daily post

Data quality is one of the biggest problems with data science projects,

I’ll be talking about these at the #AutomationGuild, here’s a quick list:

– Accuracy. Is the data accurate in the context to be used

– Validity. Is the data fresh enough, still valid?

– Consistency. Data from different sources / time frames matches

– Completeness. No parts of data are truncated / missing

– Uniqueness. Enough data to uniquely identify records

– Timeliness. Data being collected at the right time & processed in a timely fashion (efficient enough)

More on the conference here:

http://amp.gs/Dydu

#QsDaily #BigData #DataScience #Testing

For large enterprises with many interconnected applications, generating test data can be a challenge.

Even more so for big data projects, which is where masking comes in..

Take a copy of production DB from different products around the same timeline and mask the data for GDPR compliance.

Surely easier said than done, but can be very effective.

IMHO three main requirements to build this:

1. A quick guideline to identify what data will be classified as ‘sensitive’ and ensure it’s GDPR compliant.

2. A masking platform which masks value the same way across, helping with consistent data

3. Creation and usage of test environments is efficient and fit for purpose

#RedefiningSoftwareQuality #TestData #Automation

One of the main reasons for flakiness in automation is test data.

Therefore use mocks to run scripts executing in the CI pipeline.

For most folks, automation scripts are executed only in nightly runs with data created on the fly.

And unfortunately these tests are mostly UI tests, another blunder.

I gave a talk yesterday in which I said – 50% of your automation scripts should run in the CI pipeline triggered on each pull request.

And all these tests should use mocks instead of real data from downstream systems / components.

(I hope it’s needless these tests are on the services / API layer..)

So don’t forget your mocks when automating.

#RedefiningSoftwareQuality #QualityTransformation #Automation

“If I know my stuff, my employer will pay me well” that’s a lie we’re taught from childhood.

And here’s my take on this:

The industry is not there to fulfill your needs, it’s there to protect the entity’s interests.

If you want to get paid, knowing your stuff is a given,

Then there’s a heck lot more to do before you get paid well.

As engineers mostly we’re oblivious of social skills, and that’s the next big piece.

Learning how to deliver value and sell it appropriately

A few people I follow who have great content around this:

Seth Godin

Grant Cardone

Thank you Umair Sheikh for inspiring this post.

How to become a ‘well paid’ tester?

I call it “The next generation” tester and here are the main skills…

Quality Engineering
– Agile practices and DevOps culture
– Ability to match testing goals with product’s goals
– Figuring out KPIs / OKRs which help ‘deliver value’
– Enterprise & Vendor management

Testing
– Risk based testing
– Defining a practical, efficient and effective test strategy
– BDD practices
– Context driven testing

Automation in test
– Algorithm design aptitude and programming proficiency
– Automation framework design
– API / UI / Mobile Automation
– OAT (Performance and security basics)

Continuous Integration
– Branching & merging
– CI tools (Jenkins / M. Azure)
– Provisioning environments Containers / VM / Cloud)

Technical Leadership- Technical acumen
– Team leadership
– Collaborating with architects- Thought leadership

Details of each in the linked article

#RedefiningSoftwareQuality #NextGenerationTester #TestersGoingTechnical

DevOps sounds sexy, but it’s not a piece of cake.

And certainly is not just about Jenkins..

It’s foremost a mind set to deliver at speed and ONLY WHAT’S VALUABLE.

This prioritization of what’s valuable is so under-rated, and one of the biggest cause of failure.

Working on anything that’s not priority is ‘waste’, and waste is what kills it all.

I talk about this briefly in the linked article

Picture – Slide from @Jez Humble workshop showing the ‘cost of delay’ of different features in a specific case study.  (It’s it ridiculous how much we would waste trying to implement all features there!)

#RedefiningSoftwareQuality #EliminateWaste #DevOps #Lean

I’ve been reading about psychological safety for while, particularly, how you implement it in a team.

Imagine my luck, had the opportunity to ask that question from @Jez Humber in a workshop..

What I understood mainly was: The leader’s behavior builds or destroys it.

If the leader punishes people for failing or calling out what they feel is wrong, then no matter what, the culture goes down the drain

As an example, he talked about Etsy where a person on the team broke production deployment while following the process, the person was given an ‘award’ for highlighting the problem instead of being penalized.

Link to post here:
https://www.ryn.works/blog/2017/06/17/on-failure-and-resilience

If you want your team to perform, build psychological safety.

To build that, the leaders must be seen as supportive to failures, whistleblowers, people who highlight problems instead of shooting the messenger.

#RSQ #PsychologicalSafety #leadership #HighPerformingTeams

I very much dislike automation trainings starting with teaching Selenium,

For starters that’s the wrong place to do automation. One should focus more on API tests and do a handful UI checks.

Secondly, automation is a means to an end, not an exercise for the heck of it. Therefore talk about why to automate.

Thirdly, in some cases folks learning automation don’t really learn how to test well either. If your testing is bad, your automation is just going to amplfy that bad.

Lastly, without knowing how the underlying architecture of the product works, IMHO one cannot truly learn how to test well.

So next time you plan an automation training, make sure you get the fundamentals right first.

#RedefiningSoftwareQuality #TestAutomation #Training

Anyone who has written automation scripts has faced flakiness, and we all hate it to our bone.

But everyone defines a flaky test a bit differently..

For me, a flaky test definition could be “a test running multiple times under the same conditions gives different results”

How would you define flakiness?

#QsDaily #automation #flakytests

Mostly when companies talk about transforming, automation come up as the most hot topic.

While that certainly is important, without your quality strategy & foundations it’s not going to work.

Immediately people start running towards buzz words and new shiny tools.

Knowing what quality means for your product, what to test and how to test should precede any automation effort.

If you used so ship garbage at a slow pace, with new shiny tools and terms it’s going to garbage coming out double time..

#QsDaily #Automation #Testing