White Paper

Analytics Collaboration:
A Primer
 A N A LY T I C S C O L L A B O R AT I O N : A P R I M E R W H I T E PA P E R

Introduction
This paper examines the state of collaborating around data and analytics in
organizations today. It also identifies and describes issues with current practices and
describes how they should evolve in order to achieve an optimal data-driven culture
within an organization.

As you read this paper, the most important thing to keep in mind is that analtyics
collaboration is a still nascent process where best practices and methods are still
emerging. As organizations are evolving their analytics for deeper insights that answer
increasingly complex questions, they are recognizing that collaboration can help them
get there faster and with greater accuracy.

In order to discuss analytics collaboration, we must first define what it is…but also what
it is not. Afterwards, we will discuss the different ways it manifests itself, the benefits
that can accrue to organizations practicing it and, finally, the different methods of
collaboration.

PAG E 2
 A N A LY T I C S C O L L A B O R AT I O N : A P R I M E R W H I T E PA P E R

Why the need for analytics collaboration?

While we just stated that the area of analytics collaboration is young, a good
percentage of organizations claim to collaborate around analytics today. In a recent
survey, data and business analysts were asked to describe how their team collaborates
on analytics, with over 60% claiming to collaborate on most or all of their projects as
can be seen in Figure 1 below.

Q. Please describe how your team collaborates on analytics

We collaborate on all our projects

We collaborate on most and do a

few individually

We collaborate on a few, but

mostly work individually

We work on all our projects

individually

0% 10% 20% 30% 40% 50% 60%

Figure 1: Analytics collaboration survey

The core need for analytics collaboration is to increase the analytics team’s knowledge
around data. A greater understanding on the analytics assets at their disposal
facilitates greater trust, brings new ideas on how to use the analytics to derive new
insights, and generally produces faster insights with greater detail and accuracy.

PAG E 3
 A N A LY T I C S C O L L A B O R AT I O N : A P R I M E R W H I T E PA P E R

What exactly is analytics collaboration?

The word “collaboration” itself has multiple meanings. In the most used one, it refers to
the action of working with someone to produce or create something. This accurately
captures what we mean when we talk of analytics collaboration; namely, the act of
working with people in an organization to produce or create something of value with
the organization‘s data and analytics assets as the foundation.

We also note that we are focusing this paper on internal analytics collaboration among
team members within the same organization. There is also collaboration with external
business partners, clients and suppliers – which is also a worthwhile goal – but carries a
different set of requirements and challenges which we do not tackle in this paper.

The very term “analytics collaboration” is itself in its infancy, with the industry not having
fully coalesced around a canonical set of capabilities that define it. To that end, this
paper examines the different methods by which analytic teams collaborate around data
and proposes a critical set of capabilities required to fulfill analytics collaboration needs.

Methods of analytics collaboration

Analytics teams, comprising of data analysts, business analysts, data scientists and
management, will collaborate in multiple ways, all in an effort to share analytics assets
and knowledge about those assets:

• Collaborating around discovery, creation, sharing and use of analytics assets

• Sharing personal knowledge of analytics assets
• Recommendations on how to use the assets based on personal knowledge
• Automated recommendations on best use of assets based on how others use
them
• Shared automated (AI-driven) knowledge about the assets, including intimate
understanding of the data, its uses, connections and other characteristics
gleaned from user activity and detailed analyses of the underlying data
• Shared derived insights from the assets – both personal and automated
• Discussions, chats and collaborative documentation of assets
• Social feeds, sharing and discoverability of asset characteristics and insights

We will expand on these collaboration methods in the following section. But before
we get to that, we also need to define what is not part of analytics collaboration.

PAG E 4
 A N A LY T I C S C O L L A B O R AT I O N : A P R I M E R W H I T E PA P E R

What analytics collaboration is not

Today, analytics collaboration is a process, set of best practices, and series of
supporting capabilities inside of software products to facilitate the collaboration. Given
the rather amorphous state of the collaboration capabilities at the moment, some
vendors might categorize features from their business intelligence platforms as being
part of analytics collaboration. But we would beg to differ.

For example, the following products or features of larger platforms are not by
themselves classified as analytics collaboration:

• Enterprise data catalogs

• Source code management and code repositories
• Business Process Modeling (BPM)
• Shared data visualizations
• Shared enterprise ETL / ELT workflows

Some of these products might have a feature or two to support collaboration in some
form but by no means do they cover all the aforementioned methods of collaboration.

