The Cost-to-Serve Curve: Why MSO Marketing Economics Deteriorate Between 10 and 59 Locations
Driver-based cost modeling across a multi-location service portfolio shows marketing contribution per FTE falls as location count rises, producing a structural dead zone where mid-size MSOs are too big for boutique service and too small for enterprise economics. The mechanics, the price floors, and the way out.
There is a quiet assumption inside every MSO roll-up model: marketing gets cheaper per location as the platform scales. Shared brand, shared agency, shared tooling; the spreadsheet says the line item compresses.
We built a driver-based cost-to-serve model from the actual labor and tooling consumption of a live multi-location service portfolio, and the data says something more specific and less comfortable: certain costs compress, but the coordination stack grows faster than the shared-services savings through most of the mid-market range. The result is a curve with a sweet spot, a dead zone, and a second wind, rather than a smooth decline.
The model
Cost-to-serve here means the fully loaded cost of operating competent digital marketing for a location: campaign management, tracking/attribution maintenance, reporting, budget governance, creative refresh, and the coordination overhead of doing all of it across N locations that share a brand, a market, or a budget pool.
The model is driver-based. Each cost line is tied to a countable driver (accounts managed, tracking joins maintained, reporting surfaces, inter-location budget decisions) with consumption rates taken from observed operations rather than industry benchmarks.
What the curve shows
| Location count | Economics | Why |
|---|---|---|
| 1-3 | Expensive per location, but simple | Every cost is direct; no coordination layer; boutique service is affordable relative to revenue |
| ~9 | Sweet spot | Shared infrastructure amortized; coordination still fits inside one operating cadence and one senior brain |
| 10-59 | Dead zone | Coordination overhead (cross-location budget allocation, attribution reconciliation, per-location reporting) scales super-linearly; too small to justify enterprise tooling and dedicated data engineering, too big for the boutique model to hold quality |
| 60+ | Second wind | Enterprise tooling, centralized data layer, and dedicated roles become affordable; coordination is systematized instead of personal |
The counterintuitive core finding: contribution per marketing FTE falls as the portfolio grows through the mid-market range. Each additional location adds not just its own workload but a slice of coordination against every existing location: budget arbitration between sibling stores in overlapping markets, deduplicating attribution when customers cross locations, reconciling per-location P&L views of a shared budget. Headcount productivity peaks early and erodes until the platform is large enough to buy its way into systematized operations.
The pricing consequence: floors, not rates
Run the model backwards and it produces price floors: the minimum a competent provider must charge per location, by portfolio size, before margin. Three consequences follow:
- Any vendor quoting one flat per-location rate across the 10-59 range is either underserving you or losing money on you. Both end the same way, so it is worth finding out which.
- The cheapest credible bid clusters at the 9-location shape. Below the floor, quality is being silently withdrawn from somewhere. Usually that somewhere is attribution maintenance, because its absence is invisible until someone audits the revenue join. We ran that audit; the default interaction-to-revenue link rate across 1.45M records was ~0% (see the portfolio teardown).
- Dead-zone MSOs get the industry's worst service by structure, not by vendor failure. The boutique that served you well at 8 locations degrades by 25; the enterprise agency that would serve you well at 80 won't price sanely at 25. This is the curve, not bad luck.
The way out of the dead zone
The dead zone exists because coordination is being done by humans. The escape is making the coordination layer software before the location count forces the issue:
- Automated budget governance. Daily pacing against calendar across every account, with variances surfaced, replaces the largest single coordination cost in the model.
- A durable attribution join. Deterministic matching from ad interaction to repair-order revenue (caller ID and hashed identity, not modeled attribution) replaces the reconciliation meetings that otherwise consume senior time every month.
- Centralized raw-data archiving. Call detail, interaction logs, and platform exports, archived on a schedule, so every future question is answerable without re-negotiating vendor access.
In the book this model is drawn from, that software layer is what lets 31 locations run on a cadence the curve says should already be degrading. The dead zone is real, but it is an instrumentation problem wearing a headcount costume.
Methodology note: driver rates are taken from observed operations of the live 31-location book we manage (2025-2026), not from survey benchmarks. The model is deliberately conservative: coordination drivers are counted only where a named recurring activity exists. Portfolio and clients anonymized.