TEST CASES ARE NOT TESTING:

TOWARDS A CULTURE OF TEST

PERFORMANCE
JAMES BACH
WASHINGTON, USA
AARON HODDER
WELLINGTON, NEW ZEALAND

“We are seeking a Graduate Test Analyst to write and execute test cases to ensure quality is
delivered to an extensive client base”

“Responsibilities: Assist with requirements gathering and develop test cases that satisfy the
requirements”

“To be successful in this role, the following is required:...To have written manual test cases and
involved in test execution...Proven experience in writing Test Conditions & Test Scripts”

“Responsibilities:...Create comprehensive test cases relevant to test conditions...Execute test cases

and report results”

From job advertisements listed on seek.co.nz on 24th January, 2014.

OUR INDUSTRY HAS A TROUBLING OBSESSION WITH TEST CASES. WRITING TEST CASES IS
mentioned in job descriptions as if it were the main occupation of testers. Testers who approach us
for advice too often use phrases like “my test cases” to mean “my work as a tester.”

This test case focus has an innocent basis, and is for the most part well-intentioned. But it has
become toxic to the field and we, the authors, believe it’s one continuing reason why testing
commands so little respect. Test cases are holding us back.

It’s time for an intervention. In this article, we will critically examine the notion of a test case, and the
culture that so often surrounds it. We will show why testing cannot be encoded in test cases, then
suggest an alternative vision based on testing as human performance, rather than on artifacts.

! TESTING TRAPEZE | FEBRUARY 2014 30

 “
What is a test case?
There is no objective, Definitions vary from place to place. One common idea of a test case is
universal way of that it is a set of instructions and/or data for testing some part of a
accounting for test product in some way. Testers will speak of test cases that are written, or
cases about to be written. In some projects, a test case will always have a

”
unique procedure associated with it. In other situations, test cases may
share procedures and be distinguished by unique data.

There is little difference between Agile and non-Agile projects when it

comes to test cases. Test cases in Behavior-Driven Development are
often specified with a formal structure for execution by tools such as
Cucumber. Agile projects tend to be more automation-oriented and
more focused on using test cases to define “done.”

Here are two simple, contrasting examples:

First, consider a table of test cases characterised by differing

conditions. The procedure associated with these cases is implied or
documented elsewhere. Thus, test cases may be spoken of as
variations on one test idea. Often the contents of each cell are used to
report the status of the product with respect to that case. Note that it’s
not obvious how to delineate or count the cases in this example.
“Cookies Accepted,” being a distinct idea, may fairly be called a test
case, or perhaps each cell in the table is a test case.

Chrome Firefox
Cookies Accepted Pass Pass
Cookies Not Accepted Pass Fail

The second example is also commonly called a test case. It is a step-

by-step procedure:

Step Expected Result Actual Result

1. Start Chrome browser Browser starts Pass
2. Set to accept cookies Browser accepts cookies Pass
3. Attempt log in User home screen displays Pass

As above, each step might be called a test case because each step
has a verification operation associated with it. There is no objective,
universal way of accounting for test cases, and hybrids of data-like and
procedure-like cases are often found, wherein a procedure-like test

! TESTING TRAPEZE | FEBRUARY 2014 31

 case includes variables that take their values from data-like test cases
stored in tables.

These are not the only kinds of things called test cases. A tester can
make a loosely structured list of ideas such as “load a corrupted file”
and call them test cases. Considering the variety of things called test
cases around the industry, a definition that covers all of them would
have to be quite general. However, our concern in this article is mostly
with detailed, procedural, documented test cases, and the attitudes
surrounding that kind of test case.

Note: Often test cases are poorly designed. James was once ordered
to create test cases by adding the words “verify that...” in front of the
literal text of each requirement. But that silliness is not our complaint in
this article. The issues we are raising hold even if you assume that
each test case is well-designed. Our claim is that even good test cases
cannot comprise or represent good testing.

“
The Innocent Foundation
Programmers write code. This is, of course, a simplification of what If programmers write
programmers do: modeling, designing, problem-solving, inventing data explicit source code
structures, choosing algorithms. Programming may involve removing or that manifests working
replacing code, or exercising the wisdom of knowing what code not to software, perhaps
testers write explicit test
write. Even so, programmers write programs. Thus, the bulk of their
cases that manifest
work seems tangible. The parallel with testing is obvious: if
testing.
programmers write explicit source code that manifests working
software, perhaps testers write explicit test cases that manifest testing.

