Method · How Adrata measures
How we know a play worked.
Adrata compares reps who have a play with reps who don’t have it yet, in the same weeks and the same kind of deals. Every number comes with a likely range, and every number opens the calls behind it.
01
Compare with reps who don’t have it yet
A play reaches reps in waves, in an order picked at random. Each wave is compared with reps who’ll get it later, in the same weeks and segment. Comparing reps who picked up a move with reps who didn’t mostly measures who’s keen.
02
Show a likely range, not one number
Seven wins from ten decided deals could mean a true win rate anywhere from about 40% to 89%. So every rate is adjusted for how few deals it rests on, and every gain carries its likely range, such as “likely +2 to +14 points”.
03
Say “too early to tell” when it is
While the range still includes zero, a play shows as an early read, with how many more deals it needs. A play that didn’t hold up stays in the track record, with its date and the deals behind it.
04
Name a best rep at about 20 decided deals
A “best at this” badge needs at least an 8 in 10 chance that the rep is truly in the team’s top quarter for that skill. At typical team spreads that takes about 20 decided deals in the skill, not 10.
05
Every number opens the calls behind it
Click any rate, gain or badge to see the deals and call moments it was counted from. We only count what happened before a deal was decided. If you can’t check a number, we don’t show it.
Sources
Where the method comes from
Study · Hemming, Haines, Chilton, Girling & Lilford, BMJ, 2015 Study · Callaway & Sant’Anna, Journal of Econometrics, 2021 Study · Wilson, Journal of the American Statistical Association, 1927 Study · Gelman, Hill & Yajima, Journal of Research on Educational Effectiveness, 2012
