Archive for September, 2010
In response to my recent post on risk based testing fallacies, an RBCS client–I’ll refer to him by his initials AN–wrote to tell us about some fallacies he’s struggling with implementing risk based testing in his organization. He recounted a discussion he had with two colleagues–I’ll refer to them as Confused and Confused Too–who were caught in these fallacies.
Confused said, “Risk-based testing does not control the extensiveness of test design.”
I [AN] was very surprised and replied, “The tester can choose an appropriate test technique as risk mitigation.”
In his statement, AN is entirely right. Confused has fallen into the fallacy of assuming that, because risk based testing is non-prescriptive on test design, it’s silent on test design. Risk based testing does not prescibe the technique, but rather gives guidance on the level of risk mitigation that is required. Often, people use a descending scale for this extent of testing: e.g., extensive, broad, cursory, opportunity, and report bugs only. It’s up to the test engineer to select a test technique–or blend of techniques–that will yield the correct risk mitigation.
He [Confused] doubted that risk-based testing works for test design because the level of risk could not deduce the test types and/or test conditions directly. I think we can design tests using risk-based testing. The level of risk is not an absolute value, but a relative value. We can assign the resources based on these relative values when balanced against the risk inherent in the entire system under test. Therefore I think test design is a very important process for risk-based testing.
AN is again correct, and has diagosed the fallacy here. Confused has fallen into the fallacy of assuming that risk based testing is quantitative risk management. Risk based testing is qualitative, because we (as an industry) don’t have access to pools of statistical failure data such as those the insurance companies have. The test conditions to be covered are the risk items which are identified during the quality risk analysis. The degree to which they are covered is determined by looking at two factors, likelihood and impact. Based on the relative level of risk, we select test design techniques that will give the proper level of coverage.
A test leader [Confused Too] insisted that he was using risk-based testing, but did not have test design documents, only a test plan and test cases. I don’t believe it was risk-based testing.
Again, AN is correct. Confused Too has fallen into the fallacy that risk based testing can be done without any additional structure. In fact, in order to have proper risk based testing, you need some document, tool, or other structure to capture the risk items, their risk ratings, and other ancillary information. Otherwise, you’ll not be able to manage the alignment of the other testing work with the risks and their ratings.
I try to avoid using US-specific slang in my writing, especially sports-related US-specific slang, and most especially US-football-specific slang (since no one outside the US plays that game). However, it looks like one such phrase remains embedded in my writing style, as astute reader Thomas Wagner pointed out in a recent e-mail:
I am currently studying for ISTQB Advanced Test Manager by following your book “Advanced Software Testing – Vol 2. Guide to the ISTQB Advanced Certification as an advanced test manager.” I have a question: You frequently use the term “hand-offs”. What do you mean by this? Examples where used in the book:
Section 1.2, page 3, line 2 “…this is especially true at key interfaces and hand-offs.”
Section 3.3.6., page 171, line 19: “…In general, foul-ups are always more likely during hand-offs”
The ever-helpful dictionary.com defines hand-off as “an offensive play in which a player, usually a back, hands the ball to a teammate.” (In this case, note that “offensive” refers to the opposite of defensive, not that the maneuver itself is likely to offend.) And, indeed, it says that the origin of the phrase is from US football. My mistake, but how to fix it?
The problem is that a precise, universal way of saying “hand-off” might be something clunky like “intra-project deliverable.” That’s certainly not an easy phrase to write or to understand. Any time one distinct group of people (e.g., the programmers) within a project team creates a deliverable (e.g., the software to be tested) and delivers it to another group (e.g., the testers), you have, well, a hand-off. In this example, the programmers have handed the software to the testers, for the purpose of testing.
(The last part of that sentence above, that there is a specific purpose, is also important. A hand-off is not a merely informative delivery, where no action is required on the part of the recipients. The recipients–in the example above, the testers–are required to carry out a specific set of actions with the deliverable. That’s an important part of the concept of a hand-off that isn’t included in the phrase “intra-project deliverable” unless we say “intra-project deliverable given to recipients for the purpose of taking some action with it,” or “transfer of an intra-project deliverable from one group to another within a project that includes a responsibility for a specific set of activities on the part of the recipients,” and now we’re really getting into a long and tortured phrases!)
Whatever we call it, these junction points between groups in a project team are always risky. Mismatched expectations between delivering and receiving parties can result in problems (e.g., not fixing certain bugs that block some tests). Failure to deliver on time can occur (e.g., the all-too-common delay of the start of test execution due to incompleteness of the software to be tested). Failure to deliver something usable for the intended purpose can occur (e.g., the untestable test release). Miscommunications can arise (e.g., the bug report that doesn’t give the programmer enough information to debug the underlying problem).