It is seductive to think of written test cases as the “code” of testing. We

may covet the sense of accomplishment that comes from producing a
”
tangible asset. We may delight in the simplicity of direct
correspondence between test cases and written code. There certainly
are situations where thinking in terms of detailed, explicitly specified
test cases is appropriate. For instance, when we need to cover a
function that can be described cleanly and systematically in terms of a
few interacting variables, it can be sensible to model that space
formally and then formally specify which points in the space to test. Or
perhaps if we want to carry out an intricate and specific test, or a test
that requires several testers to coordinate their actions, or even a set of
simple fact checks that must be performed periodically - in any of
these cases it can help to encode them down step-by-step. And of

! TESTING TRAPEZE | FEBRUARY 2014 32

 “
course, if the operation will be performed by a machine, it must be
Chocolate-covered encoded.
french fries have
But this becomes fertile ground for a vicious cycle: in some specific
become the staple diet
situations, managers may ask to see test cases for some valid
of most of our industry.
In too many engineering reason, and testers may deliver those cases; but soon
organizations, testing is providing test cases becomes a habit, and then a tradition, and then
fat and slow. someone starts using the term “best practice” as if habitual behavior

”
had won some sort of world championship. The goal of good testing
becomes displaced by a blind mandate.

Test cases are not evil. Neither are french fries nor chocolate candy.
The problem is the obsession that shoves aside the true business of
testing. Chocolate-covered french fries have become the staple diet of
most of our industry. In too many organizations, testing is fat and slow.
We would like to break the obsession, and return test cases to their
rightful place among the tools of our craft and not above those tools. It’s
time to remind ourselves what has always been true: that test cases
neither define nor comprise testing itself. Though test cases are an
occasionally useful means of supporting testing, the practice of testing
does not require test cases.

Test Case Culture and the Factory School

Obsession with test cases is not just a habit, though. It is embedded in
a culture.

Aaron once ran a small experiment at a software testing course he was

attending. He asked fellow attendees whether they write test cases
before they start testing. He expected the answers to range from “Yes,
of course” to “No” with a healthy dose of “What do you mean by test
cases?” thrown in. To his shock, the majority of respondents just looked
at him quizzically as if he had just asked them whether they wear
clothing to work, or whether they hold their breath while swimming
underwater. Test case writing as a central practice appears to go
unquestioned in a lot of organisations.

A test case culture is not one that merely encourages using test cases
as a tool to support testing. In a test case culture, the test cases are
equated to testing; testing is viewed as a mechanistic, clerical task of
executing test cases (analogous to the mechanistic way that a compiler

! TESTING TRAPEZE | FEBRUARY 2014 33

 turns source code into object code) for the purpose of checking
specific and easily observed facts about the product.

A test in a test case culture is a concrete noun, an artifact you can

point to and say “that is a test.” Once reified in that way, it is a natural
step to treat “tests” as a commodity, like so many sacks of rice. We
often see tests counted and testing progress communicated solely in
terms of such numbers. Bugs are subject to the same reification;
thirteen bugs is one worse than twelve bugs, right? In this culture, bugs
and test cases are linked - after all, test cases find the bugs. If a bug is
somehow found without a test case (a situation test case culture views
with suspicion), we might expect some manager to ask for a new test
case to be created that exhibits that bug.

In a test case culture, the tester is merely the medium by which test
cases do their work. Consequently, while writing test cases may be
considered a skilled task, executing them is seen as a task fit for
novices (or better yet, robots).

This way of thinking is attractive because it seems to allow testing to be

managed with an unambiguous accounting system that makes testers
into fungible resources; as a sort of factory of testing. Hence we often
call this the Factory School of testing thought.

A common attitude about process in factories is that there is one right

way; and that this right way should be defined and followed. But how
that process is discovered is completely outside the scope of factory
thinking. In test case culture, this leads to a cartoonishly simplistic
understanding of test design. A common phrase in that culture is that
we should “derive test cases from requirements” as if the proper test
will be immediately obvious to anyone who can read. In test case
culture, there is little talk of learning or interpreting. Exploration and
“
Exploration and
tinkering, which
characterize so much of
tinkering, which characterize so much of the daily experience of the daily experience of
engineering and
engineering and business, are usually invisible to the factory process,
business, are usually
and when noticed are considered either a luxury or a lapse of
invisible to the factory
discipline.
process, and when
A common attitude about people in factories is distrust. People are noticed are considered
either a luxury or a
unreliable at following the one right way. People are, at best, a
lapse of discipline
transitional technology: they are tolerated until the right kind of drones

!
can be built. But even in the most successful factories you will notice

TESTING TRAPEZE | FEBRUARY 2014

”34
 that the managers don’t consider replacing themselves. At some level,
they acknowledge that human attention and action is required. Indeed,
here the authors find common ground with the Factory-schoolers that
there is such a thing as necessary humanity - except we contend that
humanity is necessary at the level where we perform the test.

