Goldilocks and Test Automation

I’ve struggled with the right level of automation for a long time and alluded to a level of salvation here.  Since that post, I’ve moved on to engage automation using Given-When-Then (and its collateral ideas) much like a sailor engages a lighthouse at night.  In my most recent experience, it was more like a sailor is on the bridge watching the light and I’m below deck providing interpretation and direction.

My testing team created an excellent suite of automated regression tests for the stored procedures used in our application.  But I’m struggling with the number of tests created for this suite.

It might be that my definition of “regression tests” is more economical than that of my teammates and I own that communication error: I probably was not clear on what I thought this task should produce.  But it also goes to the heart of what I believe about automation tests.

In 2007, I, like many others, was enamored with automated tests against the UI of my applications.  A steady diet of automation tools over the succeeding years (starting with HP QTP) were both valuable and frustrating.  I eventually gave into the frustration because the tests I, and many others, wrote were fragile and numerous.  My response was to limit what I automated because inflicting such a level  of maintenance on a product team was just wrong.

I watched as other teams continued to indulge in not only creating these tests but to develop frameworks around frameworks that provided some theoretical value and, in my opinion, dubious claims of a positive ROI.  I was unimpressed with what appeared to be the emperor (in their “new clothes”) dining happily on over-heated porridge.

With training and some practice in Given-When-Then and its description of business outcomes, I started to see practicality again in test automation.  A test that evaluates business outcome is, in my opinion, more stable, valuable, and maintainable than several poorly designed automated UI tests.  Then, there are two questions.  Always.  Two.  Questions.

Can I automate this test?  Of course, I can automate just about any test.  The avalanche of software utilities that pop up around tools for test automation is evidence of over-indulgence in coding.  Not testing.  The second question is most key:  Should I automate this test?

I advocate for tests designed for simplicity and to exercise isolated business outcomes.  The tests should be motivated by risk, value, and maintenance.  Upon a review of these criteria, I stand the best chance of balancing tests (the most value) and the code (that which must be maintained to deliver the test).

My team and I will meet to review our regression suite.  I look forward to exploring differing definitions of “regression test”.  I welcome their perspectives that could help all of us discover the combination of test and code that is “just right”.

My GWT Identity Crisis

There are a few things I’ve not tried yet.  Sadly, possibly, I’ve not tried karaoke, and I’ve not tried kite boarding.  I am however getting my first experience in applying Given-When-Then in describing product behaviors on a real project.  I’m very excited at the possibility of merging business behavior with automation.

I engaged Specification By Example, I attended hands-on training, and I have participated in periodic drive-by creations of GWT scenarios.  I felt very ready.  Hurtling into some of our project’s first story cards, I was shocked to review GWT that contained behavior AND, take a deep breath, implementation details.

As an example, I offer a scenario from an API I wrote for a website.

GIVEN active flights exist,

WHEN the website requests a list of active flights,

THEN the reply includes all active flights

AND their unique identifier,

AND their altitude,

AND their last update.

This GWT represents my interpretation of what I have learned and, as noted above, my limited experience.  Here is how the same scenario might appear if I violate one of the most fundamental rules.

GIVEN the active_flight field indicates true,

WHEN the website executes GET against the activelist API,

THEN the response contains an HTML table detailing the flight’s GUID,

AND the flight’s altitude as an integer,

AND the last update in a UTC format.

In my opinion, this is more than a violation of the “no implementation details” rule.  This description locks the behavior to a specific column of a database, defines the verb and API name, prescribes a specific rendering method, and calls out individual field data types.  Its as if a developer provided the information they needed to implement the scenario to an analyst.

While this is clearly an extreme example, I recently read such a scenario.  When I asked, the reason implementation details were included was because this is a “technical story card” (different from a “business story card”), and because it is easier to provide all the information in one place.

(ahem) Bull.  Shit.

