SaaS cost structure

SaaS P&Ls are usually read as percentages of revenue. Early on those percentages look insane. Later they should compress. The mistake is cargo-culting a public company's cost structure when you have $2M ARR — or treating early S&M above 100% of revenue as proof the model is broken.

The question is payback and whether cohorts retain, not whether this quarter's S&M "looks efficient."

The big idea

A bakery that is still finding its recipe should not have a national advertising budget that looks like a grocery chain's. A bakery that has a line out the door should spend to open the next shop.

The analogy breaks on software: R&D does not scale linearly with each new customer the way extra ovens do. That is operating leverage — once the product is built, R&D and G&A can grow slower than revenue. AI inference can take some of that leverage back by pushing COGS up.

Walking through it

Classic software COGS often leaves 70–85% gross margin at scale. AI features can push COGS up (inference).

Implementation, training, and custom engineering can raise TCV and help land elephants. They also lower blended gross margin. A common operator rule: keep services as a bridge to software revenue, not as the business. If services are more than about 20–30% of revenue forever, you may be a consultancy with a login screen. Track services margin separately from subscription margin.

Common mistakes

• Copying a public SaaS opex mix at $2M ARR. Early S&M heaviness is normal if cohorts pay back.
• Ignoring support in COGS. Human success teams scaled linearly with logos until margins collapsed.
• Letting services become the company. High TCV, low software margin.
• Treating R&D % as a moral number. On small revenue, R&D should look large.

Check yourself

S&M is 120% of revenue at $2M ARR, and new cohorts pay back in 9 months with 90% GRR. Is the model broken?

Not by the S&M percentage alone. Payback and retention say the spend is buying a base.

When do services stop being a bridge?

When they stay above roughly 20–30% of revenue and the login screen is a delivery vehicle.

Key takeaways

• Read SaaS costs as % of revenue, and expect the mix to shift by stage.
• Early S&M above 100% of revenue can be rational if payback and retention work.
• Operating leverage is R&D and G&A growing slower than revenue — unless AI COGS eat it.
• Keep services as a path to software revenue. Track that margin separately.

Further learning

SaaS Metrics 2.0 (David Skok)“If you cannot measure it, you cannot improve it” – Lord Kelvin This article is a comprehensive and detailed look at the key metrics that are needed to understand and optimize a SaaS business. It is a completely updat…For Entrepreneurs

Next: the cash problem this cost structure creates — the SaaS cash trough.