Filter before you score
First remove ideas that cannot be measured, violate the current scope or depend on unavailable resources. A test of a different product promise is not ready if the product cannot deliver it. A conversion experiment is not interpretable if the completion event fires twice.
For the remaining ideas, write a short causal statement: “Changing this part of the journey should help this customer do this action because of this evidence.” If the explanation contains several unrelated mechanisms, split the idea into smaller tests.
Use a score as a conversation tool
| Criterion | Useful question |
|---|---|
| Relevance | Does this address the largest actionable constraint? |
| Evidence | What customer observation or data supports the hypothesis? |
| Learning | What decision becomes easier if the result is positive or negative? |
| Effort | What design, engineering, budget and coordination are required? |
| Risk | Could a local win damage trust, margin or retention? |
Avoid assigning precise revenue forecasts to weak ideas. Use ranges and note the uncertainty. Two reviewers giving different scores may reveal a disagreement about the customer problem rather than a spreadsheet mistake.
Choose the next experiment
- Define the primary outcome and one or two business guardrails.
- State the unit being compared: person, account, order or campaign.
- Check that outcomes can mature within a practical review window.
- Record which other changes must remain stable for interpretation.
- Agree what evidence would justify expansion, revision or stopping.
Low-traffic businesses can still learn from interviews, usability sessions and deliberately limited pilots. Those methods answer different questions from a randomized experiment; label the evidence accordingly.
Worked example: a useful negative result
Illustrative example: a team considers a new acquisition channel, a shorter lead form and a clearer pricing explanation. Sales notes show that suitable leads repeatedly misunderstand the offer. The pricing explanation is a strong early candidate because it tests a documented objection with modest implementation effort.
If lead volume falls but the share of qualified opportunities rises, the result is not automatically a failure. Evaluate the agreed business outcome and sales capacity. A useful test can identify which people the offer is meant to serve.
Avoid experiment theatre
An experiment count measures activity, not learning. Do not celebrate dozens of launches if the team cannot describe what changed in its understanding. Keep a decision log with the original hypothesis, result, limitations and next action.
Should you use ICE or RICE?
Use a simple framework the team can apply consistently. The important part is the quality of the assumptions and the discussion behind the score, not the acronym.
Should quick wins always come first?
No. Cheap tasks can consume the calendar without addressing the main obstacle. Include effort in prioritization, but keep business relevance central.
Put this into practice
Take five backlog ideas, write the evidence for each and rank them without pretending the estimates are facts. Select the one that can most clearly change your next business decision.
Related foundation: SaaS growth strategy: build acquisition around activation. How these guides are prepared.
Ayoub Mouhachtt
Growth & performance marketing. Explore the portfolio and working background.
Related portfolio work: eGrow. The worked examples in this guide are illustrative and are separate from the portfolio’s project evidence.