In my opinion, this method of practicing and segregating GWT ignores one of its most quintessential and valuable aspects: collaboration.  Yes, I read Specification By Example and attended training.  I learned the mechanics but in both cases I left with something more.

I perceived an overwhelming feeling that during a conversation about product details between an analyst, a tester, and a developer, these few people begin to share not only their ideas but their understanding.  It IS NOT the GWT itself that encourages automation or reduces defects rather it is those precious minutes where an honest, introspective, and respectful dialog clarifies the behavior of a simple product.

Just writing GWT provides a small window into a product.  Collaborating over the value of a specific business behavior encourages teamwork and quality where everyone can take pride and feel their contribution matters.

 

Lead Testers and Lead Testing

As I exit a project with disappointing results for me personally, I’m motivated to re-invent my role by combining select experiences from past projects.  At the start of my next project, my conversation with project team members might take on this flavor.

My role as test lead is to facilitate testing, advocate for smart testing, and to guide testing within our project.  Developers may create and execute unit tests, and analysts may evaluate product definitions; both are testing activities.  I encourage them to collaborate early with project testers for feedback, to augment the testing, and to identify automation opportunities.  The goal is not a fully tested product; the goal is verifying business behavior that delivers value at a rapid pace.

I go into the project with a position of achieving 100% automation.  I’m not advocating for 100% rather I’m setting a goal of 100%.  I want to use automation to demonstrate behavior not just after a product is complete, but frequently as we make changes to the product.  I want our automation to provide confidence to our team so they can continue to add functionality, explore different implementations, optimize the product for performance, and verify we still deliver business value.

I will foster a sense of community as we explore the details of the design.  Together, I want to create definitions of our products that we can build and evaluate quickly, products that we are proud to demonstrate, value that distinguishes us in the marketplace.  As equals at the drawing board, I want our team to collaborate, to challenge, and to bring their diverse talents for a common purpose.

Our project will likely required services and integration with other parts of our enterprise.  I want to meet with them soon, share our story and engage their team to assist us.  I believe a proactive approach helps them prepare better, makes them more willing to participate, and contributes to our mutual success.

The Shift Left
The role of Test Lead is no longer confined to leading a team of testers, it must lead all testing within the project.  With “Shift Left” continuing to gain momentum, the Test Lead’s focus and responsibility grows to guide the testing of the project such that product definitions are clear, small bits of code are exercised, and value is demonstrated.  Success will depend more on your ability to support team building and encourage effective collaboration than testing or technical skill.

Technology Stacks (yawn)

I was assigned a Test Lead role for a project that, among other goals, is investigating SPA.  I thought I’d jargon-up my opening line because it looks all cool but it has a distinct “Emperor’s New Clothes” feel.  Single Page Application (SPA) is all the rage as of this writing (I may be a few months behind – “all the rage” is hard to maintain in software development).

Plainly said, I ain’t impressed.  In fact, I think we’re coming full circle from (former “all the rage”) desktop computing applications.  You may recall the shrink-wrapped, shiny boxes containing the coolest new application, requiring twenty minutes of disk swapping to install.  I grant you today’s installation is far simpler and faster but in the end, it’ll be a desktop application hosted in your browser.  Yup – full circle.

What really frosts me is developers talking about new frameworks and utilities they are using.  Not having familiarity with the names (which for my money are nowhere nearly as cool as Lotus 1-2-3, Wordstar, Paradox, or Sidekick), I did a bunch of research and experimenting.  I was able to install < insert popular framework>, its recommended testing tool <insert popular testing tool for popular framework>, and run a few tests.  As Biff exclaimed in “Back to the Future II”, there’s something very familiar about this.

It doesn’t stop there.  Since there are soooo many files used to create an SPA, you need a utility to bundle them.  These also have unique names but that hasn’t fool me.  Lastly, in a penultimate salute to developer vanity, there is even a utility to convert (sorry – they call it transcompile but I think that’s a pile of sh-t) source code to javascript.