Factory Theory Meets Practice

What happens when we try to manage a complex cognitive activity
such as software testing by reifying the activity down to a superficial
representation such as test cases?

Recently Aaron observed a class that aimed to teach the limitations of

test cases. The students were given eight identical test cases and were
instructed to execute them. In Aaron’s opinion, these were reasonable
examples of relatively good, unambiguous test cases. At the end of the
exercise, students were asked for the number of test cases that had
passed, the number that failed, the number unexecuted, and the
number of bugs found.

One of the purported benefits of a test-case driven approach is that

consistency and repeatability are ensured. Aaron expected to see
some variation in the results to demonstrate that these claimed benefits
are unfounded, but the results were even more striking than
anticipated.

Some groups reported

no bugs at all, while
one group found five.
While a few groups
reported two bugs,
upon further elicitation
it was discovered that
they weren’t
necessarily the same
two bugs.

! TESTING TRAPEZE | FEBRUARY 2014 35

 This exercise demonstrated beyond even his initial expectations the
danger of assuming that test cases can be a reliable medium of
conveying testing ideas and reporting on quality-related information.
Each group, even though they had the same test cases to work from
and were instructed identically, brought their own judgements and
heuristics to bear resulting in eight different testing performances.

To report on testing by reducing the individual performance down to a

numerical report of test cases passed vs. failed is at best, of marginal
value, and at worst, potentially misleading. Yet, this is the primary way
testing is managed in a lot of organisations.

This case is not a carefully controlled scientific study, but at the very
least this experience undermines the glib assurances by many testing
consultants and authors, going back to Program Test Methods, the first
book on testing, written by William Hetzel in 1973, that written test
cases provide a strong and stable basis for testing.

Testing is Not a Factory. Testing is a Performance.

Testing is an event; an activity; a performance. If we work backward from the moment that a tester
successfully reports an important problem, we find it results from many overlapping and supporting
thoughts and judgments and experiences. There are innumerable ways that this process can play
out, all of which require human enactment.

Testing is the evaluation of a product by learning about it through experiment; by seeing it in action.
The reason we test is to analyse product risk: the danger that the product will cause trouble for its
users or otherwise fail in some way to fulfill its purpose. In other words, we look for anything about the
product that might significantly impair its value. We are looking primarily for “bugs.” We want to find
every important bug, although there will be no way to know for sure that we have succeeded.

! TESTING TRAPEZE | FEBRUARY 2014 36

 Testing as a performance depends on the credibility of the performer,
which is improved or damaged in every interaction with the team. But
excellent testing is a complex performance that is difficult to teach,
supervise, or evaluate. While novice testers may find some bugs by
romping around like kittens, management needs confidence that there
has been a diligent search for important problems. This confidence
must come, in large part, from the personal credibility and observable
behavior of the responsible tester who vigorously questions the
product.

Can an algorithm exist that will guarantee we find all the important
bugs? No. This is a matter of basic computing theory (see the Halting
Problem) and the fact that bugs are socially constructed by users

“
rather than being something to do with the essence of the product.
Bugs are not “in” the Bugs are not “in” the product. Bugs are about the relationship between
product. Bugs are the product and the people who desire something from it. It’s possible
about the relationship for a bug to be created or resolved just by changing the stakeholder.
between the product And apart from every other problem, a total bug identification algorithm
and the people who would require a complete, unambiguous, and up-to-date specification
desire something from that is accepted by all stakeholders... and when was the last time you
it. saw one of those?

” When we test we are exploring the relationship between the product

and values about the product. To do so, we must make many social
judgments, including judging the importance of testing a specific
situation, and judging the meaning and importance of potential
problems.

