How to choose what to improve first (ICE, PIE)

Hypothesis prioritization is a way to order ideas for improving a site (change a button, rewrite the delivery info, simplify a form) by expected benefit and difficulty, so you do first what gives the most for the least effort. The best-known simple models are ICE and PIE: ideas get scores on a few criteria, and the scores are combined.

In plain words

There are always more ideas than time. Without rules, the idea that was named loudest wins. Scores don't make a decision scientific, but they force you to explain why this idea ranks higher and reduce the influence of personal preferences.

ICE and PIE

The result is the average or the product of the scores (Mida, May 2026).

Example (illustrative numbers). Three ideas for a home appliances store:

We take delivery first: it is the highest. A confidence of 7 should rest on data, for example, on the fact that extra costs are named the most frequent reason for abandoning a cart (40% in Baymard's roundup, 2025; Baymard).

Weak spots

According to Mida's analysis (2026):

• all the criteria are subjective: "impact" without a link to conversion and traffic means nothing;
• the idea's authors overstate confidence;
• "ease" can be fudged by presenting a hard task as an easy one;
• in PIE, "potential" does not tell a weak page from a weak offer;
• the models undervalue bold ideas on new pages where there is no data.

There is also a stricter model, PXL: almost all criteria are answered yes/no (is the change prominent, noticeable within 5 seconds, on a high-traffic page, backed by research). It is more objective but slower: scoring one idea, according to Mida, takes 10–15 minutes versus a minute or two in ICE.

How to make scores more honest

1. Tie scores to data: funnels, session recordings, surveys.
2. Have the team set "confidence", not the idea's author.
3. Review the list once a month and recalculate.
4. Complement scores with common sense: bugs that break payment get fixed without waiting in line.
5. To check the effect, use an A/B test if you have enough traffic; with low traffic it is better to make the obvious fixes and compare periods carefully.

There are no fresh studies proving that one prioritization model gives better results than another: these are tools for disciplining the discussion, not a guarantee of results.

Common mistakes

• Giving every idea 8–9 points.
• Counting scores and then doing the "favorite" idea anyway.
• Not recording the result and not returning to the list.
• Starting with easy tasks and ignoring big ones with high impact.

FAQ

What is ICE? Impact, Confidence, Ease, each scored from 1 to 10.

What is PIE? Potential, Importance, Ease.

ICE or PIE: which is better? The difference is in the criteria: ICE takes confidence into account, PIE the value of the page. Choose the one your team will apply consistently.

How many ideas should I score at once? To start, the top 20 ideas; Mida advises reviewing and rescoring them monthly.

Related terms

• CRO (conversion rate optimization)
• A/B test
• Funnel analysis
• Customer surveys
• Conversion rate

Sources

• Mida, Test prioritization frameworks: ICE, PIE, PXL (18.05.2026) — mida.so/blog/test-prioritization-frameworks-ice-pie-pxl
• Baymard Institute, Cart Abandonment Rate Statistics (updated 22.09.2025) — baymard.com/lists/cart-abandonment-rate