Ask ten agency owners their gross margin and you’ll get ten different definitions and three honest answers. So let’s define it properly, look at what the data actually says, and then look at what happens to the numbers when you put software you own underneath the service.
Gross margin for a service business: revenue minus everything it costs to deliver the work. That means delivery labour (including account managers who service the accounts, not just the hands-on-tools people), the tools and software used in delivery, and any direct costs like contractors or API spend. What’s left pays for sales, marketing, admin and profit.
The two most common ways owners flatter the number: pushing account managers into “overheads” so delivery looks leaner than it is, and not pricing their own time when the founder still does client work. If your gross margin only looks healthy because you work for free, it isn’t healthy.
The best recent dataset I’ve seen for our market is The Shrinking Middle, the 2026 Agency Growth Report from InsightLeaf and Minimalist Marketing Co, built on 74 Australian agencies averaging 14 people and $3.4M revenue. (Disclosure: I’m quoted in it.) The headline numbers:
Published benchmarks tell the same story from the gross side: the commonly cited target for agency delivery margin is 50-60%+ at the P&L level, while Promethean Research puts the average digital agency at around 13% after-tax net, with niched agencies reporting dramatically better margins than generalists.
Put together, the honest picture for a pure service business at 20+ headcount: gross margins mostly land in the 30-50% band (the 50-60% target exists precisely because most agencies miss it), and net lands in the teens.
| Pure service | Service With Software | Pure SaaS | |
|---|---|---|---|
| Gross margin | 30-50% | 50-70% | 80%+ |
| Scales with | Headcount | Data | Servers |
A pure service scales with headcount: every new dollar of revenue needs more people, and the benchmark data above shows what that does to margins as you grow.
A pure SaaS scales with servers at 80%+ gross, which is why everyone spent a decade trying to build one. The thin point solutions will feel pressure as customers realise they can build a 70% version in-house, but deep platforms, systems of record and products with real network effects keep their economics.
Service With Software sits in between: roughly 50-70%, scaling with your data. You still need people, because people make the calls AI can’t. But every engagement makes the system better, and the system makes every person more productive. The service funds the software, the software compounds the service. The full model: Service With Software.
I’ll be straight about where StudioHawk is on this curve: early days. The margin gain so far is real but not dramatic, and it comes through labour: the quality of service has gone up, and the output the average person can produce has gone up, so we make margin on labour without squeezing anyone.
The bigger shift is pricing power. When your delivery runs on software trained on your own data, the work stops being comparable line-by-line with commodity alternatives, and that opens the door to value-based pricing instead of hours. That’s the mechanism that moves you from the first column toward the second, and it compounds: notice it’s the same mechanism the high-growth agencies in the data are using (fewer services, repeatable delivery, return on time), with software accelerating it.
Now the warning label, because this model has a failure mode: the lighter touch your service gets, the better your margin, and the more replaceable you become.
We live this trade-off ourselves. StudioHawk runs two offerings: our core service for scaling brands, and Early Bird, a lower-priced offering for smaller businesses. Early Bird is deliberately more automatable, more systematised, more light-touch, and it’s brilliant at what it’s for: getting a brand from zero to one. It’s also, by design, the more replicable of the two. It gets clients 70% of the way.
The core StudioHawk service is the opposite: creative, judgement-heavy, strategy-driven work that gets brands from 7 to 10. That’s the 30% AI doesn’t do, and it’s where the defensible margin lives.
The lesson generalises: automate delivery aggressively, but know which layer of your offer is the moat. If everything you sell is light-touch, your margin is just a countdown clock. The escape routes are a data moat or a legislative one, and for most service businesses only the first is on offer: how to build it.