Usage-based SaaS companies often report gross margins that look similar to traditional SaaS — 70-80% is common. But the aggregate number hides significant variance by cohort, customer size, and product intensity, and that variance is where your real business story lives.
Here's the math most usage-based teams aren't doing well:
Cost per unit by customer tier. Your enterprise customers likely consume your product very differently from your SMB customers. If your enterprise accounts are high-volume AI workflows with expensive inference per transaction, and your SMB accounts are lightweight human-initiated workflows, your blended gross margin hides that enterprise is significantly more expensive to serve.
Margin by product feature. Not all features in your platform have the same COGS. Your AI-powered features may have 50-60% gross margin. Your data storage features may be at 90%. Your human-assisted services may be at 30%. The blended margin hides whether you're growing the high-margin or low-margin parts of your business.
Margin trajectory vs. margin point-in-time. Is your gross margin expanding or contracting as usage grows? Ideally, as you get more usage data and optimize inference, your AI feature costs decrease while your pricing holds. If that's not happening, you have a structural problem.
Infrastructure cost scalability. At 10x usage, does your infrastructure cost go up by 10x, or by less? The leverage point in usage-based COGS is whether your infrastructure costs scale sublinearly with usage. If they don't, your margins will compress at scale.
The benchmark that matters for usage-based companies isn't the industry-average gross margin percentage. It's whether your margin per unit of value delivered is improving over time.
Track the trend, not the point. That's where your business model is revealed.