life questions from the Talent analytics leaders have
HR leadership team. This gives found different ways of tackling
them hands-on experience of this challenge:
simple modeling and teaches them how to ask good
• Several talent analytics teams rely on close partnerships
questions. It also builds their confidence not only with
across the organization to identify where to focus their
using data but also with presenting and defending it to
time. By getting the business to tell them its problems,
their line leaders.
they ensure they are aligned to business priorities and
• One talent analytics leader who hasn’t been able to fund that analytics insights will be acted on. This in turn gets
this type of training instead ensures that he and his team them invited back for other projects.
evolve from simply delivering answers to teaching their
• One talent analytics leader didn’t wait for the business to
clients how to self-serve. Whenever a request comes in
tell him what it needed. Observing a disruptive change
for which this is feasible, they consult with the requestor
in the industry, he decided to have his team develop a
one-on-one to show them how to think through getting
workforce planning tool. When the business realized
the answer and teach them simple analytics skills.
what it needed, the tool was close to ready, and it is now
• This is echoed by another organization where the talent being rolled out across the organization.
analytics team forces requestors to state what they are
• Several organizations have stressed the harm in getting
going to do with the data they are asking for. The team
too sophisticated too early. One team aims to deliver the
asks the requestor to consider the questions: What do
analytics quality and insight that is required plus a bit
you think the answer is going to be? If I tell you the
more; anything beyond that risks leaving the business
answer, what will you do about or with it? Over time, this
behind and not getting used. It is more important to focus
teaches requestors to ask better questions; and in the
on application rather than on sophisticated analytics.
interim, it can drive requestors to realize when they can
answer their own questions, or that the question might • Another organization names a business sponsor for every
not be worth asking in the first place. strategic analytics project. This ensures that the business
is involved in the work instead of just receiving the
results. Talent analytics teams have to recognize that the
answers they provide require change, and this change is
Takeaway easier to execute when the business has been engaged in
the search for the right answer.
Focus on building analytics skills in every
interaction with HR data users. Start by sharing
the HRBP Guide: Deliver on Talent Analytics Takeaway
Initiatives on the CEB Corporate Leadership
Council™ website with interested HRBPs to help Don’t overwhelm your clients, but engage them in
them identify when to apply talent analytics, defining the key questions to answer with talent
how to select the right metrics, and how to analytics. See our case study of Gap Inc.’s method,
communicate talent insights. Analytics Prioritization Principles, on the CEB
Corporate Leadership Council website.

Aligning Talent
Analytics to CHROs’
Strategic Priorities
By Danielle Douglas
C
HROs know that talent analytics can play Using metrics to create the illusion of change is easy. But
a big role in achieving their goals. That’s behavior change is what truly defines a culture shift, not
the metrics organizations typically focus on (e.g., interim
why 68% of CHROs want to improve
objectives, improvement in business results, employee
their organizations’ talent analytics capabilities attitudes). Talent analytics should define metrics up front to
in 2017. To help talent analytics leaders like be able to identify whether employees are actually working
you prepare for 2017, we have taken the top in new ways.
insights from our global survey of heads of HR
2. Strengthen cross-functional analytics partnerships.
and identified five opportunities to get ahead
of potential CHRO needs. CHROs spend nearly 25% of their time on enterprise-wide ini-
tiatives, necessitating strong cross-functional partnerships.
1. Create a line of sight between culture
CHROs reported that the chief financial officer (CFO), chief
and strategy via metrics.
information officer (CIO), and chief
Culture goes hand-in-hand with strategy officer (CSO) are the most
business transformation. So given the important C-suite peers to partner with
increasing number of organizational as they look to bring attention to tal-
68%
changes, it is not surprising that 76% ent issues in new ways in 2017. For ex-
of organizations anticipate undergoing ample, CHROs can partner with these
a culture change in 2017. leaders to redesign processes that may
of CHROs want to causes managers to redirect resources
CHROs are confident in HR’s improve their talent away from large growth projects, build
analytics capabilities
ability to define a new culture, in 2017. a robust digital workforce pipeline, and
but confidence decreases by 30% create a change-capable HR function.
when it comes to embedding the
new culture. In fact, only 37% of Knowing that the head of HR will
CHROs are currently confident in HR’s ability to define need to partner with Finance, IT, and
metrics and collect data on culture-change progress— strategy leaders, talent analytics teams can (and should)
one of the more challenging aspects of embedding a begin to form their own partnerships with their counterparts
new culture. in these areas. These relationships will give the talent
Q1 2017 9
analytics team an advantage if it is asked to support cross- 1. Improving business acumen to ensure the team
functional projects because the team will know whom to understands the key priorities of the business
reach out to for different data. In addition, these relationships 2. Improving the data presentation (i.e., translating
will certainly prove beneficial to the talent analytics team for complex analysis into a consumable format) to engage
best practice sharing, data access, tools, technology ideas, the board and push leaders to action.
and more.
5. Apply leader analytics to strengthen
3. Analyze employee needs to create succession strategy.
consumer-like experiences.
Key business stakeholders are beginning to question current
External, macro trends such as the rise of big data, on- succession management approaches, and boards seem
demand service expectations, and digital disruption are
dissatisfied, asking about the succession process, current
prompting CHROs to look internally for ways to transform bench strength, and overall pipeline sustainability. It is
the HR function. One-third of heads of HR are planning to
therefore not surprising that executive succession is the top
innovate by creating a consumer-like employee experience.
talent issue on which heads of HR feel pressure from the
This transformation includes applying product management board to provide more insight. At the same time, heads of HR
discipline to HR offerings (e.g., Agile are questioning traditional succession
methods), building HR apps, and practices, such as the nine-box grid
gamifying talent management. and talent reviews.
67%
For HR to move away from the The number and depth of these
proficiency and performance measures succession doubts indicate that CHROs
it is often preoccupied with and may be looking for new-in-kind ways to
become more consumer focused, it of CHROs think they
must better communicate supplement, or even replace, existing
must draw inspiration from functions
talent management succession management practices.
that concentrate on external customers insights to the board. To get ahead of this, talent analytics
(e.g., Marketing and Sales). The first leaders should focus on leader
step to an effective consumer-like analytics—a systematic, data-driven
strategy is to understand and segment approach for identifying the drivers
your audience. and outcomes of effective leadership.
(See Seven Lessons for Impactful Leader Analytics on the
Talent analytics teams can help HR begin to understand
CEB Corporate Leadership Council™ website for more
unique employee needs by analyzing data, such as life stages
information.)
and context, that influence employee behaviors, motivations,
and actions. The goal is to build a demand landscape for Using past data on effective leadership to predict future
employees as a means of better understanding the workforce, performance, leader analytics can help organizations with
similar to what occurs in the consumer space. succession planning by more effectively identifying essential
leadership qualities and by identifying leaders earlier and
Read our interview online to learn how Hershey identified
from unexpected places.
nontraditional employee segments.
4. Prioritize business acumen and

data visualization skills.
Concluding Thoughts
As CHROs and HR continue to cement their role as a strategic
CHROs’ priorities in 2017 have important
partner to the business, the CHRO’s relationship with the
implications for talent analytics leaders and
board is more critical than ever. Yet nearly half of heads of
professionals. Taking advantage of the above
HR don’t think the board has a good enough understanding
opportunities gives talent analytics teams the
of key talent issues. To improve this, 67% of CHROs think
chance to get ahead of likely demands from
they must better communicate talent management insights
CHROs and other senior leaders and better align
to the board.
talent analytics’ strategy to the organization’s
With a new and even more influential audience to present emerging priorities.
talent insights to, data that was previously unnecessary may
quickly become imperative. Knowing that the board’s goal is
to minimize risk and maximize opportunity for the business,
talent analytics teams must go beyond merely providing data
on the overall state of talent. Instead, they should focus on
enhancing their data-sharing skills in two ways:

