It’s the question behind almost every process-improvement claim, every A/B comparison, every “the new supplier’s parts look better” observation — and it’s the one question a lot of teams skip, because picking the right statistical test to answer it has always felt like its own specialized skill. Guess wrong, and you can talk yourself into a conclusion the data never actually supported.
The Hypothesis Testing Suite exists to close that gap: 14 tests across every major category of comparison, wrapped around a plain-language layer that tells you not just an answer, but why it’s the right test to be asking.
The problem with “just run a t-test”
Most people default to whatever test they learned first, regardless of whether their data actually fits it. Comparing more than two groups? A t-test won’t tell you the whole story — that’s what ANOVA is for. Data that clearly isn’t normally distributed? The classic tests start making assumptions your data doesn’t meet, and there’s a non-parametric alternative built for exactly that case. Comparing pass/fail rates instead of measurements? That’s a different family of test entirely.
The suite covers all of it — Compare Means, Compare Variation, Compare Proportions, Compare Categories, Non-Parametric alternatives, and Correlation & Regression — and includes a detection engine that looks at the actual shape of your uploaded data and suggests which test fits, rather than leaving you to guess from a menu of names.
An answer you can actually explain to someone else
Every result comes back as a plain verdict — Reject H₀ or Fail to Reject H₀ — translated into a sentence about what that actually means for your process, not just a p-value sitting by itself with no context. For anyone whose statistics training is a few years behind them, there’s also a built-in primer on the two decisions every hypothesis test ultimately comes down to, so the tool doubles as a refresher, not just a calculator.
Verified the way a statistics tool should be
Every one of the 14 engines — every t-test, every ANOVA, every non-parametric alternative — was checked against independently computed results before release, the same standard applied throughout this whole toolset. If you upload data partway through building a case for a process change, an in-app editor lets you correct a bad entry or add a missed observation and re-run detection immediately, instead of starting the upload over.
Statistics doesn’t need to be the bottleneck between “we think this changed” and “we can prove this changed.” That’s the whole point of building the suite this way.
