alikhalid

Testing in big data area has typical challenges,

A big factor is quality of data ingested.

The analysis results has a heavy dependency on the ‘quality’ of data ingested (obviously),

What often happens is inconsistency across the data from down-stream sources, missed records, missing data within records etc and other data quality issues.

Unless these issues are flagged at the lower levels, problems creep up and start reflecting in the analytics results.

While having automated data checks on massive data stores might not be an easy job, it certainly is worth implementing.

Have you ever calculated Automation Regression Percentage?

I think it’s one of the most misused metric. Mostly used to show ‘cost savings’ on pretty graphs.

Here’s how to use the metric get some actual value:

#QsDaily #Automation #Regression #Testing #AutomationRegression

A client asked me to ‘Quote a budget and timeline for completely ‘automating testing’ while giving no context of the product.

I refused to work with them (obviously), here’s why:

Even if disregard the ‘automating testing’ ask, which is a BIG ask..

Getting a budget and time estimation of a complete project might be a common ask before any project sign off.

But I consider this the biggest anti pattern to agile.

Decades of practical experience has showed us we are horrible at estimating.

On the other hand I understand we need to budget and calculate what return we might get.

What might be more practical is to have a ball park range for budget and work we can deliver in first three months.

You can extrapolate and calculate the total estimated cost and use for long term budget planning, but remember it’s always going to be wrong.

We get so hung up with the detailed plans & good old

gantt charts (which were invented in WWI) and forget the ground reality.

#QsDaily #projectestimation #agile

Fault injection / mutation testing is among the things I miss from the old waterfall days.

More on fault injection in my talk at QA&TEST conference this month end.

In waterfall days it was okay to spend time on improving testing and coming up with better testing techniques.

After the agile apocalypse, many teams found excuses to get away with perfecting their craft under the inaccurate pretense of urgency.

In my understanding, agile never meant not to get half cooked stuff out of the door, in fact quite the opposite, important activities should be part of the Definition of Done.

The only difference was to break a big project into very small deliverable.

So essentially all the good practices from before should continue, just implement in bite sized pieces.

Fault injection is one of those practices.

#QsDaily #Agile #Testing #MutationTesting

I am furious on hearing a tester saying ‘this is out of scope’.

I’ll be talking about one way to solve this at QA&TEST this month (https://www.qatest.org/ali-khalid/?lang=en).

Most of the time I’ve seen testers say this because they don’t know how to test that feature.

Instead of trying to go under the hood and understand the architecture and code base, they just raise their hands.

What really should be done is understand how that piece of code can be tested.

Yes developers should be doing unit tests, but even then tests on module / component level should be figured out

In the world of embedded systems and especially safety critical devices, letting a functionality go untested is not in the cards.

Using fault injection / mutation testing is one technique which I’ll be discussing at the conference.

Like industry types, testing applications dealing with big data is very different.

One of the biggest quality challenge in the big data space is of ‘Data Quality’.

Big data mostly deals with analyzing large data sets and extracting useful information / insights from it.

The real currency here is data, therefore quality of the data you get will heavily affect the results, hence quality.

What ‘data quality’ would mean for each product & insight will be different.

So the first step is, figuring out the ‘definition of data quality’ for a particular product and the insights we are trying to get.

#QsDaily #BigData #DataQuality

I very much dislike the word end to end testing,

It means different things to different people, instead I use testing the ‘user journey’…

End to end can mean:

1. From the UI to the DB level and back (ends of the tech stack)
2. Can mean a user story executed from the UI
3. A feature (group of stories) demo over the UI
4. The end user’s complete journey which cuts across multiple applications

For me, only no. 4 is end to end, but it’s so hard to explain to certain people.

I find testing the ‘user journey’ much easier to get across to the audience.

And this part of testing is the most neglected, where a user flow is going through multiple systems,

Each team’s feels responsible from the start and end of their product only..

#QsDaily, #testing #softwarequality

Many re-orgs are done to change the culture and how decisions are made,

Often I’ve noticed in the process they end up centralizing decision making even more,

This notion of a select group of people having enough background to make right decisions has to go away,

The typical old management style of information flowing up and decisions coming down, like in the military just does not work.

The complexity of today’s work place is far too much for any select group to be able to make all decisions.

Although the re-orgs mean to decentralize but letting go is just so hard,

Ultimately going even further way from the goal of gaining efficiencies by reducing waste in decision making

If we stick to the fundamentals of reducing waste, making these structural changes might become more easier.

For more, refer “scrum” By Jeff Sutherland

#QsDaily #agiletransformation #scrum

Did you know ‘phased program’ – (waterfall SDLC) was developed by NASA

And inspecting the failed programs, this process was found to be extremely inefficient

The phased program / gated development process / waterfall model was said to be very good at trying to show an over-glorified picture of the software

The focus was on documenting rigorously and proof.

I guess that’s where the Agile manifesto said – we value working product over comprehensive documentation

Many of our teams might have moved into an ‘agile’ model, but this idea of valuing working software over documentation still has not gone away

We insist on over documenting nonsense stuff, and under-documenting where needed.

Classic example – still maintaining those long test cases, and having very less context around stories or importantly writing feature files to agree on the product’s ‘behavior’ with the team

#QsDaily #agile #testing

A guru of mine once said – You are paid exactly what you are worth – I thought he was joking

Looking back, I can say he was right, at least in the long run, here’s one reason why..

I was having a chat with fellow testers the other day talking about tester’s career progression.

The biggest mistake IMHO people make is to not learn how to solve problems..

In today’s world, all businesses are focused on solving their customer’s problems, especially in tech.

Folks hired in these tech firms primary role is to develop solutions to solve these problems.

A few people find this very taxing and unexciting, because they never trained themselves in problem solving.

If you are working in tech, regardless of your position in the org, your primary role is to develop solutions for the target market’s problems.

If you can’t handle that job, you have no place in tech, and let’s then stop talking about career progression (or griping about it).

If you are solving problems, and your organization is not treating you well, you can get a better opportunity and leave.

So, in the end, you are paid what you are worth..