This seems hard to accept for many casual observers, though,

because many kinds of bugs seem so obvious and uncontroversial.
Indeed it is possible to create algorithms to detect specific kinds of
problems that possess identifiable and predictable characteristics.
Many such checks are already built into compilers and application
frameworks.

However, even if all imaginable checks are performed, there is no

theory, nor metric, nor tool, that can tell us how many important bugs
remain. We must test - experiment in an exploratory way - in order to
have a chance of finding them. No one can know in advance where the
unanticipated bugs will be and therefore what scripts to write. The

! TESTING TRAPEZE | FEBRUARY 2014 37

 “
performance of testing unfolds forward in time, like a swarm of ants
foraging for food. This is a constantly shifting picture. No one can know in
advance where the
Although a tester can and should prepare to perform, testing proceeds unanticipated bugs will
in ways that cannot be predicted simply because of two things: there is be and therefore what
always more testing to do than we can afford to do, and we don’t know scripts to write.

”
where the bugs are until we discover each one. This unpredictability
requires each tester to be prepared to live in and react to the moment,
regardless of any specific plan. That’s why, just as flying an airplane,
doing surgery, or playing football are rich, complicated performances,
testing is too.

Why Testing Cannot be Literally Encoded

Distilling test ideas from tacit mental models; explicitly and precisely
describing test cases; writing them down: let’s call that “encoding.”
Encoding means the expression of ideas, explicitly, in the form of some
sort of code (such as written words or software). Test case culture
insists on encoding testing to such a degree that only trivial aspects of
the process should remain undeclared. It insists that encoding is
practical and desirable - moreover, that it is necessary; that it is
irresponsible not to encode testing.

But this is not so, because testing cannot be encoded.

There is no support in scientific or engineering literature for the idea

that testing can be encoded. Despite years of searching, the authors
are aware of no studies that have ever shown that testing should be
written down, or even that it can be written down. In fact, the opposite
is more the case. See the Sciences of the Artificial, by Nobel laureate
Herbert Simon for a deep treatment of this topic. Simon shows that
perfect rationality is unavailable to us in any but the simplest situations,
and explores the nature of design processes as “bounded rationality”
requiring heuristic solutions. Or perhaps Introduction to General
Systems Thinking, by Gerald Weinberg, who shows that observing and
describing systems requires us to simplify them, and that there is no
algorithm for knowing how to do that without losing something that
might be important. Or look at The Social Life of Information, by Paul
Duguid and John Seely Brown, who tell how copy machine repairmen
learned their craft not by reading the official documentation but rather
by socializing with each other in free-form ways. Or Things That Make
Us Smart, by Don Norman, who shows how adding a cognitive artifact

! TESTING TRAPEZE | FEBRUARY 2014 38

 (such as a test case document) to a process changes the process in
potentially unpredictable ways. If you even glance through any of these
works, you will see a rejection of mechanistic, reductionistic,
algorithmic ways of conceiving and controlling complex systems,
including social systems.

In performance terms we consider a test, in its noun form, to be the act

of configuring, operating, observing, evaluating some part of a product
in the service of a test project. So, what testers call a test is a process
embedded in a larger process called “testing” that includes reading
specifications, attending meetings, acquiring equipment, etc. The
scope of a test is elastic. It may involve the speculative exploration of a
product, or something as simple as checking the result of a function
call. There is no objective method by which we can draw sharp lines
between individual tests: it is purely a matter of convenience and

“
There is no objective
method by which we
can draw sharp lines
context how you choose to delineate them.

Testing has many levels, all of which contribute to the success of the
testing enterprise, and very little of which can be encoded:
between individual
1. A person is born. Yes, it’s important to start here. Each of us has a
tests: it is purely a
specific genetic, environmental, and cultural foundation that means
matter of convenience
and context how you we approach testing with a certain mix of talents and a certain
choose to delineate temperament. The fact that James is mathematically inclined leads
them. him to be biased in favor of analytical modeling the things he tests.

”
Other testers may approach the work in a more intuitive or social
way. There is no such thing as purely objective and unbiased
testing. Two test designers, unlike, say, two car engines, cannot be
analyzed and compared in terms of any universal model of testing
performance. Testing talent and temperament cannot be encoded.

2. A person learns to test. Some skills useful in testing are ubiquitous

among adults. Others come with general technical or scientific
education. Some are technology-specific, and some are specific to
testing itself. Learning to test begins in childhood as we play and
interact with our world. Testing skill can be acquired in a variety of
ways, but deep systematic training in testing is difficult to obtain.

