What each sample size buys you
Assuming a true rate near 30% and a 95% interval, the range around the measured number narrows like this:
| Resolved deals | Approximate margin | One deal moves the rate by |
|---|---|---|
| 30 | 16 points | 3.3 points |
| 50 | 13 points | 2.0 points |
| 100 | 9 points | 1.0 point |
| 200 | 6 points | 0.5 points |
| 400 | 4.5 points | 0.25 points |
Small samples break the cuts people care about most
The cuts with the highest management value are the ones with the fewest deals. Win rate by rep and win rate by named competitor both slice the same denominator into pieces too small to read. A single quarter of enterprise deals for one rep is a sample of maybe eight.
The fix is to change the shape of the question. Ask whether a rep is above or below the team distribution over a year rather than what their rate is this quarter, and put the resolved deal count next to every rate you publish so a reader can discount it themselves.
When the data is thin rather than dirty
Sample size and data quality are separate problems, and teams frequently blame the second for the first. ORM's position on data quality is that consistency matters more than cleanliness, since predictions hold up as long as the same errors show up the same way over time.
That does not rescue a thin sample. Inconsistent records can be modeled around. Twelve deals cannot, because there is nothing there to model. Widen the window, roll the cut up a level, and keep the granular version as a hypothesis rather than a finding. See win rate for the base metric and forecast accuracy for what happens when a thin rate is used as a forecast input.
Frequently Asked Questions
What is the minimum number of deals for a usable win rate?
About 30 resolved opportunities for a directional read. At that size a single deal moves the rate by more than three points and the 95% interval around a 30% rate spans roughly 16 points in each direction, so the number is a range rather than a value. It is enough to say a segment is weak and not enough to say it is 6 points weaker than another segment.
Can you rank reps by win rate?
Not on a single quarter. A rep who resolves 12 deals a quarter has an interval wide enough to cover almost any ranking you could construct, which means the leaderboard is mostly reordering noise. Pool four quarters per rep, or compare reps at the team and segment level where the counts are large enough to carry a signal.
What do you do when a segment has too few deals?
Widen the window before you narrow the cut. A rolling four quarter view on the same segment usually produces a workable count, and it also smooths the seasonality that makes single quarters look volatile. If the segment still lands under 30 after that, report the raw counts instead of a percentage.
Does dollar weighting help with a small sample?
It makes the problem worse. Value weighted rates are dominated by the largest deals, so in a small group a single outsized win or loss can move the number by more than the difference you were trying to measure. In small samples the count based rate is the more stable of the two, and both should be read as ranges.
Put these metrics to work
ORM builds custom revenue forecast models that turn concepts like how many deals do you need to calculate win rate? into prescriptive action for your team.
Schedule a Demo