Starting
Simple at EY
An Interview with
Blair Hopkins
EY | Assurance | Tax | Transactions | Advisory About Blair Hopkins
EY is a global leader in assurance, tax, transaction and Blair Hopkins is the Director
advisory services. The insights and quality services it of Global People Data,
delivers help build trust and confidence in the capital Reporting and Analytics
markets and in economies the world over. EY develops based in Glasgow, Scotland,
outstanding leaders who team up to deliver on its and has been with EY for
promises to all of its stakeholders. In so doing, it plays more than 10 years. The
a critical role in building a better working world for its objective of the Global
people, for its clients, and for its communities. People Data, Reporting
and Analytics team is to
EY refers to the global organization, and may refer to improve the control, quality,
one or more of the member firms of Ernst & Young and efficiency of people data, reporting, and analytics
Global Limited, each of which is a separate legal entity. to allow better decision making for leaders across
Ernst & Young Global Limited, a UK company limited the EY global Talent function and across the broader
by guarantee, does not provide services to clients. organization.
For more information about EY, please visit ey.com.
Blair oversees people data governance, helping confirm
data standards and policies are met for talent data
across EY member firms in more than 150 countries,
and the Global People Reporting team, comprising
a globally distributed center of excellence and an
India-based shared services reporting team. He is also
accountable for the establishment and management of
an advanced Analytics team focused on strategic HR
issues impacting EY globally.
Q1 2017 11
E
very quarter, we interview talent analytics of data management, from data definitions and the
leaders to gain their perspectives on processes that gather data through to guidance on how
to best leverage our data for analytics while respecting
issues facing them and their teams. This
the intent of the rules and regulations.
quarter we spoke with Blair Hopkins about his
work at EY. After we’ve set the rules, our Global People Reporting
team acts as a gatekeeper and applies these rules to data
requests. It’s an approach that works because we have
data governance as a foundation for our way of working.
When you created the Global People Data, Reporting
and Analytics organization at EY three years ago, what Let’s go back to how you generate insights from data.
did you decide to invest in first? You mentioned competency with Excel—what other
skills does an analytics team need?
BLAIR: Aside from hiring the right people—initially good
storytellers—my focus was to rapidly demonstrate the BLAIR: In the early stages, we recruited storytellers
business value of the new function to our leadership, then rather than data scientists. In our case, our first projects
focus on data governance. Technology investments came wouldn’t have been sophisticated enough to keep data
later. Decisions made on our findings impact real people scientists busy, and that’s an important consideration.
in real terms, hence the focus on data. With technology We concentrated on low-hanging fruit in order to build a
investment, too much analytical sophistication early business case for talent analytics and why it would pay off
on could have run the risk of overwhelming our internal to invest more in our team. In order to do that, we needed
clients, as advanced methods and technology can make people who were technically clever (e.g., modeling skills,
analytics seem unapproachable. Microsoft Excel) and could also tell a compelling story.
We were able to communicate a high-impact, simple, and As general advice, I would recommend to start simply
effective message through having a highly competent before pursuing more elegant analytics solutions that
user of spreadsheets generate insight from Microsoft can be difficult to explain and tie into leaders’ objectives.
Excel. When you consider many HR generalists have a Again, do not overcomplicate things at first. As we have
limited understanding of statistics, there was no need to matured, we have been able to add deep statistical
immediately dive deeply, using more sophisticated tools knowledge, psychology, modeling, and a dedicated data
like SPSS or R. scientist to manage data and technology environments. Of
note, while some of the team has HR domain experience,
Why do you think data governance should be at the we actually seek deep HR domain knowledge from our
center of your analytics strategy? clients on a project-by-project basis. While desirable,
this is not an essential hiring skill and can be developed
BLAIR: I was primarily driven by concerns with respect
on the job.
to data privacy, which carries a much greater risk when
dealing with people data compared to market, consumer, Can you give us an example of what one of those early
or financial data. The conventional starting point for projects looked like?
any analytics project is “Give me all the data, and let’s
see what we can do with it.” The challenge is that you BLAIR: One of our first projects focused on retention. We
are often limited in taking this approach, as regulations spent only a week on our initial exploratory analysis, but
covering collection of personal data often require explicit we came up with a bottom-line impact figure in excess
permission—and that can be hard to obtain. of US$100 million that could be derived from just a small
improvement in retention. That’s a compelling story to tell
There are also considerations around how data is managed. based on fairly simple, yet robust analysis.
For example, an analyst might send out analytics results,
hiding an employee list in the spreadsheet but leaving the Our global Talent vice chair took this to the global
data potentially accessible. In a lot of countries, this might executive team, leading to the initiation of a project across
constitute a breach of employee data privacy regulation, our organization—mobilizing hundreds of HR business
leaving our executives exposed to regulatory risk. partners and organization development leaders—as well
as the central development of more sophisticated tools to
How are you addressing these risks? support regular analysis.
BLAIR: I currently chair our Global People Data So storytelling skills combined with an aptitude for
Governance committee, which identifies what the numbers and an eye for business context took us farther
boundaries are and confirms we have common definitions. than if we had solely pursued data knowledge within our
We have membership drawn from our data privacy team, team. Given the success of the retention project and our
as well as IT and the business. We focus on all aspects ability to make the case that we could have tremendous

impact, I was given the resources to hire additional people year to identify how different parts of the business might
into our team. apply such tools, what data we needed to drive it, how to
transform HR data to match with financial data, etc.
This is something a lot of talent analytics leaders
struggle with, getting sufficient resources for their team. We are now utilizing and deploying a number of models
After your initial success, how were you able to secure to understand the impact at both the strategic and
the resources you needed? operational levels, a truly collaborative effort across our
business functions.
BLAIR: We certainly often
struggle with the number of
requests we receive—in some
ways, a victim of our success.
Negotiating and setting
expectations are the best ways
The important thing was to offer a way forward,
to get what we need to handle not to say “no” in the first instance due to lack of
the influx. resources. Managing expectations on time is also
There are a few options to a way to ensure progress but not necessarily add
consider when someone comes cost. Most clients are perfectly happy to wait a
to us with a request. First, if the
month or two if they know they will get the support
request involves only one country
or impacts a small number of they need.
people, we’ll always try to give
some advice but may not take on
a full project. We’ll also explore
whether they might have funding,
allowing us to contract resources for a given time. The What is the role of the talent analytics team
bottom line is a question of business value; demonstrating in workforce planning going forward?
the ROI in commercial terms is the best way to
secure resources. BLAIR: We provide the support—methodologies,
knowledge, code, and infrastructure. We work closely with
And how do you convince someone to give you their the reporting team to automate the data feeds into the
money and resources? tool. We are trying to drive a consistent approach for how
everyone approaches workforce planning and are building
BLAIR: Always be prepared to ask and negotiate. For a toolkit and a network to get the business functions to
example, we have a significant workforce planning project talk to each other and share their ideas.
underway, and one of our larger countries needed help.
We said it would take six resources to support, with This all helps drive short- and medium-term operational
additional funding for shared services support. The decisions around recruiting efforts and learning
requestor needed this completed in three months; we requirements, even assisting with revenue planning as well
countered with six months. Our client secured funding, as addressing more strategic questions. The necessary
allowing us to engage and provide service. change management is delivered at a local level with our
close support.
The important thing was to offer a way forward, not to
say “no” in the first instance due to lack of resources. The views reflected in this article are the views of the author
Managing expectations on time is also a way to ensure and do not necessarily reflect the views of the global EY
progress but not necessarily add cost. Most clients are organization or its member firms.
perfectly happy to wait a month or two if they know they
will get the support they need.
Can you tell us more about the workforce

planning project?
BLAIR: Like most sectors, ours is facing disruptive change

driven by megatrends, such as digital, robotics, the gig
economy, etc. We foresaw the need to address these
trends as the business leadership started mobilizing, and
my team started building workforce planning models
in collaboration with one of our regions. This gave us a
Q1 2017 13
Talent Analytics
Innovations: Separating
Fact from Hype
By Blakeley Hartfelder
O
rganizations today face
a rapidly expanding
marketplace of emerg- HRIS predictive Nearly half of organizations do not use their HRIS for
analytics predictive analytics, and only 30% of those that do
ing technologies and vendors capabilities are report that the features provide high value. As a result,
in the talent analytics space— insufficient. these organizations may be turning to other platforms
offering everything from data to provide analytics support, unnecessarily increasing
coordination and data compatibility costs.
aggregation to attrition predic-
tion and social network anal-
The importance
ysis. Although these vendors of data More than half of organizations use data visualization
software for talent analytics, and another one-third are
promise improvements, each visualization
drives software planning to use it soon. Of users, about 70% report
requires substantial investment receiving high value from the software and will increase
investments.
of your time and resources. So investment.
how do you choose?

