PIE and ICE Score Calculator for Experiments

The bottleneck of a testing program is almost never a shortage of ideas, it is the order you run them in. With limited traffic, every week spent on a weak experiment is a week the strong one did not run. Prioritizing by opinion turns the roadmap into whoever argues loudest in the meeting. PIE and ICE swap the argument for a number: you rate each idea on three dimensions and let the average order the queue. This tool does the math live and hands back the ranking in one click.

How to use it

1. Pick the framework: PIE when you prioritize pages and areas by potential return, ICE when you prioritize loose ideas and want to reward stronger evidence.
2. List the experiments, one per row. Describe the concrete change, not the goal: "social proof above the fold", not "increase trust".
3. Rate 1 to 10 on each of the three dimensions. Be honest about ease: a test that needs three dev sprints is not easy.
4. Read the score and the rank, which update on every rating. Use "Sort by score" to see the queue in ranked order.
5. Click Download CSV to take the prioritized ranking into your backlog in Sheets, Notion or Trello.

How it works: the PIE and ICE formula

In both frameworks, each dimension gets a rating from 1 to 10 and the score is the simple average of the three:

All three dimensions are always "higher is better", including ease: easier to build earns a higher rating, because it frees traffic sooner for the next test. That is why a plain average already ranks the list, with no axis to invert. Potential (or Impact) is how much the page can improve; the Importance in PIE is the value of the traffic passing through; the Confidence in ICE is the strength of the evidence behind the bet.

Worked example (reproduces the default result)

With the three experiments already filled in, on the PIE framework: "social proof above the fold on the homepage" gets Potential 8, Importance 9 and Ease 6. The score is (8 + 9 + 6) / 3 = 23 / 3 = 7.67. "Simplify the checkout form" gets 9, 8 and 4, so (9 + 8 + 4) / 3 = 21 / 3 = 7.00. "Rewrite the main CTA button label" gets 5, 6 and 9, so (5 + 6 + 9) / 3 = 20 / 3 = 6.67.

The ranking comes out as 1st the social proof (7.67), 2nd the checkout (7.00) and 3rd the button (6.67), exactly what the tool shows above when you open the page. Notice what the score captures: the button is the easiest of all (9 on ease), yet it still lands last, because its potential and importance are lower. That is the central lesson of prioritization, start with what moves the needle, not with what is comfortable to build.

How to read it and where it misleads

The score is a relative comparison ruler, not an absolute grade. A 7.67 only means "better than the 7.00 in the same list", not "will return 7.67". It exists to order the queue, and that is what solves the real problem: to stop running the easy test ahead of the important one. Always compare within the same round and the same framework, because mixing PIE and ICE ratings in one list makes no sense, the dimensions measure different things.

The honest limit is the subjectivity of the ratings. Two analysts give different ease scores to the same test, and the confidence rating in ICE is where bias hides most: it is easy to hand a 9 to the idea you already wanted to run. Two simple defenses: rate as a group, not alone, and anchor confidence in real evidence, not enthusiasm. If your confidence comes from quantitative data, it deserves a high rating; if it comes from a hunch, be strict. The hypothesis generator helps turn the prioritized idea into a falsifiable bet before you switch the test on.

Best practices when prioritizing

• Fix a shared scale before rating. Agree on what "ease 3" and "ease 9" mean so the rating carries the same weight across the list.
• Rate in pairs or trios. The average of two readings cuts individual bias better than any rule.
• Re-score the queue after every finished test. A surprise result changes the confidence of neighboring ideas and reorders the backlog.
• Do not chase the third decimal. A 0.1 gap between two ideas is noise; run either one first and move on.

Frequently asked questions

What are PIE and ICE prioritization?

They are two frameworks for deciding which experiment to run first. PIE (Potential, Importance, Ease) came from WiderFunnel and scores the improvement potential of a page, the importance of the traffic it receives and how easy it is to build. ICE (Impact, Confidence, Ease) came from Sean Ellis growth work and scores the expected impact, your confidence that it will work and the ease. In both you rate each dimension from 1 to 10 and the score is the average.

What is the difference between PIE and ICE?

The difference is in the middle letter. PIE asks "how important is this page" (the volume and value of the traffic passing through it), while ICE asks "how confident are you" (the strength of the evidence behind the bet). Use PIE when you prioritize pages or areas by potential return; use ICE when you prioritize loose ideas and want to reward the ones with stronger evidence. Potential/Impact and Ease exist in both.

How is the score calculated?

It is the simple average of the three ratings: score = (rating1 + rating2 + rating3) / 3. All three dimensions are "higher is better", including ease (easier is better), so a plain average already ranks them. An experiment with 8, 9 and 6 gives (8 + 9 + 6) / 3 = 7.67. The result stays on the same 1 to 10 scale, which makes it easy to compare the whole list at a glance.

Should I multiply instead of averaging in ICE?

Some teams multiply the three ICE ratings instead of averaging, which spreads the scores wider (from 1 to 1000) and punishes a weak dimension harder. In practice both orderings come out similar. This tool averages in both frameworks because it is the original documented form of PIE and the most readable one: the score lands on the same scale as the ratings, so a 7 means the same thing everywhere.

Does the highest score guarantee the test will win?

No. The score measures priority, not outcome. It tells you where the next cycle of traffic is best spent, based on what you know today. Plenty of high-scoring ideas lose, and that is normal: prioritization is about where to bet, not a guarantee of return. The confidence dimension in ICE helps calibrate that, but no framework replaces running the test and reading significance at the end.

How do I export the ranking?

Click Download CSV: the tool builds the file already sorted by score, with rank, name, the three ratings and the final score, and the download happens in your browser without sending anything to any server. The file opens straight in Excel, Google Sheets or Notion, ready to become the prioritized backlog of your testing program.

Keep going

Prioritized the queue? Take the top idea and write the bet in the A/B test hypothesis generator, size the experiment in the sample size calculator and, when the test ends, read the verdict in the significance calculator.