PAG E 5
 A N A LY T I C S C O L L A B O R AT I O N : A P R I M E R W H I T E PA P E R

Types of analytics collaboration

As seen in the “Methods of analytics collaboration” section above, there is an
abundance of capabilities that define an end-to-end analytics collaboration process.
These capabilities generally fall within two distinct types: implicit collaboration and
explicit collaboration. In this section, we present the basic characteristics of each type,
describe their benefits and outline how they would appear in a collaboration platform.
There is no implied assumption as to which type is better; both are valuable and
provide benefits, and these benefits are multiplied in a process that can seamlessly
blend both together.

Implicit collaboration types and associated benefits

We consider implicit analytics collaboration methods to be those that emerge
automatically and organically through an organization. These methods are typically
implemented in a software platform by leveraging the knowledge of all team
members’ actions and combining it with machine learning around activities, workflows
and uderlying data sets. They are also surfaced through the paradigm of social media,
utilize machine learning and recommendation engines, and are centered around
sharing and taking advantage of the power of collective team activity throughout the
platform.

The importance of implicit collaboration to delighting analytic teams and enhancing

an organization’s knowledge around analytics assets cannot be overstated. Implicit
analytics collaboration helps to surface potential insights hiding inside the assets,
create seamless workflows and ad hoc collaboration between teams, and generate
additional exploration and discussions.

Social media/ ”Netflix UX”-influenced modalities

These are recommendations on how to best use the analytics assets, under which
conditions and for which purpose, exposed through a social media-style user
interface in a collaboration platform. Based on implicit collaboration performed
ad-hoc by analytic team members through traditional ways of sharing information
(such as speaking face to face, over the phone or email) these approaches assume a
software platform smart enough to recommend the proper uses of the asset to its
users based on a variety of analysis factors. These include both the usage-derived
recommendations we explore in this section, and data-driven insights which are
described later. Discoverability and easy sharing are major drivers, and optimizing
analytic team collaboration is the largest benefit. Some hallmark methods are listed
below.

• Clear descriptions of each asset, its purpose, provenance, connections, potential

uses, ‘best for’ recommendations and other relevant facts around the asset

PAG E 6
 A N A LY T I C S C O L L A B O R AT I O N : A P R I M E R W H I T E PA P E R

• Recommendations around which assets to use in analysis and exploration

• Recommendations for similar assets to the ones currently being used

• Surfacing of relevant assets used by other team members through a Netflix-style

recommendation engine and interface

• Display of interesting facts about all assets in the organization through a ‘Did you
know…’ style of interface

• Identification of trending assets, and exposure though the interface, for possible
consideration by the team

• Highlighting of assets recently added to or created in the platform

• A generalized social media-style interface combining aspects of the above into a

holistic “newsfeed” interface for analytics assets

• Seamless discoverability and frictionless sharing throughout the platform.

This needs to be designed in from the beginning as a major characteristic of
the platform. Ideally, every piece of insight and information, generated both
automatically and through team member intervention, should be shareable by
one click and discoverable at a glance

Data-driven approaches
These are additional recommendations, informational points and other actions based
on deep statistical, machine learning, AI and other analyses of the underlying asset in a
collaboration platform. These types of approaches are not feasible with the traditional
ad-hoc way of doing collaboration that most analytic teams perform today. Below
follows a short list of useful approaches.

• Automatically generated insights based on the data in the asset. While this
feature may overlap with some of the newer capabilities in the self-service
business intelligence space, it is also highly important to drive collaboration
through unearthing connections and other insights inside the data, and exposing
them to users. This should take the form of both informational facts around
assets and recommendations for further usage and analysis

• Recommendations for additional asset exploration and the combining of assets

into an „uber asset“

PAG E 7
 A N A LY T I C S C O L L A B O R AT I O N : A P R I M E R W H I T E PA P E R

• Detailed statistics on the asset, including provenance, counts, usage

characteristics, connections, popularity and trends

• Automatic alerts for significant additions and changes to the asset

Explicit collaboration types and associated benefits

We consider explicit analytics collaboration methods to be those that require an
explicit team member action for initiation or completion. These methods typically
revolve around deliberate, team-driven requests, delivered through a variety of
shared communication channels, which should leverage features of a collaboration
platform to enable cooperative asset discovery, annotation and description. These
methods are typically implemented through a combination of interfaces, including
chat-style experiences, like/dislike (or “thumbs-up”/“thumbs-down”) feedback and a
variety of documentation mechanisms. The overall experience is centered around
explicit interactions between team members to better understand, document,
curate and collaborate around the data. While manual, ad-hoc actions around explicit
collaboration is definitely possible (but certainly cumbersome), the benefits of a
software platform that codifies, centralizes and exposes these methods for wide use by
analytic team members cannot be overstated.

