The outcome
Four things the agency could not previously have known.
Each had been silently costing money for as long as the function existed.
Finding 01
Every speed estimate was wrong, in both directions
Work was assigned against speeds the team believed to be true. Not one was right, and critically they were not uniformly optimistic or uniformly pessimistic. They were simply unrelated to reality.
Assumed pace against measured pace, per available working hour. Measured across 31,294 finds and 65,506 vetting decisions.
| Team member |
Discovery |
Error |
Vetting |
Error |
| A | 56 → 134 | 2.4x under | 100 → 182 | 1.8x under |
| B | 50 → 90 | 1.8x under | 150 → 170 | 1.1x under |
| C | 50 → 70 | 1.4x under | 150 → 127 | 0.85x over |
| D | 60 → 68 | 1.1x under | 100 → 209 | 2.1x under |
| E | 40 → 44 | 1.1x under | 120 → 94 | 0.79x over |
| F | 60 → 41 | 0.68x over | 100 → 98 | roughly right |
| G | vetting only | — | 150 → 96 | 0.64x over |
Pod-wide, discovery capacity was under-estimated by about 40% while vetting capacity was over-estimated by about 18%.
When estimates are wrong in both directions, work is not merely mis-sized. It goes to the wrong people.
Someone believed to be a fast vetter and loaded accordingly was in fact one of the slower ones, while a genuinely fast discoverer sat under-used. Every capacity decision made before this point, whether a campaign could be accepted, whether a deadline was reachable, whether another hire was needed, rested on it.
Finding 02
The survival assumption was under-provisioning every campaign
Planning assumed 75% of discovered creators would survive quality vetting. The measured rate across 68,942 real verdicts is 64.8%. That gap is not academic. It has an exact operational cost.
To deliver 1,000 surviving creators, the plan sizes discovery at 1,333 finds. At the real survival rate those yield 864. Hitting 1,000 actually requires 1,543 finds, 210 more than the plan provisions, every time.
Across the six-week window the wrong assumption accounts for roughly 9,400 vetting decisions the plan never budgeted for. That shortfall was previously absorbed as unplanned top-up work and missed targets, which read as effort problems rather than arithmetic problems.
Finding 03
A 28-point quality spread between researchers, previously invisible
Two people producing at similar volume can differ by 28 points on whether that volume is worth anything. The difference lands entirely on the vetters downstream, as hours spent rejecting work that should never have been submitted.
Survival rate by researcher, across 29,167 peer-vetted finds.
| Researcher | Finds vetted | Survival rate |
|---|
| A | 652 | 91.6% |
| B | 3,077 | 86.7% |
| C | 2,765 | 78.5% |
| D | 10,964 | 76.3% |
| E | 8,035 | 63.8% |
| F | 3,674 | 63.3% |
Before the platform, every one of them looked identical on a tracker. The same measurement runs per campaign type, where survival ranges from 33.9% to 100% across 28 categories, which is the direct evidence that a single flat assumption cannot be right for every brief.
Finding 04
About a third of all rejections were objectively preventable
Rejections are captured as structured reasons rather than free text, which makes them countable. Of 23,777 tagged rejections, 8,380, roughly 35%, failed on objective criteria: follower count, region, age range, language, inactive accounts, business pages.
All of those are knowable before anyone opens the profile. They represent work that was sourced, submitted, reviewed and thrown away.
8,380
Objectively preventable rejections in six weeks
310
Paid hours consumed sourcing and rejecting them
~20%
Of total pod capacity, spent on avoidable rework
28p
Cost to deliver one vetted creator, now calculable