I.
“Fill rate is the only marketplace metric that matters. Everything else is lagging or gameable.”
It decomposes into three drivers — match, accept, complete — each with its own owner and lever. I would refuse to goal on GMV alone.
The numbers are illustrative.
The thinking is what matters.
The only metric that tells you if the marketplace works. When fill rate breaks, client NPS breaks. Renewals slip. Everything downstream is a second derivative of this one number.
The marketplace ran at 84% for years.
Orders placed, creators matched, shoots delivered, inside SLA. The kind of number you build enterprise contracts on.
Then late 2025 happened.
The rate cut. The v3 platform launch. Payment slowdowns. Three shocks compounding into one operational collapse.
Today it sits at 72%.
A twelve-point drop in six months. The single biggest operational problem on the board.
Fill rate owns everything upstream. The first ninety days is about making this number move. Every other decision refers back to it.
Shoots filled within SLA, divided by shoot orders placed. Did clients get what they ordered? Seventy-two percent says mostly yes.
But rush orders — the segment that defines enterprise NPS — fill at fifty-four percent. That’s where contracts get renegotiated.
Shoots filled within SLA, divided by creator-hours offered. Did creators get the work they expected? Forty-one percent says no.
Most marketplaces report one fill rate. Ocus has two. The gap between them is diagnostic — and the gap here is thirty-one points.
Not about supply. Not about demand. About the work of finding the right one for the right one, fast enough that both sides stay.
Goal the matching team on share of orders matched within six hours, not on average time. Averages hide the broken long tail — which is where the business actually lives or dies.
Ocus isn’t one marketplace. It’s seventy-eight. Each metro × shoot type × turnaround is a separate matching problem. Blending them hides where the break actually is.
| Rest. Menu |
Hotel Room |
Delivery Listing |
E-comm Product |
Real Estate |
|---|
Seventy-eight separate matching problems.
Each combo has its own supply, its own demand, its own break point. You cannot fix them with one lever.
Thirty-four are structurally broken.
Clients ordering faster than creators are offering hours. Tier 2 metros on rush turnaround are the hot zone.
The biggest gap: Lyon × Rush.
47 orders a month meeting 12 creator-hours a week. Uber Eats France is outrunning supply. This is where the next contract renegotiates.
Here’s the full map.
Metro × shoot type. Red cells with hot dots = fleet acquisition priority. Green = healthy. Grey = marginal.
Stop recruiting photographers globally. Recruit for Lyon Rush, Chicago Rush, Manchester Rush — in that order. Everything else is noise.
Utilization down. NPS collapsed. Payments running slow. Retention halved. Four separate numbers telling one story.
Payment speed to fourteen days. Weekly pay for Tier 1 creators. Do this before anything else. Credibility compounds. So does its absence.
The nine metrics above are the skeleton. These are the stances I’d walk in with — places I’ve already done the thinking, things I’d refuse to give ground on without evidence.
“Fill rate is the only marketplace metric that matters. Everything else is lagging or gameable.”
It decomposes into three drivers — match, accept, complete — each with its own owner and lever. I would refuse to goal on GMV alone.
“Stop reporting thirty-five thousand creators. Start reporting activated creator-hours per metro.”
Raw count is a vanity metric that lets ops game the number. Density per metro, measured in hours of availability, is the real supply.
“Demand-side and supply-side liquidity are different problems. Don’t blend them.”
A blended fill rate obscures whether we’re short on activated creators or short on orders. The fix list for each side is different.
“Quality needs an explicit threshold with written triggers, not a culture.”
Completion rate below 70% plus three incidents in 60 days equals deactivation. Above 95% equals priority routing. If we can’t state the rule in one sentence, we don’t have a quality program.
“Matching is a product, not an algorithm.”
Match quality rots silently. The team needs explicit matching-health metrics — viable-match density, time-to-match, cold-start boost — separate from supply growth. In commoditized marketplaces, a hundred milliseconds of latency equals one percent of supply.