Explicit collaboration is important to closing the loop between team members

and the implicit recommendations described in the previous section. Through this
type of collaboration, the team cooperates to deepen its understanding of analytics
assets, above and beyond what the software platform used for collaboration has
automatically generated. These processes also result in teams collaboratively
discovering and documenting additional details around asset usage, provenance and
purpose. Through this continuously enhanced feedback loop the team generates a
permanent record that helps current and future members of the organization navigate
and understand its analytics assets.

The main methods of explicit analytics collaboration include:

Conversation and notes

Through the use of structured conversations, lightweight notetaking and annotation
capabilities, a analytics collaboration platform should facilitate frictionless
collaboration and decision-making. These notes and conversations can also be further
analyzed through a variety of AI techniques. Some techniques and approaches in this
area are listed below.

• Annotations and ‘case notes’ on a per project, product, asset, team member or
other level, in order to document in place individuals’ particular insights derived
from the asset

PAG E 8
 A N A LY T I C S C O L L A B O R AT I O N : A P R I M E R W H I T E PA P E R

• Chat interfaces around assets, ideally within the collaboration platform, or

captured through the organization’s collaborative chat software (e.g. Slack or
Teams). Machine learning and sentiment analysis can also be brought to bear on
these conversation to further expose insights to the organization

• Structured email dialog around data, ideally within the platform, or at the very
least captured though the organization’s email system through tags or plugins.
Machine learning and sentiment analysis can also be utilized to further derive
knowledge from the organization’s email exchanges

Team-driven documentation
Through the use of shareable notebooks, wikis and other techniques, a canonical set of
documentation around data can emerge through the collective understanding of the
analytic team members. These will of course need to be shareable and discoverable
and also serve as main input points in the cooperative asset invenmtory and discovery
approaches described in the next section. Approaches for team-driven documentation
include those in the list below.

• Shareable notebook-style data stories, centered around exploration of data,

testing out of hypotheses and ideas, bringing together disparate assets, and
presenting quick conclusions and interesting facts about the underlying data
analysis. These assets serve as the basis for documenting exploration and its
associated assumptions and insights, can be augmented over time by analytic
team members and serve as basis for further development of a holistic data story

• Wikis, collaboratively edited documents that distill knowledge around data

and can serve as the canonical source of organizational knowledge around a
particular set of assets

• Light workflows, centered around connections between data assets, annotated

by the team that serve as a quick way to discover non-obvious ways to combine
data assets

Cooperative data management and discovery

Building on documentation that the team has already created, as well as leveraging
automatic and user-generated insights, analytic teams can work in a collaborative
manner to analyze, prep and curate the organization’s data. In addition, the
collaboration platform’s recommendation engine can be trained to further enrich and
inform the analysis results. Typical methods of cooperative data management and
discovery are enumerated below.

• Cooperative discovery and analysis of assets, utilizing common communication

mechanisms to perform in-place data analyses, identify gaps and issues with the
data and jointly improve the quality of results throughout the organization

PAG E 9
 A N A LY T I C S C O L L A B O R AT I O N : A P R I M E R W H I T E PA P E R

• Asset modeling collaboration, either as a subset of full-blown data analysis or

as its own standalone task. Teams will particularly benefit from shareable and
documentable combining operations. A platform’s role as either as a mean
to combine assets with collaboration capabilities or as a hub to connect to
specialized tools, layering on analytics collaboration capabilities seamlessly is
important in advancing the team’s knowledge

• Cooperative data modeling, in parallel and as a result of the above-mentioned

discovery and analysis. Through the certification of data assets and data stories,
and by using the shared documentation and communication processes already
described, the team members should be able to certify data assets, analytic
results, workflows, annotations, insights and all other automatically and user-
discovered information about the assets. A byproduct of such cooperative
curation is an ad hoc, highly accurate asset marketplace for the organization

• Training of the collaboration platform’s recommendation engine by the use

of a social media-influenced like/dislike mechanism for automated and user-
generated insights, allowing recommendations, insights and discoveries of the
assets to rise and fall in relevance and popularity as the team collaborates more
closely over time

PAG E 1 0
 A N A LY T I C S C O L L A B O R AT I O N : A P R I M E R W H I T E PA P E R

The benefits of analytics collaboration

The descriptions in the previous sections of analytics collaboration processes and
methods lead us next to discuss how they add value and cause benefits to accrue to an
organization. We believe that there are six distinct axes around which benefits can be
realized and for which the need for a analytics collaboration process becomes evident
for an organization. These can be seen in Figure 1 and are described in detail further
below.

