Prioritization
Estimating reach and impact when your data is thin
The two RICE factors people fudge most are reach and impact. Here's how to estimate them defensibly without pretending you have data you don't.
Effort you can ask engineering about. Confidence is a judgment call you can be honest about. The two RICE factors that invite fantasy are Reach and Impact — so here's how to estimate them without inventing numbers.
Reach: find a countable proxy
Reach is supposed to be a real count over a time period, so the job is to find something you can actually count that stands in for "people affected." Candidates:
- Funnel data: a checkout improvement reaches everyone who hits checkout — that number is in your analytics.
- Feature usage: an improvement to feature X reaches feature X's monthly actives.
- Ticket/support volume: a fix reaches roughly the people hitting the problem, and ticket counts approximate that.
- Segment size: a feature for enterprise admins reaches the count of enterprise admins.
Pick the proxy, write it down, and you've turned a guess into an estimate someone can challenge — which is the whole point.
Impact: anchor to measured reality
Impact is the easiest factor to inflate because it runs on enthusiasm. Anchor it instead: think of two or three past features where you actually measured the result, and rate the new item relative to them. "This is probably like the onboarding change that lifted activation a few points — call it Impact 2" is defensible. "I think this is huge — Impact 3" is not. The fixed scale (3/2/1/0.5/0.25) only protects you if you tie it to things you've observed.
Use ranges, then be conservative
When you're unsure, estimate a range and take the cautious end. If reach is "somewhere between 2,000 and 5,000," use 2,000 and let Confidence carry the uncertainty. Optimistic point-estimates compound across a backlog and systematically over-rank your favorites.
Keep a record
The single best thing you can do for future estimates is write down what you shipped and what it moved. A short log of "feature → predicted impact → actual impact" turns next quarter's guesses into calibrated estimates. Most teams never do this, which is why their RICE scores never get more accurate.
- Reach: derive from a proxy you can count — active users, ticket volume, funnel steps.
- Impact: anchor the scale to past launches you measured, not to enthusiasm.
- Use ranges and pick the conservative end; lower Confidence to reflect the uncertainty.
- Comparable past features are your best estimator — keep a record of what shipped and what moved.
Try the Prioritization workflow in Cadenly
Cadenly can suggest RICE factors with a rationale you can adjust — a defensible starting point when your data is thin.
Start free →