The 7 principles of software testing, with real examples

The 7 ISTQB testing principles, each with a plain definition and a real example, from why exhaustive testing is impossible to why tests wear out.
Back to blog

The 7 principles of software testing, with real examples

Short answer

The seven principles of software testing, from the ISTQB Foundation syllabus, are: testing shows the presence of defects, exhaustive testing is impossible, early testing saves time and money, defects cluster together, tests wear out, testing is context dependent, and absence of defects is a fallacy. They tell you what to test, when, and how much.

The principles come from the ISTQB Certified Tester Foundation Level syllabus (version 4.0.1, 2024, section 1.3). They are general guidance, not rules to tick off: each one explains why a common testing habit works or fails. Below is each principle in plain words, an example, and what it changes in practice.

Article details
Question
What are the 7 principles of software testing?
Audience
Testers, developers and anyone preparing for ISTQB Foundation
BetterQA team
50+ engineers, founded 2018
Clutch rating
4.9 / 5 from 71 reviews

The seven principles in one table

#Principle (ISTQB v4.0.1 name)In one sentenceWhat to do about it
1Testing shows the presence of defects, not their absenceFinding bugs proves they exist; finding none does not prove there are noneReport coverage honestly instead of promising "bug free"
2Exhaustive testing is impossibleYou cannot test every input, path and environmentChoose tests by risk and priority
3Early testing saves time and moneyThe earlier a defect is found, the cheaper it is to fixReview requirements and designs before code exists
4Defects cluster togetherA small number of modules usually hold most of the defectsGive risky or bug-prone areas more test time
5Tests wear outRepeating the same tests finds fewer new defects over timeReview and refresh suites, add exploratory testing
6Testing is context dependentA bank app and a game need different testingPick techniques to fit the product and its risks
7Absence-of-defects fallacyA product with no known defects can still fail its usersCheck that the product meets real user needs

A note on names. Older ISTQB syllabi and many articles call principle 5 the "pesticide paradox" and principle 7 the "absence-of-errors fallacy". They are the same ideas under earlier names, so a search for either name lands here.

Each principle, with an example

Definitions follow ISTQB Foundation Level syllabus v4.0.1, section 1.3, in our own wording. The "on our own website" lines come from testing betterqa.co in October 2026.

1. Testing shows the presence of defects, not their absence

Testing can show that defects are there. It cannot prove that none are left. Even a large passing suite only lowers the chance of undiscovered defects; it never takes it to zero.

Example. You test a checkout and the discount field fails. That proves a defect exists. If every check passes, you still cannot say the checkout is flawless: you can only say these checks passed.

On our own website. We found a bug that made the FAQ questions on 37 of 39 pages turn pink on hover. Testing proved the bug was there; it could never have proved the other pages were clean.

2. Exhaustive testing is impossible

Testing every combination of inputs, preconditions and paths is not feasible, apart from trivial cases. Teams use risk and priorities to decide where to spend a limited amount of testing.

Example. A form with 10 fields that each accept 10 values has 10 billion combinations. Nobody runs those. You pick representative values with techniques such as equivalence partitioning and boundary value analysis, and spend the most effort on the fields that handle money or personal data.

On our own website. For 11 new industry pages we could have checked hundreds of screen sizes. We checked two, 1440px and 390px, because that is where visitors are, and looked for the defect that matters on a phone: sideways scrolling.

3. Early testing saves time and money

Testing activities should start as early as possible in the lifecycle, because defects removed early do not get carried into later work. This is often called shift left, and it includes reviewing requirements and designs, not only running code.

Example. A tester reads a user story and asks what happens when the coupon has expired. Changing a sentence in the story costs minutes. Changing code that was built, tested and released around the wrong assumption costs far more.

On our own website. The first version of our client logo strip had 15 logos and gold seals. On a mockup it took one look to see it was too busy. Fixing it then cost minutes; fixing it live would have meant another deploy and another round of checks.

4. Defects cluster together

A small number of components usually contain most of the defects found in testing or in production. This is sometimes described as an 80/20 split, but that figure is a rule of thumb, not a measured law, so check your own defect data.

Example. In a shop, bugs tend to gather around the complicated parts: pricing rules, discounts, tax and stock. If the last three releases all broke the basket, the basket gets more test time next cycle.

On our own website. Service-page headings showed in Times on 41 pages. It was not 41 bugs: every page asked for a font the site never loaded. One cause, one fix, and a census of all pages to prove it was gone.

