Every agency eventually hits the same ceiling: revenue is a function of hours, and hours are a function of headcount. The benchmark data makes the consequence concrete — net margin falls from roughly 19% at under ten people to about 8% above fifty (Promethean Research, 2025). Recurring revenue that is not priced in hours is the usual escape.
There are seven common models. They are not equally good, and the differences are mostly about who owns the client relationship and whose margin you are inside.
1. Referral fees
Margin: 100% of a small number. Effort: minimal. Control: none.
You recommend a tool, the vendor pays a percentage. Nothing to build, nothing to support. The problems are that commissions typically end after 12 months, you cannot influence pricing or roadmap, and you have handed the client relationship to a vendor who may later sell them services you also offer. Fine as a byproduct, weak as a strategy.
2. Software resale
Margin: usually 20–30%. Effort: low. Control: limited.
You buy licences at a discount and resell at list. Better than referral because the billing relationship is yours, which means the renewal conversation is yours too. But the client sees the vendor's brand, and your margin is set by someone else's channel policy — and can be cut unilaterally.
3. White-label platforms
Margin: high. Effort: moderate. Control: high.
You resell a platform under your own brand and set your own prices. Public reseller-margin claims in this category run 40–80%, but note these come almost entirely from vendors marketing their own reseller programmes and should be treated as promotional rather than measured.
What is structurally true regardless of the exact percentage: you own the brand, the pricing and the relationship, and your cost per additional client is close to flat. The risk is concentration — your recurring revenue now depends on one supplier's uptime, roadmap and pricing decisions. Diligence the supplier the way you would diligence an acquisition.
4. Managed services on top of software
Margin: moderate. Effort: high. Control: high.
The platform fee is small; the management is the product. This is the most durable model because you are paid for judgement rather than licences, and it is the hardest to displace. It is also the one that scales worst — it consumes senior time, the exact resource that makes agencies unprofitable as they grow. Best combined with model 3 so the platform margin subsidises the human margin.
5. Productised retainers
Margin: high if scoped tightly. Effort: moderate. Control: total.
A fixed monthly deliverable at a fixed price. Retainer adoption reached about 78% of digital agencies in 2026, up from 64% in 2023, and retainer-led agencies are reported to retain roughly 2.3× better than project shops.
The failure mode is the retainer that is secretly hourly — "20 hours a month" — which caps your revenue while leaving scope open. Price against an outcome, cap the scope explicitly, and revisit it quarterly.
6. Hosting and infrastructure
Margin: moderate. Effort: low once built. Control: high.
Reliable, unglamorous, and it makes you responsible at 3am. Worth it when it protects a larger relationship; rarely worth it as a standalone line.
7. Your own product
Margin: highest eventually. Effort: very high. Control: total.
The honest benchmark: median SaaS CAC payback is about 16 months and median growth has fallen to roughly 26%. You are underwriting well over a year of negative cash flow per customer, from agency profits, in a slower market. Agencies that succeed here usually productise something they already do repeatedly and sell it to the clients they already have — not a net-new bet.
How to sequence these
Most agencies should not pick one. A workable progression:
- Start with 5. Productised retainers need no supplier and prove whether you can sell a fixed-scope outcome.
- Add 3. White label gives recurring revenue whose delivery cost is nearly flat per client, which is the specific thing that fixes the margin-versus- headcount curve.
- Layer 4 on top. Managed services attach to the platform you now control and lift average revenue per client without a new sale.
- Treat 1, 2 and 6 as byproducts. Take the money; do not build strategy on them.
- Attempt 7 only once the others fund it.
The metric to watch through all of it is revenue per employee — benchmarked at $150,000–$200,000 for healthy agencies. Recurring revenue is working when that number rises while headcount holds flat.
Sources and how to read them
Figures below are attributed where they appear. A note on quality: agency and SaaS benchmark data is mostly self-reported survey data, and response bias runs toward firms healthy enough to answer a survey. Vendor-published numbers are marked as such, because a company selling the thing it is measuring is not a neutral source. Treat these as directional benchmarks for comparison, not as audited accounts.
- Promethean Research — How Profitable are Digital Agencies?
- GigRadar — Retainer Pricing in 2026: Benchmarks
- Haus Advisors — Marketing Agency Industry Statistics 2026
- Aleph — CAC payback period benchmarks for SaaS (2026)
- Data-Mania — B2B SaaS Benchmarks 2026
Running the numbers on your own stack
If part of your cost problem is subscription sprawl rather than headcount, the savings calculator totals what your current tools cost against running the same functions in one place. Mewayz is $39 per active user per month, taken from the payments you process, with every module included — and free to start with no card.