There is no governing body for the testing field. Therefore, there is

no generally accepted Body of Knowledge, or taxonomy of required
skills. Commercial certification programs are controversial, but even

! TESTING TRAPEZE | FEBRUARY 2014 39

 “
if one accepts their view of testing, they do not even attempt to
assess practical skill. It is not unusual for people with no significant It is not unusual for
experience in testing to become a certified tester. This is not people with no
possible in reputable fields such as medicine or air transport. significant experience
in testing to become a
The result is that testers vary quite a bit in their practices and their certified tester.

”
grasp of different aspects of testing. Each tester’s education is local,
conditioned by the idiosyncrasies of specific technologies,
companies, and projects. And on top of all that, much of testing skill
is comprised of inherently tacit skills such as questioning,
collaboration, and systems analysis, which cannot be made explicit
and mechanical.

No one even attempts to encode the fine details of their own testing
skills.

3. A tester joins a project. Joining a project is a complex social event.

As part of that process, we learn whom we serve. We come to
understand the scope of our mission as testers. We commit to the
project. This context conditions everything we do as testers. While a
mission statement and elements of context can be written down,
there are innumerable ways that one might interpret those things
and act accordingly. There is no calculus for determining how all
that influences testing.

A tester’s sense of and response to context can be sketched,

perhaps, but not fully encoded.

4. A tester learns the product. Each of us must construct a mental

model of the product, its context, and its uses. A more familiar way
of saying that is we have to learn all about it, and the result of that
learning is a mental structure from which we can design tests. This
model itself cannot be encoded in any explicit form (it’s neurons,
baby). But we can, if we choose, produce some formal and explicit
projection of our mental model.

Therefore while some of our learning can be encoded, most of it will

never be, for at least two reasons:

a. We have no algorithm or mechanism for doing a “brain dump”

that accurately reflects the state of our knowledge about anything.

! TESTING TRAPEZE | FEBRUARY 2014 40

 That means we can never rule out the possibility that there is a fact
we know about and yet have not put into our model.

b. Even if there were such a mechanism, the totality of what we

learn is overwhelming. For instance, to list our true expectations for
the behavior of a browser that displays a simple webpage, such as
Google’s home page, would require prohibitive time and space.

Oh, and that doesn’t include an even bigger dynamic: the process
of learning the product is also testing. The product is operated,
observed, and evaluated during the learning process. Therefore,
even to the degree that some aspects of testing can be formalized,
that can only occur after a substantial amount of un-encodeable
learning and exploring work has already been done.

Therefore, the bulk of our product learning, and testing while

learning, cannot be encoded.

5. A tester enacts the testing according to some idea. This is the act
of experimenting on the product, apart from all the processes of
preparation that support or inform it.

It is traditional to divide testing into test design, test execution, and

result evaluation. Test design may be further partitioned, perhaps
into coverage modeling, data modeling, procedure design, oracle
design, and tooling. Perhaps this tradition has not served us well in
one respect: it seems to imply that there are clear divisions between
these activities, and that they are independent of each other. This is
not the case. Although when training testers it often helps to focus
on each of these in isolation, the practice of testing brings them
together in an evolving, exploratory process.

“
This interplay cannot be encoded. The most we can do is write
extensive notes about our thinking in every session of testing, but
Although when training
testers it often helps to writing it down, beyond a certain point, interferes with testing. And
focus on each of these even perfect notes about our thinking would not be an encoding of
in isolation, the practice the process of that thinking - in other words we can’t write a
of testing brings them program, while we are working, that duplicates the workings of our
together in an evolving, minds.
exploratory process.
Picture the process of testing: You look; ponder; try something. You

!
”
TESTING TRAPEZE | FEBRUARY 2014
see what happens and ponder that. You have a question, then

41
 conceive of an observation that might answer the question. And so
on. Skilled testers perform hundreds of what might seem like
discrete tests in a session. Few of them need to be repeated. Most
tests performed are informed by the results of the previous test. The
value of a test may not be known until it has been performed, or
possibly much later.

Testing seems encodeable because we can crystallize - from out of

this thinking-learning-trying soup - acts of configuring, operating,
observing and evaluating the product. These moments are
embedded in our evolving concept of risk and of the status of the
product. But to an outsider, who is not privy to the workings of a
tester’s mind, they may seem to stand alone. James Bach and
Michael Bolton, in their Rapid Testing Methodology, call such acts
“checks” if they can be performed, in principle, by a machine (http://
www.satisfice.com/blog/archives/856). It’s useful to talk about
checking because it is a task, embedded within testing, that might
be accelerated or substantially supported with automation.