If I have this straight, there is a “new” IDE for coding, a “new” utility for “compiling”, and a utility for bundling all to support SPA.  Maybe it’s just me but I’m pretty sure I can do all that in Visual Studio today.  With the Community version, I can even do it for the same price.

This journey started because we wanted to pursue SPA.  Its shiny and new.  As we experimented with it, we discovered the technology stack.  It’s as if we were walking along in the woods, and found a neat stick we could use.  When we picked it up, the other end disturbed a bee’s nest.  When you pick up a stick, you always pick up both ends.

 

Test Automation is an Investment

I submit that test automation can improve your project’s pace.  I realize that some believe it helps detect defects or that it can be developed independently from project execution.  I recommend you reconsider those positions.

Ever since I read Specification By Example, my understanding and attitude towards test automation has changed.  It did not change how I design or build test automation, it changed how I see its purpose.  For this post, I refer to test automation in a spectrum from evaluating newly checked in code to that used in regression testing.

I’ve been pitching what I believe is a familiar notion around test automation in testing circles.  But it was suggested to me that I provide some context.  Put simply:

The purpose of test automation is to enable frequent evaluation of application behavior in the face of change.

I did not arrive at this conclusion by myself.  Adzic’s book was a great catalyst to think about test automation more, and to discuss it with other test engineers.  A context took shape for me and its present form is my declaration above.

Defect Detection is a Narrow Focus
Test automation costs too much to make it beneficial in just detecting defects in the short term.  Indeed, test automation developers find defects many times when they are developing (not executing) the automation.  That means that by the time the test automation is completed and verified, it’s possible someone else could have located that defect AND corrected it.

I recommend you invest in your test automation today.  You will realize the real value when it detects a defect in a succeeding iteration, or a succeeding project.  When it does, it demonstrates how recent changes have introduced errors in existing behavior.  In that manner, it is acting to preserve your business intent.

Independent Development Costs More
Let’s say you wait to develop test automation.  At some point in the future (the next iteration or the next project), the testing team is discovering defects in code that was completed AND working.  The current pace begins to slow:

• Defects in existing code must be corrected so testing can evaluate the newly deployed code.
• There is a good chance that correcting defects will introduce more defects.
• When the code is operating again, you may find defects in the newly deployed code.

If there was an investment in test automation, frequent execution would detect changes to existing functionality at the time they were introduced.  That is, defects would be detected when developers execute the test automation, and they would correct them.  This saves time because the project team continues building products rather than fixing products; the project maintains its pace.

Your Investment Pays You Back
Test automation pays dividends in protecting your investments in building and expanding business functionality, and detecting unintentional changes to business behavior early.

• Time is saved because the team executes test automation often to detect behavior changes early – before the products are deployed for testing.
• System and integrated testing is more effective because simple defects are corrected before deployment for testing.

Lastly, the project team builds confidence in their products because the testing team can spend more time exploring products broadly, deeply, and for risk.

The Explorer

Imagine you are an explorer.  You often explore jungles and forests, and occasionally cities and villages.  People regard your exploring skill highly but with some reserve.  One day, a builder comes to you who wants you to explore a bit of pasture.  They are interested in changing the pasture so that many people might live there.

• Given your skill and experience in exploring, when you go out to the pasture, then the builder wants to know your opinion on having people live there.

You decide to visit the pasture.

After a day’s journey, you arrive at the pasture.  From your vantage point, you can see it is bordered on one side by a large hill.  While there are many groups of healthy trees, there are also inviting clearings.  You think that if people lived here, they would enjoy the area.  As you walk around the pasture, you take notice of the types of trees, variety of animals and insects, and appreciate the small river.

Upon your return, you describe what you saw and what you found.  In your opinion, people would be able to live in the pasture.  The builder asks many questions and you respond with your experience in the pasture.  The builder asks if you found any bugs.  You reply that you found many bugs for they were plentiful on the ground and on the trees.  The builder thanks you and gives you a special token of their appreciation.

