Tracking outcomes

An approved proposal becomes a test, and the test produces a verdict. That verdict is recorded against the original proposal, so over time you accumulate a record of which kinds of change actually move your numbers.

Reading an outcome

A completed proposal shows what was predicted, what happened, and the size of the difference. Three cases:

• Won — the variation beat control. Implement it permanently; a test left running forever is a maintenance cost.
• Lost — control held. Useful. Record why you thought it would win.
• Inconclusive — not enough difference to call, or not enough traffic. Either run longer or accept the change does not matter enough to pursue.

Shipping a winner

Spectry applies a winning variation through the SDK, which is fine indefinitely but adds a small render cost to every page load. For a change you have decided to keep, build it into the page properly and stop the test.

The compounding part

Individual wins are usually small. The value is in the loop: propose, test, record, repeat. After a dozen cycles you know things about your visitors that no general best-practice article could have told you — and the proposals you get become more specific to your site.

Pair this with your CRO profile, which tells Spectry what your site is for and which metric you actually care about. A proposal aimed at the wrong goal wastes a test cycle.