But we must always remember that checking does not represent or

comprise testing itself, just as a hammer does not comprise
carpentry. We can encode a check, by definition. We cannot encode
the process of conceiving, designing, implementing, re-evaluating,
or judging the meaning of the results of performing a check.

Some would say we can encode testing simply by recording

keystrokes or videoing the test process. Those recordings can be
helpful, but they are mere echoes and hints of the testing thought
process. They don’t encode the richness of the bug seeking and
finding intelligence and experience. At no time, when replaying
keystrokes, will your test tool stop and say “wait a minute, I think I’m
looking at the wrong thing.”

Expectations also cannot be encoded. In James’ classes, he

demonstrates that by asking students to list their expectations for
the output of a simple and well-understood feature of an everyday
product. Then James proceeds to list dozens of expectations that
each student agrees with - but did not think of listing. If testers who
try very hard to list expectations can’t do it completely even for the
simple functions, it is outrageous to think that testing can be fully
encoded.

! TESTING TRAPEZE | FEBRUARY 2014 42

 “
Therefore, with the exception of certain acts of fact checking,
Test case culture is a enacted tests cannot be encoded.
ceremonial approach
6. A tester reports, explains, defends, and amends testing. Testing
to testing. It is, quite
doesn’t end with the output that the system-under-test produces.
simply, fake testing.
The results must be made relevant to the project. This process of

” reporting bugs and status and concerns happens throughout the

test performance process, and it influences that process. It is
probably not sensible to separate this process from the operational
performance of testing.

A test report may be partially encoded, but there are innumerable

judgments to be made about what to say and what keep silent.
Reporting involves responding to questions, too. There is no way to
encode an algorithm for that. The act of reporting may spur testers
to redo or add to the testing performance, and that also cannot be
encoded.

Yes, bits of testing can be encoded. We can use encoding to create

useful anchor points for the test process. We might use test cases or
other kinds of lists, diagrams, or references to formalize parts of
testing. These should be considered tools that support testing, not
testing itself.

We contend that factories don’t apply to testing. While industrial

factories are productive (say what you want about how the iPhone is
manufactured - you can’t deny that it IS manufactured) testing factories
do not work. Testing is not manufactured. Testing factories are a big lie.
Test case culture is a ceremonial approach to testing. It is, quite simply,
fake testing.

But How Can Fake Testing Seem to Work?

Answer #1: testers may be secretly not faking it. Brian Osman dubs this practice “stealth testing.”
This is skilled testing, done for the good of the project, kept hidden due to a management culture that
demands performance while at the same time mandating processes that undermine performance.
Stealth testing, while well intentioned, helps to perpetuate the test case myth. This is a double-edged
sword. If stealth testing finds important problems, and finds them quickly, the tester doesn’t get the
credit; the approach they actually used doesn’t get the credit. The publicly avowed process gets the
credit.

Answer #2: the product may be good enough even with poor testing. Quality comes mainly from

! TESTING TRAPEZE | FEBRUARY 2014 43

 developers who do a good job. A wasteful test process might amount to little more than a sanity
check, and yet the product simply has no important, deep bugs to be found. And face it, many
products are pretty bad, and yet still put out there to torment users. The market for software is not an
efficient one with respect to quality.

Answer #3: it is easy to shift the blame for it not working. Ironically, the inefficiency and
ineffectiveness of test factories can be used as an excuse to invest more in them. Once a gullible
business has been convinced that testing must be structured in test cases, then any problem that
escapes testing seems due to not having enough test cases. If you believe in test factories, any
problems are either down to the factory not being big enough, or bad people who are sabotaging it.
Yet, behind the expensive high walls of test case documentation and the publicly avowed processes
that go with them, broken practices of testing can easily hide, and there is little incentive to improve.

Toward a Performance Culture

A performance culture for testing is one that embraces testing as a
performance, of course. But it also provides the supportive business
infrastructure to make it work. Consider how different this is from the
factory model:

• Testing Concept: Testing is an activity performed by skilled people.

The purpose of testing is to discover important information about the
status of the product, so that our clients can make informed decisions
about it.

