Live demo

The best rep on this leaderboard has closed three deals.

Forty reps, one year of qualified opportunities, and a true close rate injected into each of them that no ranking ever gets to see. Everything below is computed in your browser from the outcomes. At the end we check both leaderboards against the truth, which is the one thing you can never do with real data.

Close rate, trailing four quarters40 reps5 major rank changesteam rate 30.0% · prior worth 39 opportunities
What the dashboard shows
ranked by raw close rate
#RepOppsRateBecomes
1Priya N.joined Jul 2026560.0%#11
2Kwame A.joined May 20261656.3%#4
3Hannah B.64046.7%#1
4Tomas V.joined Jun 20261145.5%#12
5Paulo R.21041.0%#3
6Rafael C.52040.4%#2
7Ruth C.10540.0%#5
8Ines B.joined May 20261338.5%#14
9Ivy M.47037.0%#6
10Greta L.38036.3%#7
Ranked by shrunk estimate
each rate pulled toward the team mean in proportion to its noise
#RepOppsRateWas
1Hannah B.64045.8%#3
2Rafael C.52039.7%#6
3Paulo R.21039.2%#5
4Kwame A.joined May 20261637.6%#2
5Ruth C.10537.3%#7
6Ivy M.47036.5%#9
7Greta L.38035.7%#10
8Samir P.61034.5%#12
9Yuki S.19534.1%#11
10Claire V.41033.8%#13
Priya N. closed 3 of 5 opportunities and tops the raw board at 60.0%. Her injected true rate is 26.0%, below the team’s 30.0%. She is not the best rep. She is the rep with the smallest denominator, and on a raw leaderboard those are the same thing.

The free lunch

Scored against the truth, the corrected board wins by 7.6×

Every rep in this data has an injected true close rate. Sum the squared error of each leaderboard against it. This is the comparison you cannot run at work, and it is the reason to trust the method when you get there.

Raw squared error
0.311
sum over 40 reps
Shrunk squared error
0.041
same reps, same data
Error removed
86.8%
not on average, on this draw
Prior strength
39 opps
fitted from the team, not chosen
Total squared error against the injected truth
Raw average0.311Shrunk0.041

James-Stein says this is not luck. For three or more groups, shrinking toward the population mean beats the raw estimates in total squared error. Always, not usually. The result is from 1961 and it is still almost entirely absent from business tooling.

The bottom of the board is the same illusion upside down

The four worst reps by raw rate are Sofia L. (1/6), Marcus O. (1/7), Ben H. (1/8), Dani R. (1/9). All four joined this year. After correction they sit at #23, #26, #29, #30, and one of them has a true rate sitting on the team average. A performance plan built on the raw bottom is a performance plan built on four coin flips.

How far each rep moves

Three opportunities is worth almost nothing, and the math says so exactly

The pull toward the team mean is the fitted prior strength divided by that plus the rep’s opportunity count. It is not a rule of thumb or a minimum-sample cutoff. A rep with 5 opportunities moves 88.7% of the way; a rep with 640 moves 5.8%.

0%25%50%75%100%10 opps50 opps200 opps1000 oppsPriya N. · 5Hannah B. · 640

The eight red dots are everyone who started carrying a quota this year. They are the only reps whose numbers move materially, and on the raw board they occupy both ends of the ranking.

Who is actually different

Ranking forty reps is forty simultaneous tests

Screening a whole leaderboard for outliers means testing every rep against the team rate at once, and at forty tests you expect two false flags at the 5% line before anyone has done anything unusual. Benjamini-Hochberg controls that. It costs you findings: 16 reps clear an uncorrected p < 0.05, 12 survive FDR control.

Reps screened
40
one test each
Uncorrected hits
16
p < 0.05, no correction
Survive FDR
12
Benjamini-Hochberg at 5%
Genuinely different from the team rate of 30.0%
RepOppsRawShrunk95% intervalqDirection
Hannah B.64046.7%45.8%42.0%49.5%<0.0001Above team
Rafael C.52040.4%39.7%35.6%43.7%<0.0001Above team
Paulo R.21041.0%39.2%33.2%45.3%0.0036Above team
Ivy M.47037.0%36.5%32.3%40.7%0.0047Above team
Greta L.38036.3%35.7%31.1%40.3%0.0268Above team
Samir P.61034.8%34.5%30.8%38.1%0.0356Above team
Kai N.90025.9%26.1%23.3%28.9%0.0268Below team
Mira Z.54524.4%24.8%21.3%28.3%0.0189Below team
Iris F.35522.0%22.8%18.6%26.9%0.0047Below team
Otto W.42521.9%22.6%18.8%26.4%0.0020Below team
Lior T.30520.3%21.4%17.1%25.8%0.0020Below team
Vera H.61018.0%18.8%15.8%21.8%<0.0001Below team

Not one rep who joined this year appears on this list, in either direction. That is the correct answer: with fewer than twenty opportunities there is no evidence about anyone yet, however extreme their number looks. 6 reps are genuinely above the team rate and worth studying; the rest of the top of the raw board was noise.

The calculator

How much of that number should you believe?

Someone shows you a close rate. Before you promote, investigate, or copy whatever they are doing, put the denominator in. The prior here is the one fitted from this team, worth 39 opportunities.

What they report
60.0%
5 opportunities
Pulled toward the team
88.7%
team rate is 30.0%
What to believe
33.4%
the prior carries most of it
At 5 opportunities, more than half of your estimate is coming from the team average rather than from this person. 60.0% is a headline, not a measurement. A promotion, a PIP, or a "go learn what they do" decision made on it is a decision made on 5 coin flips.