Experimentation News: July 2026

The through-line this fortnight is tools trying to do the boring parts for you: generating the next test idea, moving gate approvals out of the console, and putting the qualitative “why” next to the numbers. Three platforms shipped changes that touch how a small team actually runs its backlog, plus a quieter defaults tweak worth flipping on. Here is what changed and whether it matters to your programme.

Optimizely’s Idea builder turns Opal into a test-idea generator

Optimizely added Idea builder to Web Experimentation, documented on 22 June. It sits inside the experimentation workflow rather than a separate chat: you give it the page, the goal and any research, and Opal returns specific test ideas you can save and refine. Optional inputs like a heatmap image, a qualitative research doc or analytics notes sharpen what it suggests, and it also reads your programme’s past results so the ideas are not generic. For a lean team the risk is obvious, a firehose of shallow ideas, so treat the output as raw material for a proper hypothesis rather than a queue to run as-is. A generated idea still needs a testable prediction and a metric before it earns a slot, which is exactly the discipline in our CRO hypothesis framework. The feature is detailed in Optimizely’s Idea builder documentation.

Statsig opens the full feature-gate review lifecycle over its Console API

On 22 June Statsig exposed eight new Console API endpoints covering the entire gate review flow: create a review with a proposed change, then edit, approve, reject, cancel or list it without touching the console. A day later, on 23 June, it added follow and unfollow for Dynamic Configs so people get notified when a config value changes rather than finding out after something breaks. Taken together these are governance features: they let a team wire approvals into a CI pipeline and keep an audit trail on who signed off what. If your experimentation setup is drifting toward config sprawl and unreviewed changes, this is the kind of control that separates a mature platform from a light one, worth weighing when you read our best A/B testing tools guide. Both changes are logged in Statsig’s release notes.

VWO puts heatmaps and session replays inside feature experiments

VWO announced on 18 June that every feature experiment, rollout and personalisation campaign now carries integrated heatmaps and session recordings. The point is to shorten the loop when a test underperforms: instead of seeing only that a variant lost, you can watch where users clicked, hesitated or dropped off and tell whether the cause was the design, the copy, a bug or the placement. For a low-traffic programme that qualitative layer often finds the fix faster than waiting for a result to reach significance, and it is the habit we push in our CRO research methods guide. The update is on VWO’s product updates page.

PostHog lets you set a default minimum detectable effect

A recent PostHog change added a configurable organisation-level default minimum detectable effect, so every new experiment inherits your MDE instead of teams re-entering it or, more often, leaving it at a value nobody chose. It is a small setting with an outsized effect: the MDE is what decides how long a test must run and how much traffic it needs, so a sensible default stops people quietly starting underpowered tests. Set it to the smallest uplift that would actually change a decision, not the smallest number you can imagine, and sanity-check the runtime with our minimum detectable effect calculator. The change is in PostHog’s changelog.