Given the difficulty of thinking up a good alternative phrase, I’m going to keep using “hand-off”, though I’m glad Thomas sensitized me to the cultural difficulty of the phrase. Call it what you will, any time one group transfers something to another group during a project, take care. Especially as testers, being downstream of just about everything else that happens on a project, we have a lot of opportunities for bad hand-offs.
Risk based testing is a phrase that we hear many times in testing. Many people know many facts and have many opinions. The trouble is, in many cases, these facts are actually wrong or based on a poor understanding of risk based testing, and thus many opinions about risk based testing are incorrect. There are many risk based testing fallacies. In this first post (in what is likely to be a series of occassional posts on this topic), I’ll start with five frequently encountered fallacies.
- Risk based testing is just a method to cut corners (part 1). This whole idea for a series of blog posts came about when someone said to me, “Well, risk based testing means not testing everything.” Well, right. So does every kind of testing. There are an infinite number of tests you could run, and you are going to select a finite subset from that infinite set. The only question is whether you are going to select that subset intelligently, with an understanding of the likelihood and impact associated with potential problems. Risk based testing allows you to do that.
- Risk based testing is just a method to cut corners (part 2). Sometimes when people say this, they mean that risk based testing does not cover all the requirements. Unfortunately, some people have promoted an approach which they call risk based or risk driven testing that involves exactly that: Selecting which requirements not to test based on risk. While in some cases it is appropriate to skip testing some of the requirements, as a general rule we want to cover not only the important risks but all the requirements (at least those which are specified). By ensuring that every requirement has at least one associated risk item and at least one associated test case, you can do so. This is an example of a blended strategy of risk based and requirements based testing.
- Risk based testing is all about technical risk. Some people have put forward this idea that risk based testing is a form of reactive testing where we wait to see what the system does (i.e., no planning, analysis, or up-front test development), then use experience, defect taxonomies, and other aids to predict and find as many bugs as we can in a limited period of time. To me, that approach is just a big geeky bug hunt; it does not cover all of the strategic objectives most organizations have for test teams. Yes, we should consider defect likelihood when analyzing quality risks, but we should also consider the impact of potential defects as well.
- Risk based testing can be done entirely by the test team. Those who believe this fallacy simply analyze requirements or other information, in isolation from other project and product stakeholders, and then test based on that analysis. Sorry, but that’s just a risk-aware form of requirements based testing. What makes true risk based testing truly powerful is the consideration of input from a cross-functional team of project and product stakeholders. When we help clients start doing risk based testing, I always emphasize that getting the right quality risk analysis team together is more important than the right process or templates.
- Risk based testing only influences selection of test cases. It’s true that one major benefit of risk based testing is the smart selection of test cases. However, with risk based testing you can also report test results in terms of residual risk, which makes test status truly clear to non-test project team members. You can also run tests in risk priority order, which maximizes the likelihood of finding important bugs first. And, if you do get squeezed for time, you can triage your test cases based on risk, ensuring that the most important tests get run.
I hope this blog entry has helped to dispel some of these fallacies. I’ll return to it in a later post someday to try to dispel more such fallacies. In the meantime, you might want to check out the videos on risk based testing, found in our Digital Library, for more information about what risk based testing really is and how to make it work for you.
As part of the continuing series of video blog entries, I’m throwing out to the wider software testing and quality community a question: How do we balance the desire–especially in Agile and Lean/Agile projects–to enjoy the efficiency of lightweight record-keeping, while at the same time having enough visibility into project, process, and product status, including testing and quality metrics?
For a little video context on my question, check out: How to Balance Metrics, Lean, and Agile when Measuring Software Testing and Quality
I’d be interested in hearing from people as to how their projects are achieving balance here, and also in anecdotes about when they are not achieving balance.
I recently gave a workshop at the STANZ conference, first in Wellington and then in Sydney. In this workshop, I mentioned that connecting software testing to business value is a key test management challenge of the 2010 decade. (Of course, it’s really been a challenge for the entire time there has been software testing, but it’s a challenge we’ve yet to resolve.) Everyone in b0th audiences agreed, and a number of people offered examples of how this challenge was affecting them.
Earlier this week, I gave a webinar on how to calculate the return on the software testing investment. You can listen to a recording of that webinar if you missed it.
In that webinar, I walked through a case study of calculating software testing ROI. This case study was described in an article originally published in Software Test and Performance magazine, and you can now find the article here on the RBCS website.
After the webinar, a bunch of people sent e-mails saying, “Hey, could you please post the spreadsheet that you walked through during the webinar?” Here at RBCS, we like to say yes to our friends, clients, and supporters, so we did. You can find the free software testing ROI spreadsheet on our Advanced Library now, under the name Case Study Info Appliance Test ROI.xls.