A few weeks later, the builder asks you to explore a city.  They are interested in making changes in the city.

• Given your skill and experience in exploring, when you go to the city, then the builder wants to know your opinion on making changes in the city.

You decide to visit the city.

After a day’s journey, you arrive at the city.  From your vantage point, you can see a few buildings, neighborhoods, streets, and people.  Beyond the city limits, there is a forest.  You think that if the city had some changes, the city would readily adapt.  As you walk around the city, you take notice of the population, the clean streets, the commerce, and the services.  You recall the builder asked about bugs so you also pay special attention to the ground and the few trees you find.

Upon your return, you describe what you saw and what you found.  In your opinion, the city was ready for changes.  The builder asks many questions and you respond with your experience in the city.  Some of their questions about infrastructure and services were challenging so you express your regret in not gathering more information about them.  The builder asks if you found any bugs.  You reply that you found no bugs.

Surprised, the builder asks, “Where did you look for bugs?”

“On the ground and in the trees.  The ground was often covered in concrete and there were few trees.”

“Did you look under the concrete, or in the buildings?”

“I did not.  Since I found bugs on the ground and in the trees in the pasture, that is where I looked for them in the city.”

The builder explains how the city is different from the pasture.  While the bugs were easy to find in the pasture, bugs in the city may hide in buildings or houses.  They may also hide in the way services operate in the city.  The builder suggests you review information about how cities are different from pastures.

The builder expresses their disappointment in your report.  They believe the changes to the city will be delayed because your information may not be accurate.  They thank you and leave.

A few weeks later, the builder asks you to explore the city again.  Not to be disappointed, they ask about your understanding of cities.

“I have explored cities occasionally and they contain many things that I don’t understand.”

The builder thanks you and says they know of another explorer more familiar with cities.  They will ask them to explore the city.

Imagine YOU are the explorer.  What would you do next?

 

 

An Experiment for 2016

I missed writing.  I think about it every day but it’s like drooling over a pizza ad.  It’s nice but pizza with the family is much more satisfying, leaves an impression, and has more impact.  My inspiration to return came from Julie Zhuo reminding me of many great things I have read about writing – thank you Julie!

I focused a bunch of time on building a new web site for my hobby.  I like building it and I like testing it.  It gives me an opportunity to stretch my testing muscles lest they atrophy from the daily experience and negotiation of leading a testing team.  I enjoy the test lead role but when I need to do something to relax, I code and I test.

With a large and successful project behind me, I had an opportunity to catch up with our testing community.  Notably, I finally read Specification By Example (SBE) by Gojko Adzic (I realize I’m late to the party on this one).  I was seven chapters deep and realized I’ve miss so many opportunities to implement SBE ideas.  Then, there was a post on an internal blog recommending all testers read the book.  There was a momentarily satisfaction that my actions and the universe had aligned.

Ever since I started automating test cases, I was apprehensive about their existence beyond the end of the project.  We live in a very dynamic industry so the solution we implemented would inevitably change.  My tests would fail when change hit some critical mass.  I resolved to do better but better was to recommend caution around what was automated.  My constant challenge was writing automation that could survive change AND deliver a valid result.  I was a little successful but not always satisfied.

SBE crystallized my reasons for cautious automation.  I was unable to articulate why because I hadn’t found a suitable alternative.  But the logic and wisdom of automating business outcomes resonated with me.  The business outcome has a much lower frequency of change so automation makes sense.
It also defined why I had such apprehension.  What I have seen is automation of not only business outcome, but automation of test cases that provide little information just because they could be automated.  I also saw automation evaluating lower level functionality – which is valuable – but it was executed from the UI.  Both of these added technical debt and impeded project pace.

With the new year and a new project, I’m ready to experiment (it feels more like a hurricane of righteous purpose but I’ll calm down after a bit).  I discovered that our Requirements/Business Analyst community also favors the approach.  So far, I’ve suggested SBE to two other projects and the response is positive.