Workforce All talent analytics teams that currently use advanced
To separate fact from hype, we
planning analytics services for workforce planning report they
conducted a survey to objectively technology is will maintain or increase their levels of investment. This
map the trends and future importance becoming more is likely due to the technology’s current impact—about
of over 50 established and emerging popular. 70% find it delivers high value.
talent analytics innovations and
technologies. We’ve begun to
analyze the results, and the figure to Wearable Only 25% of organizations currently use or plan
the right summarizes what we have technology is to use data from wearable technology for talent
learned so far. failing to win analytics. Interestingly, that number may soon
over talent decrease, as one-third of current or potential
analytics users will decrease their level of investment in the
leaders. near future.
Organizations
plan increased Few organizations currently use consultants to conduct
investment one-off analytics projects, but the number may increase.
in analytics About 80% of organizations think these services provide
consulting. high value, and nearly half plan to increase investments
in this area.
Source: CEB 2017 Talent Analytics Innovation Survey.

CEB HR Analytics
Leadership Academy
Develop your team’s skills to move
from talent reporting to talent analytics.
Just 8% of senior HR leaders

believe they are getting
significant returns on their
analytics investments.
Learn how the CEB HR Analytics Leadership

Academy can build your HR staff’s ability to
provide actionable insights based on data to
improve business decisions.
Learn More
http://ceburl.com/1qwc Q1 2017 15
Beginner’s Guide
to Text Analytics
for HR
By Andrea Kropp
W
ords underpin nearly every aspect would prove a difficult starting point for teams delving into
of business—from e-mails between text analytics for the first time because they require multiple
text processing and analytical techniques. A good rule of
team members and phone-based
thumb is to scope your initial text mining projects so that
customer service requests to live sales visits
each has no more than two required steps or methods your
with prospects and résumés submitted by team has never encountered before. This is especially true
candidates. Add to that list the more-structured if the team is learning the software and methods at the
final reports and presentations stored on your same time.
company’s network and the less-structured
This article details three foundational capabilities and some
instant messages and breakroom chatter.
related projects you can use as starting points in your text
When you begin to think of words as data, new analytical analytics journey. As your team masters these capabilities,
possibilities emerge across all the major HR domains. For you can build up to more complex analyses by chaining the
example, you may wonder if it’s possible to: basic methods together. In this way, you can take incremental
steps toward big ideas.
• Monitor employees’ engagement levels passively
via their language, A Note on Software Choices and Finding Help
• Predict how well a person will perform in a role from
their interview responses, This article will not list or contrast the many text-handling
and text-analysis software choices available. The suite of
• Identify future leaders from the content and style
open-source and proprietary tools to facilitate text mining
of their communications,
is growing every month. For most common tasks, excellent
• Assess how well a team is working together from their written and video tutorials are available, and many include
exchanges during a meeting, or sample code created by a very active user community.
• Create tailored coaching for individual sales
One consideration is that text mining cuts across nearly all
representatives based on transcripts of their customer
academic fields, and it is not uncommon to jump between
interactions?
resources created by computer scientists, linguists,
These types of big ideas will help talent analytics leaders historians, political scientists, psychologists, and more when
make the business case for seeding and nurturing text researching a new method. Don’t expect or try to find many
mining competencies within their teams. However, these HR-specific examples.

Unless you have a compelling reason to begin elsewhere,
we’d recommend starting your text analytics journey with
Python (or R as a second choice). It may take longer to get
results from your first project than it would using software An Example of Tokenization
that has a graphical user interface and point-and-click The Human Resources team is fantastic.
menus, but in the long run, Python will be more flexible and • Unigrams: fantastic (1), human (1), is (1),
powerful. This will make it easy for your IT team to implement resources (1), team (1), the (1), . (1)
the findings of your work into production systems.
• Bigrams: human resources (1), is
Capability 1: Counting Words and Phrase Frequencies fantastic (1), the human (1), resources
team (1), team is (1)
What and Why: The ability to count how often certain words
Source: CEB analysis.
or phrases are used is a prerequisite for nearly every other
text mining activity. Combining word counting with some
basic statistical techniques such as t-tests and regression
equations will unlock many possible questions to answer,
Start by counting words in a single document, and move on
such as “How do the concerns of junior versus senior staff
to creating frequency counts across multiple documents.
compare on engagement surveys?” or “What word and
phrase combinations on résumés correlate with future Relate your normalized counts to other data you have about
success in sales roles?” the authors, such as their age, job role, or performance.
How: The methods you’ll need to apply are tokenizing and Examples: The MyPersonality Project, which relates over four
n-gram counting. Tokenization is the process of breaking million Facebook profiles to over six million psychometrics
a passage of text into words, phrases, symbols, or other tests, has resulted in the publication of hundreds of papers
meaningful elements, called tokens. The tokens can then be and interactive applications that show how word and phrase
counted and reported as single words (unigrams), two-word
use relate to demographics, personality, and mental and
phrases (bigrams), three-word phrases (trigrams), and so on.
physical health (see figure 1). Explore the demos available
The result will be dominated by functional words such as
on the World Well-Being Project website (wwbp.org/demos.
“and,” “in,” or “the,” which can be removed by specifying a
list of stop words. Experiment with removing all punctuation html), and try the four ideas in the project suggestions table
before tokenizing; this will cause “e-mail” and “email” and at the end of this article for some quick wins using only this
“Ph.D.” and “PhD” to become the same token. first capability.
Q1 2017 17
Figure 1: MyPersonality Project Words, by Age excited
:)
i am thankful
Standardized Relative Frequency
3
wonderful
2
(1)
(2)
(3)
10 15 20 25 30 35 40 45 50 55 60 65
Age
Source: University of Pennsylvania Positive Psychology Center: The World Well-Being Project, “Words Across Age,” 2017,
http://wwbp.org/age_plot.html.
Capability 2: Identify the Main Topics and Themes
What and Why: Substantive text analytics requires moving

beyond the words people choose to the ideas they wish Figure 2: Topic Modeling Example
to express. Effective text analytics methods must include
an understanding of semantics—the meaning of the text. Topic 1: Topic 2:
Organize Mercury
This includes negations, questions, sarcasm, analogies, Lead Pollute
Committee Arsenic
metaphors, rhetorical statements, and other literary devices Direct Lead
used in oral and written communications. The natural
language semantics tools that exist now aren’t very robust
or complete, but progress is rapid, and the most essential
component in the toolkit—the ability to find topics—is in
Leadership Committee Heavy
widespread use. Today Structure Metal
at the EPA Poisoning
How: A rudimentary way of moving from words to topics
is stemming. Stemming is the process of reducing inflected Source: Ted Underwood, “Topic Modeling Made Just
words to their word stem or root form. For example, the Simple Enough,” 7 April 2012, The Stone and the
Shell, https://tedunderwood.com/2012/04/07/topic-
individual tokens “manage,” “manages,” and “managing” are
modeling-made-just-simple-enough/.
collapsed to the common root “manage.” However, many
English words have multiple meanings. The word “lead” (the
heavy metal) shouldn’t be counted with “lead” (the act of
leading). Conversely, “wages” and “pay” are conceptually
related but don’t share a common word stem. Topic modeling Topic modeling is a precursor to any document classification
is able to handle both these challenges. problem. Take, for example, comment classification from
an anonymous review website such as Glassdoor. First, the
Topic modeling is an exploratory method used to find the analytics team conducts an exploratory topic modeling
major recurring themes in a collection of text (see figure 2). exercise to identify the topics and elicit the words that
A “topic” consists of a cluster of words that frequently occur define the topic. Then, each comment is compared to the
together. The researcher must specify the number of topics defined topics for classification. The researcher can define
to be extracted (e.g., 10 to 100), and it is a good idea to test whether the comment should be forced into a single topic
multiple values. If the researcher believes the resulting topics or can belong to multiple topics. In this way, organizations
have too much overlap, a principle component analysis or can monitor comments over time to see which topics are
factor analysis can be used to reduce them further. rising in frequency and which are falling.

