Testing scientifically - Scientific method in testing

Scientific method in testing

begins here

Aristotle is believed to think that women had less teeth than men. At the moment we can't say for sure if he could not count or women in Greece of his time experienced problems with nutrition.
A fact.

To tell the truth best ever definition of scientific method is already provided by wikipedia and according to it

-- method should be empirical and have measurable evidence [2]
-- method should be proving or disproving a theory/assumption [3]
-- approach and analysis should be as unbiased as possible [4]
-- knowledge should be documented and sharable [4]
-- results should be reproducible [4]

All these map on what we know about testing perfectly. In fact any good testing professional has been doing this for years but did not always realise that there was a solid scientific background all the time. Let's look at each of the above points closely.

Well, being empirical. Unless you are in SQA department your job 100% emprirical anyway.

Measurability. In this case measuring is not about centimeters, but should be used in a bit wider sense, say, you do not expect your test to return some ok result, you expect it to return either specific number of lines, or no error message, or what not. But you need to be specific.

Being an applied technique testing is quite remote from most of theories and deals with more low-level thing like requirements and assumptions (derived directly from said requirements). As such your test is supposed not just to do something abstract but to either prove or disprove an assumption. Suppose that under certain circumstances some button is  expected to tunr blue. Assumptions are:
-- button must look blue given preconditions are met
-- button should not look non-blue
-- etc

It is possible to talk endlessly on human biases how testers being human suffer from them severely. Here are some of them
-- fear of looking incompetent (and not reporting a potential issue)
-- fear of aggressive reactions (and not reporting a potential issue)
-- being lazy to deal with consequences (and singing it off in hope that by the moment it fires you'll be far away)
-- etc

Sharing knowledge. Ahh. Knowledge means power and unless you are lucky enough to work for a company with right corporate culture you will have to beg for it until you build your own informational capital and be able to trade. There is a lot of literature on business and functional analysis with a log of explanations of why documenting is important so I'll do it briefly here:
-- if information is documented and shared you do not have to waste anybody's time and effort to get it
-- if information is documented and shared it is less likely to be lost if key persons leave
-- if information is documented and shared it is much easier to make sure your tests are well aligned with requirements
-- etc

Reproducible results. Tiny but crucial difference between a bug and a glitch. To prove there is a bug it is important to be able to show that a very specific cause leads to a very specific result. Being inaccurate in this usually leads to non-reproducibles and I-do-not-wanna-know-it-works-on-my-machine stuff. Golden rule: make sure to reproduce it twice at least before you claim it works ok or it is a bug. As I mentioned earlier there are both false positive and false negative pitfalls.

-------------

Footnotes:
[2] "To be termed scientific, a method of inquiry must be based on empirical and measurable evidence subject to specific principles of reasoning." Source: http://en.wikipedia.org/wiki/Scientific_method

[3] "The chief characteristic which distinguishes the scientific method from other methods of acquiring knowledge is that scientists seek to let reality speak for itself, supporting a theory when a theory's predictions are confirmed and challenging a theory when its predictions prove false." Source: http://en.wikipedia.org/wiki/Scientific_method

[4] "Scientific inquiry is generally intended to be as objective as possible in order to reduce biased interpretations of results. Another basic expectation is to document, archive and share all data and methodology so they are available for careful scrutiny by other scientists, giving them the opportunity to verify results by attempting to reproduce them." Source: http://en.wikipedia.org/wiki/Scientific_method

-------------

To be continued in part 3

Testing scientifically - Testing and quality

Testing and quality

Quality is conformance to requirements, not 'goodness' or 'elegance'
usually attributed to Phil Crosby


Testing is measure taken in order to make sure you get exactly what you expect. Or very much like it. Or has some of expected features. At least. Testing deals with how you put your expectations into words and as such is closely connected with message-sent-is-not-the-same-as-message-received thing. This is where requirements come in and analyst plays his/her part.

Ok. Suppose we agreed on meaning of words and shared goals, how do we prove we got what we wanted? And this is where testing starts and test specialist appears on the scene. In an everyday common context test stands for "checking". In a narrower scientific context it implies a number of actions that prove some statement. Do you remember functional logic? If something in a statement contradicts actual reality or common sense then statement is considered to be false. Otherwise it is said to be true. It works pretty similarly with the testing except that reality and quite often common sense are replaced with requirements. If test results contradict requirements then test is failed. If test result are in perfect harmony with those requirements then test is passed or, in other words, we got what we wanted (or agreed to think that we wanted this).

NB: Sometimes results turn out ot be false positive or false negative. This may happen due to a number of reasons like poor understanding of what is going on, ambiguos requirement or something being wrong with the test design. We are dicussing this a bit later.

Roughly all tests may be split into two categories: tests that try to confirm that anything works alright and tests that try to prove there is something wrong. Scientifically these two approaches are called verification and falsification and their history goes as far back as Karl Popper's writings and even further then that. It is important to remember that this in no way contradicts or replaces specifict methods of building test coverage such as equivalence partitioning, boundary values analysis, cause and effect, error guessing or exhastive testing.[1]

Why mess with science when the only thing we want is to make sure it works during sales demo? In fact this question is crucial and the answer is simple. If your goal is just to sell something that somewhat works then testing is not necessary. Testing is only required if you expect to get something specific and getting what you expect constitues quality. In case you need quality messing with science is unavoidable.

-------------

Footnotes:
[1] Based on this page (Russian with automatic traslator): About testing

-------------

To be continued in part 2