This calculator scores a list of features against a list of criteria you weight yourself, then ranks the features live as the numbers change. Score each feature 1 to 5 on every criterion, 5 is best. For a cost criterion such as effort, score it the same direction as the others: 5 for the lowest effort, 1 for the highest, so a cheap win does not get penalized for being cheap.
| Criterion | Weight | Normalized | Remove |
|---|
| Feature | Weighted score | Remove |
|---|
The rows above are an example. Clear them and score your own list.
How do I pick criteria and weights?
Four or five criteria, not ten. Customer demand, revenue impact, strategic fit and effort cover most product decisions on their own, and every criterion past that point mostly restates one already on the list in different words, which quietly double counts it.
Weights are a policy decision, not a measurement. Nobody discovered that revenue impact is worth thirty and strategic fit twenty. Somebody who owns the roadmap decided it, wrote it down, and now applies it the same way to every row. Write the weights down before scoring the first feature, because a weight chosen after you have already seen the scores tends to land wherever it needs to for the feature you already wanted to win.
Weights do not have to sum to 100. The calculator above normalizes whatever you enter, and shows the note when they do not, so entering 35, 30, 20 and 15, or 7, 6, 4 and 3, scores the same way either time.
The worked example, explained
The five rows above score five feature requests against four criteria: customer demand, revenue impact, strategic fit, and effort, scored so the smallest effort gets the highest number.
Saved filters wins because it is cheap and reasonably wanted across every column, not because it is the biggest win on any single one. API webhooks comes second on strategic fit and revenue impact alone, even carrying a high effort cost. Dark mode tops customer demand with a 5 and still lands third, because it scores 2 on revenue impact and on strategic fit. Its low effort is what keeps it a whisker above SSO login, 3.35 to 3.30, and that is the kind of close call the weights exist to make visible: raise revenue impact's weight a few points and the two swap.
Building the same model in Excel
A spreadsheet needs two tables and one formula: a short list of criteria with their weights in column B, and a feature table where each row's weighted score reads SUMPRODUCT of its scores against those weights, divided by the weight total.
B6 =SUM(B2:B5)
F9 (row 9) =SUMPRODUCT(B9:E9,$B$2:$B$5)/$B$6
SUMPRODUCT multiplies each feature's score by the matching weight and adds the results in one step, so the formula does not need four separate multiplications added together. Dividing by the weight total normalizes the result, which is why it still works when the weights do not sum to 100.
Take the xlsx, the same five rows and four criteria above with that formula filled down, the header frozen and a filter on every column. It opens in Excel, in Numbers, in LibreOffice, and Google Sheets imports it with the formula intact. It may be copied, changed and used commercially with no attribution and no link back.
Two mistakes that wreck a weighted score
Too many criteria. Past five or six, most of the new ones restate an existing one in different words, which gives that underlying idea two votes instead of one without anybody deciding that it should have two. A long criteria list also makes every feature look similar once the scores average out, because more terms pull every total toward the middle.
Scoring effort the same way you score value. The model adds its columns, so if 5 means the most effort, the most expensive feature collects the most points for being expensive. Flip the scale for effort, 5 for the least work and 1 for the most, and the weighting does the rest.
How this differs from RICE
RICE multiplies reach, impact and confidence, then divides by effort. A weighted scoring model adds, not multiplies, and the criteria are whatever you decide matters, which is why it can hold strategic fit or a sales deal size that RICE has no column for.
The two also fail differently. RICE's reach is often an estimate that looks like a count. A weighted scoring model's weights are an openly stated policy, so when two people disagree about a ranking here, the argument is about the weight, not about whose guess was better. The RICE score calculator is the other version of this exercise, for a backlog where reach can be counted.
Where a request board helps
Customer demand is a criterion on the list above, and a criterion needs a real number behind it rather than a guess about who is asking.
VoteFirst is a feature request board where visitors vote with no account. Each vote is a count against a specific request, so the demand column in a weighted scoring model can read straight off the board instead of being estimated. On the Lite plan, from $9 a month, a vote can also carry the weight of what that customer pays, imported from Stripe or Paddle, so a request asked for by ten paying customers scores differently from one asked for by ten accounts that never paid. How that weighting actually works is on its own page. A calculator for just that part runs the vote weighting alone, separate from the four criterion model here.
14 days, a card at signup, then $9 or $39 a month.