01 · Headcount By Phase

Small team, big leverage.

Running a marketplace is running a separate business. Different unit economics, different risks, different daily work — refereeing strangers’ money instead of shipping our own boxes. It deserves what any real business deserves: specialists, dedicated and properly staffed, not spare cycles from teams that have none. Escrow, spec-verification, and AI assist keep the headcount lean — but every role below is a dedicated, net-new hire, and we scale them responsibly against the ratios, never behind the volume.

The Name Of The Game Is Reputation

A marketplace has exactly one asset that compounds: trust. Every late claim response, every mis-graded set that slips through, every seller who waits on a payout spends it down. Staffing this table properly isn’t overhead — it’s how the reviews stay at 4.9, and the 4.9 is the moat. We’d rather scale hiring one quarter early than reputation-repair for a year.

Phase 1 · Pilot (mo 1–6)

Prove it

~150 orders/mo · ~1K live listings

4 net-new hires

Ops Lead · Marketplace Support · Trust & Moderation (one combined role) · Seller Growth — all dedicated to the Exchange from day one. Claims escalate from the T&M hire to the Ops Lead.

Phase 2 · Growth (mo 6–18)

Feed it

~600 orders/mo · ~5K live listings

6 dedicated FTE

Trust & Safety and Moderation split into their own seats; support goes to two. Ops Lead · 2 support · 1 T&S · 1 moderation · 1 seller growth. Engineering drops to maintenance (~0.5 shared).

Phase 3 · Scale (18mo+)

Run it

~2,000+ orders/mo · ~15K live listings

8–10 dedicated FTE

Support pod of 3 · 2 T&S · 1–2 moderation · 2 seller growth/community · 1 dedicated engineer. Ratios below govern from here.

02 · The Roles & Scaling Ratios

Who does what, and when to hire the next one.

Role	Owns	Scaling metric (AI-assisted)	Pilot	Growth	Scale
Marketplace Ops Lead / GM THE day-one hire	P&L, policies, Sharetribe Console, escalations, vendor relationships, this whole plan	1 — always. Hire for marketplace or ecomm-ops background	1	1	1
Marketplace Support	Buyer/seller questions, order issues, payout help — the Exchange's own dedicated line	1 FTE per ~1,500 orders/mo assumes ~10% contact rate · 30+ tickets/day/agent w/ AI drafts	1 (dedicated)	2	3
Trust & Safety / Disputes	48-hr inspection claims, evidence review, fraud escalations, KYC edge cases	1 FTE per ~2,500 orders/mo at <2% claim rate · ~30 min/claim w/ photo-compare tooling	1 combined w/ moderation	1	2
Listing Moderation / Spec-Check	The "needs human" queue only — the bot auto-passes catalog matches, auto-rejects rule-breakers (DOT age, bent wheels)	1 FTE per ~4,000 new listings/mo assumes 15–20% need human eyes · ~3 min each w/ AI triage	(in the T&M seat)	1	1–2
Seller Growth / Community the "sales" function of a P2P marketplace	Supply acquisition: seeding sellers from the gallery, power-seller program, YT explainer content, events	Driven by listing targets, not orders — supply is the growth constraint in year one	1 (dedicated)	1	2
Engineering	Build (10–12 wks), then integrations, template updates, wallet service	2–3 devs during build → 0.5–1 run-state	2–3 (build phase)	0.5	1 (dedicated)
Finance / Reconciliation	Nightly three-way recon (Stripe · Sharetribe · wallet ledger), credit liability, 1099-K oversight	Automated recon + exception review	automated + exception review	0.25	0.5
Dedicated FTE total	—	—	4 net-new (+build eng)	≈ 6	8–10

Why these ratios beat classic marketplace staffing

Traditional P2P marketplaces staff support at ~1 FTE per 600–800 orders/mo. Ours target ~2× that leverage because the product removes the ticket-generators: escrow kills "where's my money," spec-verification kills "it doesn't fit," the condition-report contract kills he-said-she-said disputes, and AI assist (below) handles the first draft of nearly everything. If the claim rate runs above 2% or contact rate above 12%, the fix is product, not headcount.

03 · The AI Assist Layer

Every role gets a copilot.

Baked into the ops tooling from day one — this is what makes the ratios above honest instead of optimistic.

Support drafting & deflection

Help-center answers resolve the common questions before a ticket exists; agents get AI-drafted replies with order context pulled in — review and send, not compose.

Spec-check triage

The bot matches listings against the catalog: auto-pass clean matches, auto-reject rule-breakers (DOT age, bent/repaired), route only ambiguity to a human — the ~10-minute promise at near-zero marginal cost.

Claim evidence summarization

Listing photos vs. buyer photos, side-by-side with an AI-written discrepancy summary — the referee reads a brief, not a folder of JPEGs.

Fraud & anomaly flags

Velocity spikes, mismatched geos, serial-return patterns, wallet abuse — flagged before a human would notice, routed to T&S with context.

Listing quality coach

Nudges sellers at creation time — missing angles, thin descriptions, price way off comps — fixing listings before they generate tickets.

Seller growth targeting

Gallery members whose builds changed wheels recently = a ranked outreach list of people literally holding inventory. The community team works a queue, not a hunch.

04 · The Dashboard The Ops Lead Lives In

Five numbers that trigger hiring.

<10%

Contact rate
(tickets ÷ orders)

<2%

Claim rate
(disputes ÷ orders)

<4 hrs

First response
(business hours)

<10 min

Spec-check
turnaround

<24 hrs

Claim resolution
(evidence in hand)

The hiring rule

When two of the five KPIs miss for two consecutive weeks at the current volume, the ratio table says who to hire next — staffing decisions become arithmetic, not debates. The ops console (spec-check + claims queues) already sketches the tooling these numbers come from.

Illustrative planning model for the board — ratios reflect AI-assisted marketplace-ops norms; validate against pilot actuals before Growth-phase commitments.