Pairwise testing, also called all-pairs testing, picks test cases so that every pair of values from any two parameters appears together at least once. It answers a practical problem: three browsers, three operating systems, two payment methods and two kinds of user already make 36 combinations, and every new parameter multiplies them.
A worked example
The sample has four parameters, Browser (Chrome, Firefox, Safari), OS (Windows, macOS, Linux), Payment (Card, PayPal) and User (Guest, Member), and one forbidden combination: Safari on Windows. Every combination would take 36 tests. The tool generates 10 tests and its verification reads “Verified: all 36 pairs are covered by 10 tests. 1 pair is excluded by your rules.” That is 72.22% fewer tests than every combination, rounded down.
Strength 3
At strength 3, every triple of values from any three parameters appears in some test. That covers failures that need three values together, at a cost: the sample needs 16 tests instead of 10, and large models grow quickly.
Forbidden combinations
Some values can't go together. A forbidden line lists assignments that must never all appear in one test; a single assignment removes that value from every test. Combinations no valid test can hold are reported as excluded by your rules, not as gaps. If the bounded search can't settle a combination within its step limit, the tool says it could not be covered and why; it never drops one silently.
How it works
The generator is deterministic and greedy: it builds one test at a time, giving each parameter, in order, the value that can still cover the most combinations, ties going to the earlier value. The same input always gives the same tests. Finding the smallest possible set is a hard problem, so the result is small but not guaranteed to be minimal, and other tools such as PICT may produce more or fewer tests. A separate verifier then checks every combination, rule and row.
What it doesn't do
Pairwise testing is a heuristic. A widely taught idea is that many failures are triggered by one parameter or by two together, but that is not a guarantee: a bug that needs more values together can be missed unless you raise the strength, and the tool doesn't know the expected results. Pick parameters and values from the requirements, add boundary values and risky single cases by hand, and use synthetic test data, not real personal data.
Privacy
The model is processed in your browser and never leaves it. Like every page on this site, the page records a visit when it loads, but not what you type.
How to use
- Enter one parameter per line, as Name: value, value, or load the sample.
- Add forbidden combinations if some values can't go together, and choose the strength.
- Read the verification line, then copy or download the tests.
Features
- A separate verifier checks every result before it is shown as complete.
- CSV, Markdown, JSON and text output, with spreadsheet formula protection.
Common uses
- Browser and OS matrices. Cover browser and system pairs without running every combination.
- Configuration testing. Pair settings such as database, cache and region, or feature flags.
- Form and API parameters. Combine dropdown values or request options.
- Test plans. Paste the CSV or Markdown into spreadsheets, test management tools and wikis. RoboQA runs website tests you describe in plain English.
Frequently asked questions
What is pairwise testing?
A way to choose test cases so that every pair of values from any two parameters is tested together at least once, with far fewer tests than every combination.
What is all-pairs testing?
Another name for pairwise testing.
How many tests will I get?
Without forbidden combinations, at least the product of the two largest value counts, and usually somewhat more. The summary shows that lower bound and the actual count.
Does pairwise testing find all bugs?
No. It covers every pair of values, not every combination, and it doesn't know the expected results.
What is the difference between pairwise testing and testing every combination?
Every combination grows by multiplication: the sample needs 36 tests. Pairwise covers every pair in far fewer, 10 for the sample, but skips most larger combinations.
What do forbidden combinations do?
They keep values that can't go together out of every test. The pairs they rule out are listed as excluded, not as missing.
Will the results match PICT?
Not necessarily. Different methods give different sets, sometimes smaller, sometimes larger; this tool makes no claim to match any other tool.
What does strength 3 mean?
Every triple of values from any three parameters appears in some test. It needs more tests than strength 2.
Is my data uploaded?
No. The model is processed in your browser and never sent anywhere or saved. The page itself records a visit when it loads, as every page here does.
Need to test a whole website, not just a snippet?
RoboQA tests real pages and user journeys in a browser: logins, forms, clicks and full flows. See what RoboQA can test or compare plans.