Examples: We’ve seen very simple topic models applied to 3. A final category of features comes from merging
employee survey comments generate extremely actionable your data with outside contextual data. For example,
insights for HR executives. Such insights might otherwise from the name of the college a person attended, you
have been buried because they were only mentioned by can derive features such as school size, urban versus
about 1% of tens of thousands of employees. rural location, and average SAT score using a merge
table from a source such as the National Center for
One topic model we ran found a collection of comments Education Statistics.
pertaining to a shift from semiannual to quarterly
performance reviews. Employees disliked the change, The ability to derive such additional features is vital in all
claiming there wasn’t enough time between administrative predictive analytics projects. Features generated in this
cycles to make meaningful changes in performance. Another way will be more reliable, more generalizable, and fairer
than individual n-grams and will greatly improve your
topic model revealed that a required health questionnaire for
model performance.
a new insurance provider contained very personal questions,
which employees resented. How: Standard transformations, such as identifying parts of
speech or calculating reading grade levels, can be readily
To put this in perspective, in an organization with 50,000
accomplished with open-source tools. Look for Python or
people, 1% equates to 500 people. If 500 people choose
R packages purpose-built for each task.
to write about the same policy change when given the
opportunity to write about anything in an open comment Many free and paid word lists are available for the application
field, that policy change deserves a closer look. Topic of word dictionaries. The most-cited free list is probably
modeling is really the only way to bring
these insights to the surface.
Capability 3: Derive Additional

Contextual Features from Text Substantive text analytics requires moving
What and Why: The final capability beyond the words people choose to the ideas
is deriving additional features from they wish to express.
your text by applying custom and off-
the-shelf rules and transformations.
Known as “feature generation” in
machine learning circles, the goal is to
view the text through many different WordNet, developed by Princeton researchers, and the most-
lenses. There are three basic categories of derived features cited paid list is probably the Linguistic Inquiry and Word
to consider, which we’ll call standard transformations, word Count (LIWC) tool. You can also develop your own lists from
dictionaries, and outside context. scratch or modify an existing list to better suit your needs.
Most organizations will need to develop custom lists for
1. Standard transformations you might want to apply to items such as “senior leadership team names,” “products
your text include part-of-speech tagging, readability we sell,” and “IT systems we use” so they can be grouped
scoring, and spelling- or grammar-error detection. together as a frequency count.
These procedures will generate features such as
percentage pronoun usage, Fleish-Kincaid score, Outside contextual data can come from many sources. You
and spelling errors per 1,000 words. For example, might want to merge in industry classifications based on
perceptions of group identity within a team are subtly company names, news events based on dates, or regional
encoded in the pronouns team members use, where housing prices and unemployment figures based on city
plural versions (“we”, “us”, “them”) signal that the names. Always look for an open-source resource as a starting
person feels like part of the group.1 point, then modify the tool to suit your needs, and make it
available to others in your organization. It’s a good idea to
2. Another feature generation method is the application start a central repository for reference tables and reusable
of word dictionaries. Other researchers have already code and to implement a version control system.
compiled extensive word lists for “quantifiers,”
Example: Suppose you wish to identify which attributes of
“tentativeness,” or “positive emotions,” which you can
a résumé correlate to attrition prior to six months on the
apply to your data to get word counts by category.
job. To find out, you would obtain the original applications
Unlike topic modeling where the aim is identify the
of all your hires and separate them into two groups—those
content of the text, word dictionaries such as these
reveal how a person organizes and expresses their James Pennebaker, The Secret Life of Pronouns, (New York: Bloomsbury
1
ideas and how they feel about the topic. Press, 2011) 229–231.
Q1 2017 19
who stayed past six months and those who didn’t stay for
six months. Then you’d use the application data to generate
as many features as possible. Your feature set would include
n-grams (capability 1), topics (capability 2), and derived
features from all three categories (capability 3). It would
not be uncommon to have 5,000 or more features to work
with at this point.
The more rational transformations and features you create

before you begin modeling, the better your results will be.
Your next steps would be feature selection (reducing the
5,000 or more to a manageable amount of truly important
predictors) and then model development.
Testimonial: The Joy of the First Word Cloud
Allow me to close with a personal story that exemplifies the There is much to learn,
profound impact text analytics can have on your analysts and
organization. A few months ago, I participated in a corporate
and these building blocks
event where teams of four people had roughly three days are just the starting point.
to develop a working prototype of a big data product. The
participants were mostly midcareer professional IT staff,
data scientists, and computer scientists, but many had never
worked with text as data. Around 10:00 p.m. on day two,
I was roaming the halls to stretch my legs when I heard a
woman from another team exclaim in an exhilarated voice,
“It worked! I made a word cloud! Come look.”
Producing a single word cloud at the end of day two meant

this team was far behind, but the woman’s excitement
was undeniable. The sheer joy in her voice transported me
back to the day I made my first word cloud and the pure
excitement I felt. Although I’ve been analyzing numbers for
most of my professional career, there was something special
about analyzing words for the first time.
There is much to learn, and these building blocks are just

the starting point. Keep an eye out for the person in your
organization who shouts the loudest about making his or her
first word cloud. Give this person a sequence of cumulative
text mining projects, and you’ll be well on your way to
answering some of HR’s most tantalizing questions.
About CEB Talent Management Labs
This article is contributed by our Talent Management

Labs, a team dedicated to helping CEB and its
customers lead the analytic transformation of talent
management. The Labs’ scientists conduct research to
ensure our talent management products and services
are built using the most advanced thinking from a
broad range of established and emerging disciplines.
For more information on CEB or the

Talent Management Labs, please contact
[email protected].

Project Suggestions
Building Block Do This First to Learn Do This Later to Build on the Work
Count Word Count words and phrases in a single open-ended Examine how responses differ by employee segments
and Phrase employee survey field with a variety of responses. such as level, location, or ethnicity.
Frequencies
Count words and phrases in the course comments for Provide feedback to individual instructors.
an internal training course that is taught by multiple
instructors.
Count words and phrases in your organization’s reviews Monitor changes over time.
on Glassdoor or a similar website.
Count words and phrases in the hiring manager’s notes Correlate success in the role with the manager’s notes.
on candidate interviews for a specific role.
Identify the Apply stemming to your text before word counting to In future projects, try stemmed and un-stemmed text to
Main Topics group words with the same root (e.g., nurse, nursing, see which yields the better result.
and Themes nurses, nursed).
Create a word cloud of all job postings from the past Limit your data to a single job title or category, and
year after removing stop words. compare your historical job postings with those of your
main talent competitors.
Create a topic model of your organization’s reviews on Examine topic frequency by job title and location as
Glassdoor or a similar website. (Tip: Keep the pros, cons, reported on Glassdoor.
and advice to management separate.)
Apply an open-source sentiment dictionary to open- Manually read and tag the comments as positive,
ended comments from internal surveys. negative, or neutral, and assess how accurate your
chosen sentiment tagger was. Learn how to make it
more accurate for your own use cases.
Derive Use a part-of-speech tagger to calculate what Correlate stylistic language features to outcomes of
Additional percentage of a speaker’s or writer’s words are verbs, interest. For example, there is substantial literature on
Contextual adjectives, proper nouns, etc. the relationship between personal pronoun use and
Features leadership.
from Text
Use the name of the college listed on a résumé to link Look for patterns among successful and unsuccessful
to an external table of data about the institution (e.g., hires in what type of college they attended.
urban or rural location, public or private, average
SAT score).
Apply a word categorization dictionary such as WordNet Use linguistic word categories to infer psychometric
(open source) or LIWC (fee applies) to a collection of traits of the author, such as personality or
documents, e-mails, or comments. cognitive ability.
Apply named-entity recognition (NER) to employee Aggregate all the survey responses that pertain to a
survey comments to identify the people, places, certain person or department. (NER is more powerful
departments, programs, etc. most often referred than a keyword search because it will find instances
to by name. where pronouns refer back to the same entity.)
Constructs and Outcomes Revealed • Personality

from Language • Mood
• Cognitive ability
Peer-reviewed research studies have • Mental health
linked individual language use to all the • Physical health
following categories. • Social dynamics and group cohesion
Q1 2017 21
The Advantages of
Machine Learning in
Candidate Selection
(and Pitfalls to Avoid)
By Marianne Stengel and Omar Hoda
W
hile many talent analytics teams This automated prediction replicates what the human mind
are still struggling with cleaning attempts to accomplish when selecting from the applicant
pool, but at a much higher level of sophistication.
data or getting buy-in to invest
in data warehouses, some talent analytics The Benefits of Machine Learning
leaders are turning their attention to machine for Candidate Selection
learning. Here, we consider machine learning’s Several decades ago, organizations made the move away
application to candidate selection, outlining its from relying entirely on human judgment to administering
possible benefits as well as its potential risks. predictive tools such as psychometric assessments before
hiring a candidate. Psychometric assessments allow
What Is Machine Learning? companies to understand a candidate’s competencies and
potential success in the role and are more objective than
Machine learning is the automated version of human learning.
relying on a hiring manager’s reaction alone (e.g., through
It mimics how our brains automatically recognize patterns
unstructured interviews).
based on previous experience.
Organizations now find themselves at a similar turning point as
Machine learning capabilities enable the creation and they consider including algorithm assessments as an integral
implementation of algorithmic assessments (using component of the candidate selection process. Traditional
algorithms to score large amounts of application data to assessment tools, such as psychometric assessments, tie
determine the most qualified candidates for the role). By candidate competencies, abilities, behavioral styles, and
applying machine learning principles during the algorithm preferences to outcomes like performance and retention.
development process, organizations can predict which Now psychometric assessments can be complemented by
features show up disproportionately in successful versus cutting-edge algorithmic assessments, where advanced
unsuccessful hires. The statistical engine powering the machine learning and semantic tools are used to predict
machine learning platform can compute large amounts of specific business outcomes, such as retention in role, using
data quickly and accurately, considering every feature— application data alone. This dual-pronged approach provides
and combination of features—available on the candidate’s companies with superior efficacy and versatility in improving
application or résumé. post-hire outcomes.

no statistically valid way to gain any insight—all outcomes
How to Apply Machine Learning
are likely driven by chance. For example, if you have 200
Machine learning capabilities can analyze every variable, data points, and all that data is used to build and test the
word, and phrase in a candidate’s application and evaluate model, then the model becomes very specific to that
them for potential predictive power. As a result, machine particular set. This is called “overfitting,” and the result is a
learning algorithms remove the biases inherent in human model that does not work very well on additional data.
decision making and focus on verified data to increase the
The best companies hold out a subset of the data (e.g., 20%
likelihood of hiring the best-fit candidates. You can achieve
of the total sample). Once they’ve optimized their models,
this outcome in two ways:
they test it on that held-out data subset. This methodology
1. Maximize the use of predictive information, instead proves the model’s validity on not only the tested dataset
of using discrete filters. Hiring managers and recruiters but also any future analysis on additional data.
often filter out candidates based on specific key words and
3. Ignoring adverse impact: Compared to manual
attributes (e.g., “customer service experience”) without an
processes, the algorithmic assessment is more objective
empirical basis for favoring candidates who demonstrate
and insulated from potential conscious or subconscious
those attributes. Furthermore, filters lead to simplistic “yes
biases. Any potential for adverse impact can be vetted and,
versus no” decisions based on a single trait, but the truth is
if necessary, corrected prior to a deployment through back-
often more complex than that. By contrast, algorithms that
testing. Through historical analysis, algorithmic assessment
use machine learning are firmly grounded in hard data and
can demonstrate to statistically significant levels that
can apply advanced statistical models to weigh positive and
recommended candidates do not have a lower proportion of
negative attributes.
any protected class, such as females, minorities, veterans,
2. Continuously update predictive features based on and disabled persons.
learning from outcomes. A static screening tool never
Conclusion
learns from actual outcomes—individual or organizational—
and therefore does not grow in effectiveness. Machine Machine learning holds a lot of promise for organizations
learning algorithms, however, can be trained to pick up cues that want to take the next step in using their talent data to
from outcome data and constantly update their predictive improve human capital decision making, but it’s not a risk-
features (and therefore the underlying scoring model). This free endeavor. By proactively managing machine learning’s
brings a level of intelligence and accuracy that is unmatched potential pitfalls, talent analytics leaders can protect their
by any tool or human processes. organizations from legal or reputational risk and set the
stage for next-level advanced analytics.
Pitfalls to Avoid
The risk of using algorithmic assessments incorrectly is high

to those unfamiliar with the approach. Before considering any About CEB Precise Fit
machine learning solution, carefully consider the following
risks and ensure the solution has addressed these issues: CEB Precise Fit is a flexible candidate selection
product that improves hiring outcomes and the
1. Unknowingly perpetuating bias—If you employ an candidate experience by attracting the best while
automated selection approach that closely replicates delivering value to all applicants. Our robust, flexible
human decision making, you are potentially just perpetuating architecture allows you to use the power of machine
existing human bias. For example, if an applicant’s zip code learning and application analytics to quickly identify
correlates with candidates who fail to receive a job offer, top candidates based on key features associated
your finding might seem objective. However, if the zip code with high performers within a specific company and
is predictive because it primarily comprises low-income role. This unique insight enables recruiters to make
houses that are predominantly minorities, you are replicating data-driven decisions and identify the best-quality
either a preexisting intentional discrimination or a prior hires. Read our white paper, Candidate Selection
process that unintentionally discriminated against minorities. Without Compromise, for more information.
Attempting to automate processes without deeper analysis

can be dangerous. Examples of such analysis include
using historical data to back-test for adverse impact and
predicting who actually does the job well, not just who
gets the job.
2. Overfitting data (or poor statistical modeling)—

Applied to analytics, the phrase “garbage in, garbage
out” is especially true. If your dataset is too small, there is
Q1 2017 23
How Do You Predict
Top Performers?
All we need is their application.
Learn more at
cebglobal.com/sunstone-analytics.
Drawing Bold
Conclusions
at AstraZeneca
An Interview with
Christian Cormack
AstraZeneca About Christian Cormack
AstraZeneca is a global, science-led biopharmaceutical Christian Cormack is the

company that focuses on the discovery, development, Head of Workforce Analytics
and commercialization of prescription medicines, at AstraZeneca. He partners
primarily for the treatment of diseases in three with leaders around the
main therapy areas: Oncology, Cardiovascular and business to develop insights
Metabolic Diseases, and Respiratory. The company that lead to better workforce
also is selectively active in the areas of autoimmunity, decisions.
neuroscience, and infection. AstraZeneca operates
in over 100 countries, and its innovative medicines Christian began his career as
are used by millions of patients worldwide. For more a scientist before spending
information, please visit www.astrazeneca.com and time in consulting with KPMG and subsequently joining
follow their organization on Twitter @AstraZeneca. AstraZeneca in 2005, initially with the reward team.
Christian also competed at the 2004 Olympics and is a

winner of the Oxford Cambridge Boat Race.
Q1 2017 25
E
very quarter, we interview talent analytics for data visualization. Another teammate comes from a
leaders to gain their perspectives on clinical development background, where attention to detail
is critical. The third teammate came from a consulting
issues facing them and their teams. This
background, so can manage stakeholders and keep the
quarter we spoke with Christian Cormack about
commercial perspective in mind.
his work at AstraZeneca.
We don’t have a data scientist on our team, but we are
able to collaborate with statisticians and data scientists
Tell us about your team at AstraZeneca and your main in other parts of the business. My advice to other
responsibilities. organizations about recruiting is that you don’t always
have to select the strongest person overall, or those with
CHRISTIAN: As a company focused on science, our superior technical skills; instead, you should select the
HR function aims to be a true strategic partner to the person who best fills the gaps in your team.
business, using data, analytics, and planning to inform our
people strategy. The final—and perhaps most important—capability is to
be able to build a compelling story based on the data. If
The HR analytics team consists of four people—myself you want analytics to result in action, then this is critical.
and three analysts. The primary job of our team is to A table of numbers doesn’t drive action. But a compelling
ensure that leaders in the business and HR have the story, which describes the trends and implications
right information to make the right decisions around our positioned in the context of the business, can. To do this,
workforce. Overall, our team does four things: you need to spend time with leaders and really understand
• Describes what is happening through quarterly the business.
summaries of key metrics
On the topic of skills, what are some good training
• Conducts more sophisticated deep-dive analytics or development experiences to develop talent
projects to increase understanding of the workforce analytics skills?
• Uses predictive analytics to understand the long-term
impact of current trends CHRISTIAN: Picking up on the last point, spending time
with the business is a critical development experience.
• Uses our understanding of the current state and future Workforce analytics is about providing the business
trends to support workforce planning with the information to make better decisions to drive
performance. But if you don’t know the business, you
What is your role on this team?
can’t do this; you end up shooting in the dark, developing
CHRISTIAN: My role is largely focused on stakeholder a solution in search of a problem!
management and how to position and communicate
analyses to have the highest impact. Fortunately, we For the broader HR function, we offer some formal training
have a separate team for reporting that covers basic data on how to make the best use of talent analytics. However, I
requests. Since we are a small team, we aren’t able to think the most impactful experiences come from working
provide every manager in every part of the business with in partnership with others. Whenever possible, I try to
insights, so we tend to focus on the requirements of our involve the customer in the analysis we do, which leads
senior leaders. to a higher-quality output and ensures they understand
the thought process and the limitations. By doing this, the
What are the skills that are critical for talent analytics? next time we work together they tend to be clearer and
Do you have any advice for where organizations should more specific about what they want to achieve.
look for them?
CHRISTIAN: The skill sets

needed to build a good analytics
team are not commonly found
in HR. The constructs of the
workforce can be learned very “Whenever possible, I try and involve the customer
quickly, but if an employee in the analysis we do, which leads to a higher-quality
isn’t analytical, it is difficult output and ensures they understand the thought
to turn them into someone
who is. I have one teammate process and the limitations.”
who is not only very technical
but was also previously an
artist, meaning he has a talent

