Reach

When a feature argument stalled, the line that usually decided it was "a lot of people use this." There is no scale in "a lot," so the loudest voice won.

Pagedeck had exactly that argument: fixing search result ordering versus fixing the document permissions screen, with both sides claiming more users. The logs said 2,000 accounts had used search in the previous quarter and 180 had opened the permissions screen. The argument ended five minutes after the numbers appeared. Reach exists to produce those five minutes by fixing a user count into the first slot of the calculation.

Count the users a change touches inside a fixed window and use that count as-is. Where records exist, count rather than estimate.

1. Fix the window first, usually a quarter, and apply the same window to every candidate. Score one item quarterly and another annually and the second one is four times larger for no reason.
2. Fix the unit too. Decide between people and actions and don't mix them. If the unit is people, someone who searches 60 times still counts as one.
3. Pull the value from the records. If 2,000 accounts opened the search screen last quarter, reach is 2,000. For something that doesn't exist yet, the number of people who passed through the screen it would live on is the ceiling.
4. Don't use total user count. With 30,000 accounts and 180 people opening the permissions screen, work on that screen has a reach of 180. Put overall scale in this slot and every candidate scores the same, which empties the calculation.
5. Multiply the result by impact and confidence, then divide by effort. A reach of 2,000 times an impact of 2 and a confidence of 0.8, divided by three person-months, gives 1,067 — a figure that means something only against the 240 the next candidate scored the same way.

"What's the reach on this? Not total accounts, accounts that open the screen."

Asking for the count of people the change touches instead of overall size.

"Reach is high but impact is a 1, so the score doesn't clear the bar."

Plenty of people are affected and the per-person effect is too small to move it up the list.

"The feature doesn't exist yet, so I used screen visitors as the reach."

Where the value can't be counted, the ceiling stands in for it.

Mia filled in reach for eighteen candidates in one pass using a single table built on one condition. Each candidate got the screen it touches, plus the count of accounts that opened that screen last quarter. Five of the eighteen came back with two-digit reach, and those five were moved to the next quarter's list without scoring the remaining terms.