I just figured this out.

At some point you've got more experiment ideas than you have time or budget to run. So how do you decide what to run first?

We use a simplified version of the infamous ICE framework. Impact-Probability (IP).

Impact is the degree that change could shift your problem. Define it as what potential uplift that would have in your key metric next month.

Probability is what % likelihood you have that result will happen.

Expected value (EV) = Forecast uplift in key metric * Probability of event occurring.

Let's say your core metric is contribution margin.

EXP01: you could double your conversion rate meaning based on next month's forecast you increase contribution margin by £250k. But the probability of that happening sits at 20%. EV = £50k.

EXP02: can only generate £75k of CM gains but you've got 80% probability of success. EV = £60k.

Start with number 2.

You don't have time to analyse everything. Apply pressure and fast. EV helps you do that without guessing.

Do you have a system for prioritising experiments?