Can you tell us about some of the work you are doing?
CHRISTIAN: We recently looked at recruitment data in one

part of the business and found that proportionally fewer
internal candidates were invited to interviews compared
to external applicants. However, if they were invited to
an interview, the internal candidates were 2.5 times more
likely to be offered the role. We also found that employees
who went through the employee referral process did not
perform any better than those who weren’t referred. Our
findings influenced our talent acquisition strategy and
approach to recruiting.
We’re also increasingly getting involved in workforce

planning, which is heavily dependent on workforce
analytics, as you need a clear picture of the current state
in order to build robust plans. We help define the current
state and then support the work on modeling and different
scenario planning.
What do you think contributes to your success?

And where do you still want to improve?
CHRISTIAN: I think a large part of our success comes from

our leadership teams seeing the value of talent analytics.
We are lucky that the head of HR in AstraZeneca really
values analytics and uses it to influence the business and
define strategy. Since we are a scientific, data-driven “Be bold with your
company, our business leaders inherently understand the
insights from workforce analytics. We are pushing on an conclusions; don’t just
open door. show a trend, but express
In terms of improvement, I think we could still do more an opinion based on that
to “operationalize” the data, ensuring it is integrated trend in a language the
into day-to-day decision making by being bolder about
telling the story to the business, expressing opinions
business will engage with.”
about the trends we are seeing, and then linking it to
business priorities. I think the more we do this, the more
engagement we will see.
Any last thoughts or advice?
CHRISTIAN: Don’t wait for the right time to “get going”

because it’s never the right time. Publish things, even if
they are imperfect. Only by publishing them will you get
feedback, drive accuracy, and be able to improve. Finally,
be bold with your conclusions; don’t just show a trend, but
express an opinion based on that trend in a language the
business will engage with.
Q1 2017 27
Lessons from Marketing:
Communicating
Data-Centric Insight
By Susannah Schools
T
oo many talent ana- Marketing research team at CEB, which Learning from these results, talent
lytics professionals recently looked at four critical success analytics teams can rely on four
factors for getting stakeholders to act simple communication principles to
are faced with a frus-
on insight: create a more effective strategy for
trating challenge: they work
communicating data-centric insight.
extensively on a new report 1. Communicating data-centric
Apply these principles both when you
or analysis only to have their insight
present data in person to a stakeholder
work glanced at and set aside 2. Attracting and developing and when you design reports or
with no follow-up. In fact, analytics talent databases for self-service.
only 15% of business leaders
3. Creating momentum 1. Real-Time: Keep data up to
report that talent analyt-
for analytics date, minimizing lag time.
ics has recently led them to
change a business decision. 4. Establishing a data 2. End-User Focused: Present
infrastructure
data in a way that is tailored to
HR can invest more in talent analytics,
Communicating data-centric insight the audience for easy use.
but HR must find a way to help
stakeholders (business leaders, HR emerged as the most important factor,
3. Visual: Present data using
COE staff, HRBPs) to act on the creating twice the impact of attracting
graphs, charts, and pictures
analytics produced. And it’s no easy and developing analytics talent and 2.8
times the impact of establishing a data rather than in more raw forms
task—many stakeholders have limited
infrastructure. These results make sense like database records and
time, limited data experience, and
preconceived notions about how given that effective communication spreadsheets.
things are done. strategies help you avoid the risk
of leaving your stakeholders bored, 4. Interactive: Present data in a
Knowing we have much to learn from confused, frustrated, or in danger of way that allows the end user to
those who came before HR on the misinterpreting or not seeing value in analyze and interpret it himself
analytics journey, we turned to our your findings. or herself.

1. Real-Time Case in Point: AstraZeneca
A stakeholder’s level in the organization plays a
In organizations where change is a constant, talent data big role in the types of decisions he or she will
can quickly become outdated and irrelevant. To convince be making and the data he or she will care about.
stakeholders they should use the data you are presenting With this in mind, AstraZeneca created business
to inform current decisions, ensure you are presenting performance tools for employees at different levels
up-to-date data that is relevant to current strategy and of the organization. For instance, one performance
circumstances. tool gives high-level data to executives, but more
junior employees, such as field sales representatives,
Quick Tips for Applying This Principle:
receive granular, territory-specific data. This tailoring
• Always check to ensure you are presenting the most saves employees time and effort when finding the
recent data when delivering reports or presentations. information they need.
• Let users know when the data you are presenting was
3. Visual
collected or how often databases and reports
are updated. Simply put, visually appealing presentations and dashboards
• Add automation to self-service tools to provide real- excite users, making them more eager to use the data.
time data with little maintenance. Visual appeal also makes analytics more relatable and helps
stakeholders understand more quickly how to use data.
2. End-User Focused
How data savvy are your internal clients? Are they able
• Create graphs that visually engage audiences but can
to detect trends on their own? Are they comfortable
be easily understood at a glance.
with details? Understanding your end user is essential to
encouraging them to use your data and ensuring they can • Transform data into illustrations that tell a logical story
apply it to their decision making. Successful talent analytics tied to business and talent outcomes.
professionals tailor their deliverables, taking into account • Select a creative communication vehicle to engage your
decision-maker preferences and needs. audience with analytics.

• Identify the decision-making and learning style needs
5
of your stakeholders.
Questions to Improve Your Data
• Gauge your stakeholders’ level of comfort with data Visualization
(figure 1).
1 Are we presenting or circulating
• Outline the decisions your stakeholders are information? Presentations require less
trying to make. detail and more conciseness.
2 What kind of chart or table is best?
3 Does the visual bring attention
to my message?
Figure 1: Typical Stakeholder
Decision-Making Profiles 4 Do the visuals accurately represent
Based on Reliance on Data Versus the numbers?
Reliance on Intuition 5 Is this memorable?
Visceral Decision Maker
• Seldom trusts data Source: CEB analysis.
High
analysis
• Makes decisions
unilaterally
Reliance on
Informed Skeptic
Intuition
• Applies judgment to
Case in Point: Beta Company1
analysis
• Listens to others and Thinking about how to use a visually creative
is comfortable with communication vehicle for your data can help it
dissent
capture stakeholders’ attention and enable you to
Unquestioning narrate how the data should be used. Beta Company,
Low
Empiricist
Low High
• Trusts data over
for instance, published FAQs from customers in
Reliance on Data
judgment a visually appealing coffee-table book. Using this
Source: CEB analysis. • Values consensus format made it much more likely that the recipients
will indeed read the list of FAQs and think about how
the questions apply to the work they are doing.
1
Pseudonym.
Q1 2017 29
4. Interactive
Leaders are used to analyzing and

dissecting issues and approaching
possible solutions with skepticism.
Presenting only your conclusions for
Figure 2: Seagate’s Filtered Data Visualizations them to agree or disagree with is not
likely to build their trust in the data or
see the benefits of analytics. Instead,
Workforce Planning Discussion let leaders interact with the data
themselves and draw conclusions from
it. After all, leaders are more likely to
Business trust something they’ve worked with
HRBP
Leader themselves and more likely to act on
insight they had a hand in providing.
Quick Tips for Applying This

Seagate’s Approach: Filtered Data Visualization Principle:
for Leader-Led Diagnosis of Workforce Health • Allow stakeholders to
change the way the data is
40%
Current State selected, aggregated, and
Industry represented rather than
Benchmark presenting a static report.
Internal Goal • Frame interactions around talent

20% analytics as discussions, not
presentations.
• Allow leaders to see what various
future possibilities are based on
0% the data.
Graduate
Intermediate
Senior
Staff
Senior Staff
Principal
Fellow
Case in Point: Seagate

Rather than reporting exhaus-
tive workforce data in planning
discussions, HRBPs at Seagate
use analyses and visualizations
Workforce Distribution to help leaders diagnose chal-
lenges themselves. Getting lead-
er involvement builds credibility
1. Highlights the talent condition in the identified problem, gives
that needs to be diagnosed leaders ownership earlier in the
decision-making process, and
makes a clear statement that HR
2. The current state and industry
is using its research to support
benchmark
leaders in creating a more in-
formed business agenda.
3. The challenge diagnosis leads Concluding Thoughts
to discussions about the ideal
internal goal. As more HR functions look to inform
talent decisions with data, learning
how to communicate data-centric
insight is imperative. By learning
Source: Seagate Technology LLC; CEB analysis.
how to engage stakeholders with
these findings, HR functions can
gain a seat at the table to help their
partners make better, more-informed
talent decisions.