5. Tests wear out

If the same tests are run again and again, they become less and less effective at finding new defects. The older name for this is the pesticide paradox, by analogy with pests that become resistant to the same spray.

Example. A regression suite that has been green for two years is either guarding a stable product or no longer looking at the places where the product changed. Review it: retire checks that can never fail, add cases for new features, and schedule exploratory sessions in areas the scripts do not reach. The syllabus adds a caveat: repeating the same tests can still be useful, for example in automated regression testing, where the aim is to confirm that nothing has broken rather than to find new defects.

On our own website. A deploy check of ours reported "no drift" while a file on the server was empty. It had passed so often that nobody questioned it. We now compare file size and fingerprint after each deploy, a check the old one could not fail.

6. Testing is context dependent

Testing is done differently in different contexts. The right approach depends on the product, the risks, the lifecycle and the people who will use it.

Example. A healthcare app that stores patient records needs heavy security, privacy and audit-trail testing. A casual game puts more weight on performance, device coverage and how it feels to play. Applying one checklist to both fails one of them.

On our own website. Our ranking pages keep two copies of their content: the one visitors see and an older one the page no longer shows. Checking the wrong copy passes every test and changes nothing on screen, so we only trust checks that read what the visitor sees.

7. Absence-of-defects fallacy

Testing every specified requirement and fixing every defect found can still produce a product that does not meet the needs of its users. Checking a product against its requirements is verification; the syllabus says validation, checking that it meets what users and other stakeholders need in their real setting, should be carried out as well. A system can be close to defect-free and still be the wrong system.

Example. A booking flow works perfectly, but the people who use it wanted to compare two dates side by side, which it cannot do. Zero open bugs, and the release still disappoints. Usability testing and user acceptance testing exist for this gap.

On our own website. On our own site, the calendar loaded, every link worked and no error appeared. But the card beside it never told a buyer what the call was for. Bug-free is not the same as useful.

Using the principles together

The principles pull against each other, and that is the useful part. Because exhaustive testing is impossible (2), you choose by risk, and the places that cluster defects (4) are usually the highest-risk. Because tests wear out (5), the suite you chose last year needs a review this year. Because a clean run proves little (1) and a bug-free product can still fail (7), you finish by checking the product against what its users need.

If you are building a test strategy, read how to build a robust QA strategy and why early testing matters. For what a late defect costs, see bug fixing costs throughout the SDLC. Studying for the exam? Our ISTQB exam advice covers the rest of the syllabus.

Where our engineers use these principles

On client projects we plan test effort by risk (principle 2), keep a map of the modules where defects keep appearing (4), and review regression suites on a schedule so that checks that never fail are retired and new scenarios are written (5). Exploratory sessions cover the areas automation does not reach.

If you would rather have a team apply this to your product, our managed testing services cover the planning, the testing and the reporting.

BetterQA team working together at the long table in the Cluj office

Frequently asked questions

The ISTQB Foundation syllabus lists seven: testing shows the presence of defects, exhaustive testing is impossible, early testing saves time and money, defects cluster together, tests wear out, testing is context dependent, and absence of defects is a fallacy. The table near the top of this page gives a one-line meaning for each.
It is the older name for the principle "tests wear out". If you run the same tests repeatedly, they stop finding new defects, like pests becoming resistant to one pesticide. The fix is to review the suite regularly, write new cases for changed areas and add exploratory testing.
Because the number of input combinations, paths and environments grows far faster than anyone can run. A form with 10 fields of 10 possible values each already has 10 billion combinations. Teams use risk-based testing and techniques such as boundary value analysis to cover what matters most.
It means there is no single correct way to test. A medical device needs rigorous safety and traceability testing, a marketing site cares more about layout and speed, and a payments API cares about correctness and security. The industry, the risks and the users decide the approach.
That phrase usually points to the seven quality management principles in ISO 9000: customer focus, leadership, engagement of people, process approach, improvement, evidence-based decision making and relationship management. They are about running an organisation, not about testing, and they are a different list from the seven testing principles.
The seven principles are not stages. The software testing life cycle has its own phases, and our STLC guide walks through the six we use, from requirement analysis to test closure.

Where these definitions come from

ISTQB Certified Tester Foundation Level Syllabus v4.0.1 (2024), section 1.3, published by ISTQB. ISO 9000:2015 for the quality management principles. The examples are ours.

Need help applying these principles?

Tell us what you are testing and we reply within one business day. Our engineers can add testers to your team or run QA for you.

Book a call Managed testing services
Share the Post: