HADI loop
Also known as: HADI cycle, Hypothesis-Action-Data-Insight loop
A HADI loop is a four-step cycle for testing a growth idea: Hypothesis, Action, Data, Insight, run in that order so a team learns something specific from every test instead of just running it.
Each letter is a step, and skipping one is what turns a test into a guess. Hypothesis states what the team believes and why. Action is the smallest version of the idea worth running. Data is what actually happened, recorded before anyone forms an opinion about it. Insight is the decision the team makes because of the data, stated plainly enough that someone outside the meeting could read it and understand what to do next.
Example
A team believes that adding a testimonial near the pricing table will raise the checkout conversion rate. The hypothesis: “Adding one testimonial above the pricing table raises checkout conversion, because buyers hesitate on trust rather than price at this step.” The action: add the testimonial to 50% of traffic for two weeks, leave the other half unchanged. The data: the test group converts at a higher rate than the control group over that period. The insight: trust, not price, was the blocker at checkout, so the next test targets a second trust signal lower on the page, such as a security badge near the payment field.
The specific conversion numbers are left out here on purpose, since a real result depends on traffic and baseline rate that vary by site, not on a fixed figure.
How to use it
Write the hypothesis down before the action starts, not after the data comes in. A hypothesis written after the fact tends to fit whatever happened, which defeats the reason for testing in the first place. Keep the loop short, usually one to two weeks, so a team runs many small loops instead of one long one that ties up the whole quarter on a single idea.
Common mistakes
The most common mistake is running the action without a written hypothesis, which makes it impossible to say afterward whether the result confirmed or denied anything. A second is calling every observation an insight, even when the data was inconclusive, which fills a backlog with vague notes nobody acts on. A third is changing more than one variable in the action step, so a positive result cannot be traced back to the change that caused it.
A HADI loop is the engine behind most work tracked on a KPI tree: the tree shows which number needs to move, and the loop is how a team tests its way toward moving it.