• Recruitment: Hire people as testers who demonstrate curiosity, enjoy

learning about technology, and are not afraid of confusion or
complexity.

• Diversity: Foster diversity among testers, in terms of talents,

temperaments, and any other potentially relevant factor, in order to
maximize testing performance in test teams.

• Training: Systematically train testers, both offline and on the job, with
ongoing coaching and mentoring.

• Peer-to-peer learning: Use peer conferences and informal meetups

to build collegial networks and experiment with methods and tools.
Occasionally test in group events (e.g. “bug parties”) to foster
common understanding about test practices.

! TESTING TRAPEZE | FEBRUARY 2014 44

 “
• Openness: Foster the ability to narrate, explain, and defend testing
Testing depends upon performances. Create a culture of normalcy about working together
testers who have pride and sharing work.
and integrity in their
work, and who strive to • Transferring work: One tester may take over from another with the
learn their craft. help of concise documentation, discussion and demonstration, or
simply by starting from scratch. Among skilled testers, this is rarely

” the problem that non-testers fear it will be.

• Personal Excellence: Testing depends upon testers who have pride

and integrity in their work, and who strive to learn their craft. Part of
the reason performance culture is not more accepted in the industry
is the lack of trust by management that testers will perform.

• Team Integration: Foster a mutually supportive attitude between

testers and development. As trust develops, everyone’s performance
becomes more fluid and collaborative.

• Preparation: Detailed and meticulous planning is rarely cost effective

in a high innovation environment such as software development, but
that doesn’t mean we can’t benefit from good preparation. Learning
about tools and technologies and developing test ideas in concise
form is part of performing at our best.

• Responsiveness: We recognize that time is of the essence. We look

at the product as soon as it is available, and if someone taps us on
the shoulder and asks "How did the testing go?" we strive to answer
with useful information, confidently and immediately.

• Cyclic, Exploratory Process: Performing feeds on itself. When we

test, we are also uncovering better test ideas as we go.

• Agility: In performance culture, agility is easier, because we aren't

traveling with all that baggage of documentation. That means we can
respond more rapidly and productively to changing context.

• Metrics: Metrics may be used to provoke inquiry, but do not use them
as the basis of decision rules to control a social system such as
testing. Any metric put in place by management to control people will
be used by people to control management.

! TESTING TRAPEZE | FEBRUARY 2014 45

 • Documentation: Prefer concise • Tools: Use tools under the direction of testers
documentation, such as lists and mind-maps, in ways that augment any aspect of tester
that are less expensive to produce and performance. Do not equate checking tools
maintain. Develop a discipline of personal with human cognition.
note-taking.
• Stopping: Testing is finished when the clients
• Management: Test leads must supervise junior of testing feel that every important question
testers or have them work with senior testers about the status of the product has been
until they are ready to take full responsibility for answered. This feeling is arrived at by
their own performance. First level management discussion of the testing and test results
must be involved in testing on a regular basis. throughout the project.
Give autonomy to qualified testers to choose
The Context-Driven testing (CDT) movement,
their style of work. Celebrate successes, but
has, for years, been promoting a humanist,
also celebrate honest, hard-won failures.
performance-oriented vision of testing. There are
• Process control: Focus on heuristic rather now two international organizations and several
than algorithmic process controls. Focus on conferences devoted to CDT. Although CDT is
discussion rather than numbers. Focus on not against any practice, it is against
trusting people who have earned credibility methodological chauvinism. The practices we
rather than on inanimate controls and use should be the practices that work well and fit
surveillances. If more formal controls are the context. It is through ongoing study and
needed, consider using an activity-based skeptical self-examination that we free ourselves
approach such as session-based or thread- from bad habits and inappropriate practices.
based test management.
After more than 40 years of trying, the factory
• Regression Testing: We may use automated approach to testing has not solved the world’s
checking tools to help detect obvious testing problems. Enough is enough. Abandon
problems at each build. These tools are the swamped, lumbering barge of test case
supervised by testers who take responsibility culture. Re-discover testing as an intellectual
for them. n addition to any checks, however, pursuit. The complexities and risks of our world
regression risk may require the tester to enact demand that we do this.
new tests, or refresh performance of previous
The authors thank Michael Bolton and the
tests.
Testing Trapeze review team for their invaluable
review and comments.

Re-discover testing as an intellectual pursuit.

“ The complexities and risks of our world demand that we do this.

”
! TESTING TRAPEZE | FEBRUARY 2014 46