Grey is made up of Black and White

There's an idea that has been on the back of my mind for a while now.

In certain testing (and development) circles, one hears about "good enough" software. I get it. Nothing's perfect and no software can ever be completely bug free, so the best you can hope for is to have caught and fixed the most important bugs (to your stakeholders) before you ship. Or something like that. I'm paraphrasing.

The thing that irks some people is that "good enough" is not carved in stone. It has a human element to the decision-making process that some bean counters might prefer to do without.

I think I've come to appreciate that definition over time because I've had the benefit of seeing many projects of different types for different companies go out over the last decade. With experience, you develop a sense of when things are in your comfort zone (i.e. acceptable risk) compared with when they are not (i.e. this risk gives me the willies - ship it if you dare but I'm going to duck and cover).

That part I'm okay with. That's not what's troubling me. The thing I'm wondering about is where do you start to teach someone this?

So, when I first started to Test, I started by working on some automation tests, and then moved to developing some written test cases and test plans. Over time, I came to appreciate that the brain tends to move quicker than paper and that there are some testing practices out there that can help you test more and provide you with more good information and feedback in the time given on a project. You can choose to spend that precious quality time with paperwork or do more testing. The choice is yours.

Fast forward a few years. I've chosen to teach Exploratory Testing to people who are new to the Testing field. It's an interesting and challenging approach and I love how it gets you to think bigger than you've thought before.

So, you decide to test something. Did the test "Pass"? Did it "Fail"? How do you know? Are you sure? Can you check? ... and so on...

This leads to discussion of Oracles - i.e. a principle or mechanism by which you can tell something is a bug. Funny thing about Oracles - they're like siblings. Each one is unique, they're similar/related in some ways but different in others, and they all think they're the most important and want to be in front.

As a tester, you need to decide *which* is the most relevant Oracle to help you decide if something is a bug or not.

For example, I often explain to new testers that the app might match the specification when you compare them, so is that a Pass? The simple answer is "yes"... assuming the only thing you care about is comparing the application to a particular claim. So what's the problem?

Well, the problem I have is the assumption that the spec, the claim, is correct in the first place. What if they both match each other but I don't care because it goes against what I *expect* as a user? That is, when you use computers for a while you develop a general understanding of how many apps work in similar ways for similar functions. And then along comes some hot-shot spec-writer who thinks they're going to be awesome and develop something completely new and different for ... oh, I dunno, the 'Print' function.

"Okay," you say to the spec-writer, "you might do that.. but then again, reinventing something that everyone has a certain idea of how it should work might lead to mistakes, frustration and user/operator errors that are not so good."

So, back to the tester, the app matches the spec, but they're both wrong I say because neither matches my expectations. Where's the Pass? Where's the Fail? It's not so easy anymore. In fact, you might say there are shades of grey when it comes to determining Pass/Fail - especially if you find yourself in the situation where there are 3 or more applicable Oracles. (eek!)

That's where I think my line of thinking went wrong a while back. It's not about shades of grey, it's clearly about determining which sibling, which Oracle, is the most applicable here. You *have* to pick one. There has to be a binary, logical Pass/Fail answer to the question "Is there a problem here?" One Oracle needs to edge out more than the others. You may not be able to make that decision on your own, and that's okay. But someone, or some team of people, needs to reach a consensus about whether or not something is a bug according to some oracle that they care about most in the given situation.

At the end of the day, it doesn't matter if your brain is firing on all cylinders or none, a "black and white" decision needs to be made about whether or not something is a bug. Computers are logical things. I think it makes sense that we need to apply logic in this situation too.

But hold on here, where does this fit in with the whole "good enough" software thing I started talking about a few minutes ago? That doesn't sound very logical. In fact, it sounds kind of like the opposite, doesn't it?

Why, yes, it does sound illogical, and yet there is logic in it too. The idea came to me when standing in a Wendy's looking at a poster while waiting to order lunch. The poster had a picture made up of smaller photos.. a photo mosaic it's called. (Aside: Google search turns up lots of cool info on these.)

And then it hit me. Aha!

Determining "good enough" is like finally recognising the picture of the quality of the release when you spend every day looking at all of the smaller 'quality' photos. These dozens, hundreds, or thousands of 'black and white' decisions that are made over the course of a project (i.e. the bugs you find) paint a picture. Some people pay attention to these, while others do not (to their detriment).

Good development teams come up with Release criteria that make sense and often contain human elements in the final go/no-go release decision. Different project team members may have different pieces of the picture. As a Test Lead/Manager, you bring information about the tests that have been run, the risks that were covered, and the kinds of bugs that were found, reported and fixed (and the ones left outstanding). Sometimes you might even have other metrics that you feel are important in helping you recognise "good enough" when you see it.

I like the idea of a photo mosaic as an analogy for good enough software. You may never get to see the complete picture, but you might not need to either. When the picture is too fuzzy, then you should recognise that you might not be collecting all the information you need to help make a good timely decision.

As a tester, you need to be concerned with all the logical decisions in the Pass/Fail criteria world that helps you identify bugs on a daily basis. But don't get attached to any one particular bug, because in the grand scheme it may not be as important as you think it is. (It *might* be, but be aware that it might *not* be too.)

As a Test Lead or Manager, you need to step back from all the separate little details and ask yourself what patterns you recognise in the complete body of [testing] work that has been performed? One might even use the cliché "can you recognise the forest for the trees?" but I like my photo mosaic analogy better. ;)

I like how the colour grey is made up of mixing the colours black and white together. And similarly, good test management decisions, which may seem 'fuzzy' to some, are also based on the understanding of all the little logical decisions made by the testers on the team.

These are different skills (between tester and test lead). I haven't been specifically teaching testers to see the big picture and patterns in work beyond their own. Where does that fit in? When? I don't know that I'm comfortable doing that just yet.

I think I got the teaching testing part down, so I guess the next logical step is looking into the skills required to manage good test teams.

Any thoughts or recommendations? Is it just something that you learn by trial and error, by osmosis, working on different projects for different companies? (Nahh, that's how I learned.. it takes too long.)

This, I think, is going to be the next big need for training in the testing field. Let's say for a moment that you've trained testers in good/different testing practices. So how do you train the managers or leads to manage in new, intelligent ways to complement the productive, agile testing effort? Part of it will be in developing new test management tools, but part of it, clearly, will be in skills development and education too. I just don't see that available right now.

I'll have to keep my eyes open the next time I'm out for lunch. ;)

1 comment:

  1. Here's something that you might find helpful.

    In the end, if you're a tester, you don't get to decide whether something is a bug or not, unless you're also the product owner. But you do get to report on why it might be a bug. If you're reporting a problem, and you include why you think it's a problem (that is, the oracles that tell you that it's a problem), then the product owner can add your information to his/her basket of other information that informs the decision as to what matters. You, yourself, don't have to make the overall pass/fail decision for a given test. You might offer "passes by this oracle, fails by that one", and leave the decision to those responsible for making it.

    ---Michael B.

    ReplyDelete