Deeper Enhanced
Knowledge Increased Enhanced Catalog ac-
data under- analytic
sharing trust Curation celeration
standing provenance

Figure 2: The analytics collaboration axes

Axis 1: Knowledge sharing

If we could pick only one reason to invest in a analytics collaboration process, it would
be for its ability to enable broad knowledge sharing across an organization. Currently,
most collaboration is done in an ad-hoc manner, often through endless email back
and forth, resulting in knowledge output that is siloed, ineffective, incomplete and
not easily shareable or discoverable. Analytics collaboration has the potential to
significantly improve on this experience, removing barriers to entry and most of the
friction involved in working with others and sharing discoveries and data insights.

Axis 2: Deeper data understanding

By enabling more people to jointly work together on analytics and by centralizing
the insights and information uncovered, the organization becomes wiser over time
about its analytics assets and their associated characteristics, usage, connections and
popularity. In addition, as the understanding around analytics deepens, new uses for
the analytics emerge and can lead to opportunities that otherwise would not have
been easy to predict.

Axis 3: Increased Trust in Data and Analytics

Increased knowledge and transparency around analytics builds greater trust in how
to use the assets and the information they produce. This benefit shows up with
the analytics producers – analysts and data scientists – enabling them to produce
results faster and with greater detail and accuracy. These benefits also bubble up
to management – who can create data-driven strategies from the analytics with
confidence, and execute at an agile pace.

PAG E 1 1
 A N A LY T I C S C O L L A B O R AT I O N : A P R I M E R W H I T E PA P E R

Axis 4: Enhanced curation

As the organization deepens its knowledge of the analytics assets, it is presented
with a big opportunity to organize, tag, document and thoroughly curate its formerly
uncategorized assets, as a certified set of canonical analytics assets. In addition,
through the sharing and easy discoverability that a analytics collaboration process
enables, these curated assets become widely available as golden sources for myriad
uses, displacing the current ad hoc, local, manually intensive and only partially
accurate methods of local data extracts and marts.

Axis 5: Cataloging and Inventorying acceleration

As many organizations who have invested in data catalog software in the past few
years can attest, getting an accurate, continuously maintained catalog of golden
source data to be relevant and used is a mighty task. Analytics collaboration methods
significantly accelerate adoption through the collective power of team members’
analyses, iterative and perpetually improving processes and the easy discoverability
and sharing that is at the core of the methods described above. In addition, analytics
collaboration covers any analytics asset, not just data, helping teams link any of
these assets to results and their analysis at hand. Organizations can move away
from manually-maintained, infrequently-updated and sparsely-used catalogs and
leverage the power of analytics collaboration to always access the most accurate, best-
maintained and meticulously-documented analytics assets.

Axis 6: Enhanced data provenance tracking / auditability

Lineage, master data management, governance, traceability, provenance, auditing,
as well as other, related technologies and their associated software platforms, are all
in use today, across a variety of organizations, to mostly middling levels of success.
An analytics collaboration process – by virtue of its design as a user-powered set of
methods and tools that make it easy to discover, share and curate analytics assets
– is in the best position to answer and document questions related to the facets
mentioned above. In addition, by leveraging the process to enable a widely-used,
centralized asset marketplace in the organization, significantly increased accuracy and
usage can be achieved.

PAG E 1 2
 A N A LY T I C S C O L L A B O R AT I O N : A P R I M E R W H I T E PA P E R

Conclusion
Analytics collaboration is a complex, still-evolving category that, while serving a real
need in an organization, has not yet reached maturity or gained critical mindshare in
the industry. The situation is rapidly evolving, however, with the benefits becoming
increasingly apparent to organizations of all stripes, the need for well-defined
collaboration processes becoming better understood and a software ecosystem
gradually emerging.

In the near-term, the industry will begin to mature and dominant vendor products
will establish themselves. A seamless combination of implicit and explicit analytics
collaboration capabilities will remain the holy grail of the market for some time.

Vendors that enable the processes described in this report, who invest in a frictionless
user experience, and avoid simply rebranding their business intelligence products, will
establish themselves as leaders in the marketplace. Similarly, customer organizations
that fully buy into the ethos of end-to-end analytics collaboration will realize most
of the benefits, enhance the quality of their data, build a genuine data-driven
collaboration culture and gain significant insights into their business through a
combination of robust processes and world class software platforms to enable them.

PAG E 1 3