In the News
CEB’s Take on What Others
Are Writing About Talent Analytics
Q1 2017 31
The Challenge of Mainstreaming Human Capital Reporting
By Matt Dudek, Jessica Knight, and Blakeley Hartfelder and originally appearing on Talent Daily
I
n an article for CFO,1 David among customer-facing professionals. This makes sense; some objective
McCann takes a look at The metrics that come out of the human capital measures link directly
investor groups’ engagements with to the bottom line in retail (e.g., if staff
two groups—Principles
retailers may be used to compare the aren’t on the floor or knowledgeable
for Responsible Investment companies with one another, but it’s about products, customers won’t buy).
and the Human Capital unknown how granular the information But those links are weaker in other
Management Coalition is, so therefore it’s unknown how useful industries.
(HCMC)—that encourage such comparisons will be.”
Deciding how to measure certain
organizations to publicly This discussion is evidence that metrics is another challenge.
disclose human capital metrics investors are starting to understand Absenteeism and turnover have
(e.g., turnover, absenteeism, the importance of human capital near-universally accepted definitions,
employee engagement). It’s when looking at a company’s value, but metrics such as engagement
an uphill battle for these and they are looking for ways to get and productivity have competing
data and information related to that. definitions used by different vendors.
advocates, McCann writes,
But actually understanding what the Therefore, investors and analysts
as not everyone is convinced metrics indicate and how to interpret cannot benchmark across firms,
these metrics are worth using them is harder than it may seem. This limiting the momentum of mainstream
or disclosing. might be an opportunity for talent reporting for these metrics.
analytics professionals, and HR leaders
For the opposition, McCann cites Tom more broadly, to step in and educate Improving human capital data
McGuire, former finance and marketing their peers in corporate leadership collection and reporting are worthy
executive and now talent strategy endeavors to better understand and
about which metrics to share as well
leader at Talent Growth Advisors. communicate the value of talent. But
as how to think about them. People
McGuire states, “Whether a company given these challenges, it’s not hard to
data is only going to be as strong as
is well run is a good question, but a see why efforts to make human capital
the story built around it.2
more relevant one is, how do its people reporting a standard practice have
impact its value? To understand that, One potential barrier to increased been slow to gain steam.
you need to look at the company’s reporting of human capital metrics David McCann, “The Holes in Human Capital
1
intellectual capital—patents, brands, is the need for guidelines equivalent Metrics,” CFO, 11 October 2016, http://ww2.cfo.
and proprietary technologies and com/human-capital-management/2016/10/
to the generally accepted accounting
holes-human-capital-metrics/.
methodologies. The only source of any principles. Not having such guidelines
of those things is people.” may also limit that data’s usefulness in 2
For more information, see our article at
cebglobal.com/talentdaily/hr-needs-to-know-
comparing companies’ relative value sentiment-analysis/.
McCann also quotes Jeff Higgins, CEO and performance. To address this
of the Human Capital Management issue, the Center for Talent Reporting
3
CEB Learning & Development Leadership
Council members can watch our 2014
Institute: “If you lose 20% of has created what it calls Talent webinar with Dave Vance, the center’s
management in a year, that’s way too Development Reporting Principles executive director, at cebglobal.com/member/
learning-development/events/replays/14/
high. Losing 20% of your call-center (TDRp) as a standard for human talent-development-reporting-principles-new-
workers is okay. It’s also fine if 20% capital reporting.3 standards-for-the-m.html.
of a retailer’s customer-facing staff is
lost. But it’s disastrous if a professional McCann notes that HCMC has started
services firm has 20% turnover to push for disclosure among retailers.

Winning on HR Analytics Book Review
By Caitlin Dutkiewicz
In Winning on HR Analytics, Ramesh Soundararajan The book also succeeds in dispelling common myths in talent
and Kuldeep Singh address how to take “an analytics—for example, that you need to have perfect data or
analytic view to HR.” With chapters focused make large technology investments to run successful analyses.
on key HR domains—organizational design,
Soundararajan and Singh’s work however, sweeps over some
talent acquisition, learning and development,
of the difficulties talent analytics teams face today—challenges
engagement and retention, competencies, and
such as accessing and synthesizing data from different
compensation and benefits—the book walks
systems, developing strong internal partnerships with data
readers through traditional forms of measurement
owners and line leaders, and creating clear processes for
and highlights ways organizations can mature their
project scoping and prioritization.
analytical approach.
But if you’re new to HR measurement and analytics, this
The book strongly emphasizes grounding analytics work in
book will serve as a good primer for the possibility and
business questions by aligning to key strategic initiatives. As
power of data for the HR function. It’s full of examples and
we know, solving leaders’ most pressing challenges not only
offers thoughtful questions and tactical advice to get you
drives impact but also secures buy-in and creates momentum
approaching HR analytically.
for analytics projects moving forward.
How Can Talent Analytics Create More Business Value?

By Marianne Stengel and originally appearing on Talent Daily
I
n an article for the Harvard As evidence, Veeramachaneni cites identified their main challenges
Business Review,1 MIT his experience participating in a panel as deriving better insights from
of machine learning enthusiasts: “I the data they were collecting and
research scientist Kalyan
polled the audience of about 150 communicating those insights to their
Veeramachaneni details internal clients. Many participants said
people, asking, ‘How many of you
some research into the have built a machine learning model?’ their organization was still ineffective
question, “What would it take Roughly one-third raised their hands. at using talent analytics to drive
for businesses to realize the Next, I asked, ‘How many of you have business decisions.
full potential of their data deployed and/or used this model to
In HR, analytics teams generally do not
generate value, and evaluated it?’ No
repositories with machine have too much data or models that are
one had their hand up.”
learning?” One key challenge too complex. We hear quite often that
he identifies is that “machine In other words, the machine they know better than to hire data
scientists who are only interested in
learning experts often didn’t learning experts wanted to spend
creating beautiful models. Rather, their
build their work around the their time building models, not
first priority is building a team that can
processing massive datasets or
final objective—deriving translate whatever data they have into
translating business problems into
business value.” meaningful reports that drive business
prediction problems.
leaders to act. At a time when many
“In most cases,” Veeramachane- To get more value from their data, organizations are just starting to build
ni states, “predictive models are Veeramachaneni concludes that their talent analytics capabilities, it’s
meant to improve efficiency, in- a good reminder to keep aspirations
companies “need to focus on
crease revenue, or reduce costs. small and to not be discouraged if your
accelerating human understanding of
But the folks actually working analytics efforts don’t yet live up to the
data, scaling the number of modeling
on the models rarely ask ‘What stories you’ve heard about the Googles
questions they can ask of that data in
value does this predictive and IBMs of the world.
a short amount of time, and assessing
model provide, and how can we their implications.” These are valuable 1
Kalyan Veeramachaneni, “Why You’re Not
measure it?’ Asking this ques- principles for talent analytics leaders
Getting Value from Your Data Science,”
Harvard Business Review, 7 December 2016,
tion about value proposition to keep in mind. https://hbr.org/2016/12/why-youre-not-
often leads to a change in the getting-value-from-your-data-science.
original problem formulation, Business impact is not easy to achieve. 2
For more information, see our article at
and asking such questions is In a peer benchmarking conversation2 cebglobal.com/talentdaily/4-trends-in-talent-
analytics/.
often more useful than tweak- at our September ReimagineHR event
ing later stages of the process.” in Miami, talent analytics professionals
Q1 2017 33
A Lesson in Using Predictive Analytics to Drive Retention
By Blakeley Hartfelder and originally appearing on Talent Daily
W
hen we talk to Wright then provides some background what is driving its most important
HR leaders about on this school’s effective process: “In employees to leave the organiza-
2001, the school undertook a new tion—and stop them.
predictive analytics,
technology initiative spurred and
the first thing they usually funded by the No Child Left Behind 3. Anderson has gone beyond analysis
want to do with this new legislation. Using Ultimate Software’s and numbers to develop clear plans of
advanced tool is improve UltiPro, Paulk said the company ‘grew… action when it does discover that a key
employee is thinking of leaving.
retention. That’s definitely and kept head count flat, reduced paper
easier said than done— [processes] by 95%, and increased the
To be able to act on insights and data
especially if you want the time spent on employee development
from predictive retention analytics, the
by 30%. The software also allows
project to actually drive most successful organizations think
staff to manage time and attendance about how they will use those results
results instead of just being from anywhere [and yields] improved even before the analysis starts coming
an interesting research topic. reporting and compliance. in. Who will they share the information
Aliah Wright at the Society for Human with? What steps will HR and managers
“‘The software also helped us avoid the
Resource Management highlights1 the take to encourage employees to stay?
loss of key talent with predictive tools.
success story of one organization And are those steps the same for
It’s really powerful, and it’s astonishing
that had a strong need to retain its all employees?
the results we’ve seen,’ Paulk said.
highly skilled employees and used ‘[The tools] helped us understand our These best practices apply whether you
predictive analytics to help meet that challenges and put the puzzle pieces are managing retention with advanced
goal: “When a top employee at the together.’” analytics or with low-tech tools like stay
Anderson Center for Autism, a private
interviews and career conversations.
school in Staatsburg, NY, handed It looks like Anderson has done a
CEB Corporate Leadership Council™
in her resignation, the school’s HR few things really well, which makes
members can read more about how
department was expecting her. The it a great example of how to apply
to measure and mitigate attrition risk
HR staff had been using a predictive analytics most effectively.
in our HRBP Quarterly.
analytics program to help them gauge
retention. 1. Anderson didn’t just choose to Aliah D. Wright, “How One Company’s Foray
1
focus on retention because it was into Predictive Analytics Aids Retention,”

