Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Metrics
September 2010
During the past few years, I have been involved in producing and it3.
reporting testing indicators. I have observed good and less good
indicators, learned from my own and others successes and mis- Note: Following the spirit of this article, these are only indicators
takes, while reaching agreement with stakeholders on format. and not absolute facts. Like any indicators, you should use them
only as aids for the evaluation of your indicators.
I would like to share my own experience, both good and bad, and
Useful Indicators:
give indicators to assist you in evaluating the usefulness of your
own testing indicators. • Indicators produce a discussion that initiates decisions
I am going to specifically discuss my experiences with quality in- A good indication that our indicators are useful is when the-
dicators (bug counts) and with test execution indicators. se three parts - Indicators, Discussion and Decision - exist and
connect in a cause and effect connection. Otherwise indicators
Before we start, we must define what “useful” means in our case. are just wasting time for the people who produce or view them.
My understanding of “useful” indicators has been shaped by trial Discussions without decisions might indicate that the indicators
and error and by reading the thoughts of others1. were interesting. but did not provide the data for decision ma-
king (unless we have a decision-making problem). If we have indi-
The most simple definition that I can define as a baseline for this cators which initiate decisions by themselves, we probably have a
article is: problem of over- valuing the indicators.
Discussions and decisions will not happen each time we extract • When a second mathematical function is involved
the indicators. When defining the indicator sets, imagining what-
if scenarios and assessing whether they are capable of inducing Some indicators introduce a weight function that unifies a few
discussion that will lead to decisions, can be a good way to assess items that are measured together. For example, if there are some
them. components of our software which are more risky than others,
we will use a function that will assign a weight to components
Useless (or less useful) Indicators’ Indications:
according to our risk level analysis.
• Indicators are happy while the project is sad
This approach can be useful since the indicators looks more
While corrections, exceptions and complimentary informati- unified and complete, but the mix between the subjective inter-
on are essential to complete the picture drawn from indicators, pretations and the numbers can be misleading. In the correct
when time after time you have to complete the picture by dra- context it is definitely OK, but in some contexts, especially when
wing an opposite picture to the one that is apparent from the indicators are a sensitive subject, it’s better to have the bare num-
quantities indicators, your indicators might be wrong, or the bers followed by a verbal interpretation for discussion, so that the
whole activity is going in the wrong direction. For example, week numbers themselves will not be subject to debate.
after week getting green “pass” bars for the test execution while
you have to express your concerns that the product itself is not I hope that I have given you some good indicators and food for
useable or too buggy can indicate that you are publishing the less thought so that in future you will ask your own indicators, “Do
important indicators, or that a major activity of your test team is you work for me (and the project’s success)? Or do I work for you?”
focused on a less critical path of your product quality. – Such questioning can be a good start to better and more useful
indicators.
I could say the same about “Sad Indicators in a happy project”.
While working on this article a software developer that I know
told me about a project she was involved in. The project mana-
ger asked to delay the release because the high-severity defect
numbers were too high, but inquiry proved that the defects were
not classified correctly and they were only minor cosmetic ones.
The project manager understood his mistake and suggested to
release the product, but the developer objected. Not due to the
number of open bugs, but because the type of bugs showed that
the test team spent the time on finding minor, redundant bugs
instead of testing the relevant functionality.