An ICE prioritization template is a spreadsheet with one row per test idea and three 1-10 scores: Impact, Confidence and Ease.
Multiply the three to get an ICE score from 1 to 1,000, sort the sheet from highest to lowest, and test from the top down.
The template below adds the columns that keep those scores honest: the evidence behind each idea and similar tests that have already run.
Copy the table, use the scoring rubric to fill it in, and check the filled-in example, where every row is an idea a real team tested and logged in the LiftArchive test database.
For the theory behind each framework, see our guides to ICE, PIE and PXL.
The prioritization spreadsheet template
Make a new sheet in Google Sheets or Excel and use these columns, left to right.
Row 1 holds the headers; each idea gets its own row from row 2 down.
| Column | Header | What goes in it |
|---|---|---|
| A | ID | A short number so you can refer to the idea in meetings and reports. |
| B | Idea | One sentence describing the change. Link to the full write-up from your hypothesis template. |
| C | Page or step | Where it runs: homepage, product page, cart, checkout, pricing page, onboarding. |
| D | Evidence | What started the idea: analytics, heatmaps, session recordings, surveys, user tests, or "opinion". |
| E | Similar tests | Links to comparable tests from your own history or our database, with their result. |
| F | Impact (1-10) | How much the main metric moves if it wins, given traffic and closeness to purchase. |
| G | Confidence (1-10) | How strong the evidence in columns D and E is. |
| H | Ease (1-10) | How little design, development, QA and sign-off it needs. |
| I | ICE score | Formula: =F2*G2*H2 |
| J | Status | Backlog, scoped, running, done, or parked. |
| K | Result | After the test: the lift, or "flat" / "lost", with a link to the test report. |
To keep a ranked view on a second tab, use =SORT(Backlog!A2:K, 9, FALSE) in Google Sheets.
That sorts by column 9 (the ICE score) from highest to lowest and updates as scores change.
The Impact × Confidence × Ease formula comes from Itamar Gilad's write-up of ICE, which credits Sean Ellis with inventing the method for growth experiments.
Some teams average the three scores instead. Multiplying is stricter: a 1 on any factor sinks the idea, which is what you want for a two-month rebuild nobody has evidence for.
How to fill it in
The weak point of ICE is that each score can be a guess, and the most senior person's guess tends to win.
A written rubric, agreed before scoring, fixes most of that.
Use this one as a starting point and adjust the brackets to your traffic and team.
| Score | Impact | Confidence | Ease |
|---|---|---|---|
| 9-10 | Seen by most buyers, on the cart, checkout or a top-traffic page | Two or more evidence sources plus similar tests that won | Copy or settings change, live in under a day |
| 6-8 | Product pages or a major landing page | One solid source (heatmaps, survey, analytics) plus a similar win | A few days of front-end work, one approver |
| 3-5 | A secondary page or a single segment | One source, or similar tests with mixed results | A week or more, or needs back-end or app changes |
| 1-2 | Low-traffic page, far from purchase | Opinion only, or similar tests mostly lost | A redesign, a new feature, or legal and brand review |
A few rules that help:
- Fill in Evidence and Similar tests before any score. Confidence should be read off those two columns, not felt. Gilad's Confidence Meter takes the same line: your own conviction and colleagues' opinions count for almost nothing.
- Search the database for the Similar tests column. If ten stores tested an idea and most lost, that belongs in the row.
- Score Impact against traffic. A big change on a page with little traffic may never reach significance. Read our low-traffic testing guide before giving a high score.
- Score as a group, then compare. Have two or three people score on their own, and discuss any factor where they differ by more than 3 points.
- Rescore monthly. Every finished test changes the Confidence of related ideas. A loss on the cart should lower similar cart ideas.
Filled-in example
Here is a five-row backlog for an ecommerce store.
Each idea is one a practitioner in our database actually tested, and the scores use the rubric above based on the evidence each team said it had before the test.
The Result column shows what happened. These ran on different stores, so treat this as an illustration of the method, not a controlled comparison.
| ID | Idea | Evidence | I | C | E | ICE | Result |
|---|---|---|---|---|---|---|---|
| 1 | Move the PDP's above-the-fold USPs to the points buyers engage with | Heatmaps and scroll data showed buyers scrolling down to specific messaging, then back up to buy | 7 | 7 | 8 | 392 | +7.28% revenue per visitor (Dave Diederen) |
| 2 | Make the announcement bar clickable, linking to high-ticket bundles | Heatmaps showed visitors trying to click the bar; surveys showed what mattered | 5 | 7 | 9 | 315 | +$74,000/month (Oliver Kenyon) |
| 3 | Add urgency messaging to checkout | Similar urgency tests in the database; small, isolated change | 6 | 6 | 8 | 288 | +3.37% conversion rate, +6.59% revenue per visitor (Dave Diederen) |
| 4 | Remove the email popup | Opinion: "popups annoy people" | 6 | 3 | 10 | 180 | -20.51% revenue per visitor (Dave Diederen) |
| 5 | Full checkout redesign | Opinion; many changes bundled together | 9 | 3 | 2 | 54 | Lost: -1.18% conversion rate, -1.84% revenue per visitor (same Dave Diederen test, row 3) |
Two things stand out. First, the ideas backed by behavioral evidence (rows 1 and 2) scored highest and both won.
Second, the two ideas based on opinion scored lowest and both lost, including the popup removal, which looked like a free win because it scored 10 on Ease.
Rows 3 and 5 come from one store.
Dave Diederen tested a full checkout update first, and it dropped conversion rate by 1.18%, revenue per visitor by 1.84% and profit per visitor by 1.96%.
Testing the urgency messaging on its own lifted conversion rate by 3.37%, revenue per visitor by 6.59% and profit per visitor by 7.13%.
The redesign would have topped a sheet ranked on Impact alone. Multiplying pushed it to the bottom.
@oliverkenyon@oliverkenyon@oliverkenyon
@DaveDiederen@DaveDiederen@DaveDiederenSmall, evidence-backed ideas often deliver large results. Oliver Kenyon removed a single line of copy from a cart page for +$40,413 per month, and three survey-based reassurance lines above a collection grid added $44,890/month.
Both would score high on Ease, and the survey-based one high on Confidence too.
PIE and PXL versions of the sheet
You can swap the scoring columns (F to I) for another framework and keep everything else.
For PXL, insert columns so the ten criteria fill F to O.
| Framework | Scoring columns | Score formula | Best for |
|---|---|---|---|
| ICE | Impact, Confidence, Ease (1-10) | =F2*G2*H2 (1-1,000) | Ranking individual ideas quickly |
| PIE | Potential, Importance, Ease (1-10) | =AVERAGE(F2:H2) (1-10) | Ranking pages or templates to test first |
| PXL | Ten mostly yes/no criteria, weighted | =SUM(F2:O2) (0-14) | Teams that want fewer subjective scores |
PIE was built by Chris Goward at WiderFunnel to rank pages, as Goward's Practical Ecommerce article shows.
Run it on a separate tab with one row per page or template, then use ICE on the ideas for the top-ranked pages.
PXL replaces the 1-10 guesses with questions such as "Is the change above the fold?" and "Does it address an issue found in user testing?"
CXL publishes the criteria in its 2016 PXL post along with a free Google Sheets template you can copy.
Our PXL guide lists all ten criteria and their points.
Whichever you use, the Evidence, Similar tests and Result columns matter more than the formula.
Over time they turn the sheet into a record of what works for your customers, which feeds your testing roadmap.
For a side-by-side of the three methods, see how to prioritize A/B tests.