Society for Human Resource Management,
“‘The software is so good that we interesting or because it had the data 21 October 2016, https://www.shrm.org/re-
were developing a retention plan to explore the topic; Anderson was sourcesandtools/hr-topics/technology/pages/
for her as she was preparing to how-one-companys-foray-into-predictive-ana-
instead trying to solve a problem that lytics-aids-retention.aspx.
resign,’ said Gregg Paulk, director was affecting its bottom line.
of information technologies for the
92-year-old nonprofit organization. 2. Anderson was focused: It applied
After HR staff spoke with her, predictive analytic tools to essential
‘she actually rescinded her resignation,’ employees, not everyone. This hope-
he added.” fully means it can truly understand

Learn with Your Peers
Upcoming Live Events
Hear from Your Peers
Maximizing the Impact from Talent Recent Webinars
Analytics Investments
Driving Diversity and Inclusion
A Workshop on Allocating Resources and
Outcomes Through Analytics
Building Team Capabilities
Learn how organizations use analytics
Join us at a 2017 talent analytics workshop. Learn about the to quantify and increase Diversity and
latest insights from our research on growing talent analytics Inclusion’s impact on critical business
capabilities, and collaborate with your peers to solve the outcomes.
resourcing and capability challenges most critical to your
organization. The State of Talent Analytics in 2017
Hear about the key trends shaping talent
Collaborate on questions like the following: analytics in 2017 and how to address them
• How do we train and hire the right people to build an at your organization.
analytics team?
• How do we ensure the quality and usability of our data? Leader Analytics: Predicting How
• How can we identify the right structure for talent analytics Leaders Will Perform in Context
at our organization (e.g., centralized, decentralized)? Learn how to generate relevant, data-
• How do we determine which investments to make in driven insights on your organization’s
technology? most valuable and impactful resource—
• How do we establish processes to focus talent analytics on its leaders.
the right priorities in the right way?
• How can we build relationships to take advantage of Three Critical Obstacles to Creating
analytics capabilities elsewhere in the organization? Talent Analytics Capability
Hear about the foundational capabilities
Event Dates you need to manage data in HR and
Chicago: 29 March 2017 avoid inadvertent exposure to legal risks
New York: 27 April 2017 or invalid analyses.
Denver: 11 May 2017
Houston: 24 May 2017
Toronto: 8 June 2017
London: 28 June 2017
Brussels: 10 October 2017
Eligibility: The most senior person whose primary responsibility
is talent analytics, plus the head of HR. Up to two attendees per
organization.
Talent Analytics Webinar

Innovative Uses for Text Analytics
23 March 2017 | Noon EST
28 March 2017 | 2:00 p.m. AEDT
Join us to learn about the prerequisites for conducting your

first text mining project, sources for gathering unstructured
Register for these events, and
text-based data, and ways in which organizations are using
access these webinar replays
text analytics to solve talent challenges.
at ceburl.com/1rpq.1
1
The link is available to CEB Corporate Leadership Council™ members only. Q1 2017 35
About Talent Analytics Quarterly
We’ve heard your demand for more dedicated
support on talent analytics. Talent Analytics
Quarterly, our quarterly publication featuring
business and functional insights for talent analytics
leaders, is one of a series of new resources we have
created to meet this demand.
Feedback on what you see?

Which articles in this publication did you find
most helpful? Is there a topic or type of content
you would like to see in future editions? Send
your feedback on the Talent Analytics Quarterly
to [email protected] today.

Ceb Talent Analytics Quarterly q1 2017

Uploaded by

Copyright:

Available Formats

Ceb Talent Analytics Quarterly q1 2017

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ceb Talent Analytics Quarterly q1 2017

Uploaded by

Copyright:

Available Formats

Talent Analytics

What We Heard from

Aligning Talent Analytics

Beginner’s Guide to Text

Drawing Bold Conclusions

Lessons from Marketing:

16 Beginner’s Guide to Text Analytics for HR

22 The Advantages of Machine Learning

25 Drawing Bold Conclusions at AstraZeneca

28 Lessons from Marketing:

2 Talent Analytics Quarterly © 2017 CEB. All©rights

And of course, we continue our well-received tradition of bringing you

As always, we welcome your feedback and suggestions for how we can

4 Talent Analytics Quarterly © 2017 CEB. All rights reserved.

Where are they investing in 2017?

What projects are they working on in 2017?

Strategic Attrition/ Engagement Diversity Talent Leadership

Access our latest insights on our Talent Analytics Portal

6 Talent Analytics Quarterly © 2017 CEB. All rights reserved.

We have heard of a few strategies companies have taken

8 Talent Analytics Quarterly © 2017 CEB. All rights reserved.

4. Prioritize business acumen and

10 Talent Analytics Quarterly © 2017 CEB. All rights reserved.

EY | Assurance | Tax | Transactions | Advisory About Blair Hopkins

12 Talent Analytics Quarterly © 2017 CEB. All rights reserved.

Can you tell us more about the workforce

BLAIR: Like most sectors, ours is facing disruptive change

how do you choose?

Source: CEB 2017 Talent Analytics Innovation Survey.

14 Talent Analytics Quarterly © 2017 CEB. All rights reserved.

Just 8% of senior HR leaders

Learn how the CEB HR Analytics Leadership

16 Talent Analytics Quarterly © 2017 CEB. All rights reserved.

Capability 2: Identify the Main Topics and Themes

What and Why: Substantive text analytics requires moving

18 Talent Analytics Quarterly © 2017 CEB. All rights reserved.

Capability 3: Derive Additional

The more rational transformations and features you create

Testimonial: The Joy of the First Word Cloud

Producing a single word cloud at the end of day two meant

There is much to learn, and these building blocks are just

About CEB Talent Management Labs

This article is contributed by our Talent Management

For more information on CEB or the

20 Talent Analytics Quarterly © 2017 CEB. All rights reserved.

Source: CEB analysis.

Constructs and Outcomes Revealed • Personality

22 Talent Analytics Quarterly © 2017 CEB. All rights reserved.

The risk of using algorithmic assessments incorrectly is high

Attempting to automate processes without deeper analysis

2. Overfitting data (or poor statistical modeling)—

AstraZeneca About Christian Cormack

AstraZeneca is a global, science-led biopharmaceutical Christian Cormack is the

Christian also competed at the 2004 Olympics and is a

CHRISTIAN: The skill sets

26 Talent Analytics Quarterly © 2017 CEB. All rights reserved.

CHRISTIAN: We recently looked at recruitment data in one

We’re also increasingly getting involved in workforce

What do you think contributes to your success?