Before you use the spreadsheet, I suggest you read the article I mentioned above. The article explains how the spreadsheet works and explains the case study numbers included in the spreadsheet by way of example.
One key to quality software is the quality of the people involved in creating and maintaining it. One of the tools for increasing the quality of your team is through training of existing employees, which I’ll address in a later blog post. For this post, I want to focus on something that is often confused with training, but actually is (or at least should be) something entirely different: certification.
All IT managers–whether software test managers or other software managers–want to hire qualified people. Certainly, IT certification can be part of the qualification puzzle in many IT fields. IT professionals often use certification in key skill areas to demonstrate their qualifications. However, with all the certification programs floating around out there, how do managers and professionals distinguish useful certifications from pointless wallpaper? In this post, I’ll examine how you can pick the right IT certifications for yourself (as an individual) or for your team and the people you hire (as a manager).
Any certification worth considering will have, at its basis, a body of knowledge or syllabus. This document should describe the skills and abilities that the certification measures. Those people who have mastered most of these skills and abilities (sometimes called “learning objectives” in the syllabus) will be able to earn the certification, usually through some kind of exam.
So, the first and most important step is to determine whether the skills and abilities listed in the syllabus are useful. Does the syllabus relate to your day-to-day work? Will the benefits of achieving the certification—increased effectiveness and efficiency, credibility of the team, etc.—justify the cost?
Of course, it’s possible that your day-to-day work should more closely resemble what is described in the syllabus. This can happen when your organization is not following industry best practices. So, you should also evaluate the source of the syllabus. If the syllabus was written by a broad, international team of recognized, published industry experts, perhaps you should consider moving your practices towards those required for certification. Adopting the certification as a guideline for your practices—and hiring people with the certification—can be a good way to move in this direction.
Selecting a certification developed by a broad team of recognized, published industry experts is important because, in general, such certifications enjoy increased acceptance over certifications developed by a small clique of like-minded people. People in the industry will recognize the names of the authors and developers of the syllabus. To some extent, the credibility and thus value of all certifications rests upon the reputation and credibility of the people who stand behind those certifications.
I also mentioned that the team of experts should be international, because so often now we are engaged in globally distributed work. If you are not working in a globally distributed fashion today, you probably will be soon. So, you need certifications that have a global reach. If you want to hire (or be part of) a global team of certified professionals, a single common certification is key. This way, the whole team speaks the same language and knows the same concepts.
Of course, if you plan to hire people who hold a certification because you believe the syllabus has value, you want to be confident that those people have indeed mastered the topics in the syllabus. This brings us back to the matter of the exam.
Certification exams are a complicated issue, and some ill-informed polemics about exams occur on a few internet web sites. Proper creation of exams is the province of a profession called psychometrics. Psychometrics applies the fields of psychology, education, and statistics to the process of qualifying people through exams. Any legitimate certification body (i.e., the organization developing and administering an exam against a syllabus) will employ professional psychometricians to ensure proper exams.
In evaluating whether an exam properly assesses someone’s qualifications, you need answers to four questions. First, is the exam statistically valid, and can the certification body prove validity? Second, is the exam a quality instrument, free from grammatical and spelling errors, formatting problems, and other glitches that might distract exam takers, and what process does the certification body use to ensure quality? Third, is the exam of uniform difficulty and quality whenever and wherever it is administered, and how does the certification body accomplish uniformity? And, fourth and finally, since exam questions are developed by people, what steps does the certification body use to ensure the integrity of the exams; i.e., that the questions are not leaked to candidates, onto the internet, or to accredited training providers?
This last point—that of accredited training providers—brings us to an important consideration. It is certainly valuable to have training available to support certification programs. Accrediting training, whereby the certification body checks the content of the training to ensure compliance with and coverage of the syllabus, can help busy managers and professions narrow their search for quality training. However, when the accreditation process is opaque, when only members of the certification body offer accredited training, or, worse yet, when accredited training is required to take an exam, you are not looking at a real certification: you are looking at a marketing vehicle for some company’s or cartel’s training programs. You should pick certification programs that have open, transparent processes for accreditation, with a diverse, competitive field of training providers, and which do not require any training at all to take the exams.
Certifications can help IT managers and professionals grow their teams and their skills, if chosen carefully. If you select the right bodies of knowledge, developed by the right people and delivering the right skills for your work, certification can lead to improvements in effectiveness, efficiency, and communication within teams. It’s also essential that the certification body follow best practices in the creation and delivery of the exams. And, if you decide to use training to help achieve certification, make sure to pick a program where the training supports the certification, not vice versa. If you follow these basic concepts, you can obtain good value from IT certification programs, both as a professional and as a hiring manager.