The GaaS Rule of 40, Reconsidered: Why the SaaS Yardstick Breaks on Agents

Table of Contents

• What the Rule of 40 Actually Measures
• Why GaaS Breaks the Two Hidden Assumptions
• The Growth Term Is Noisier Than SaaS Ever Was
• The Margin Term Has a Floor SaaS Never Had
• A GaaS-Adjusted Rule of 40
• Worked Example: Two Agent Companies, Same Score
• When the Rule Still Earns Its Keep
• Insights Most People Overlook
• References

What the Rule of 40 Actually Measures

The Rule of 40 is a back-of-napkin health check for software businesses: add your year-over-year revenue growth rate to your profit margin, and if the sum is at or above 40%, you're considered a balanced, fundable company. A startup growing 80% while burning at a negative 40% margin scores 40. A mature company growing 15% at a 25% margin also scores 40. The point was never precision. It was a single number that captured the trade-off every software CEO actually faces, spend to grow, or hold back and harvest, and let investors compare a hypergrowth startup against a profitable incumbent on one axis.

It caught on because SaaS made the inputs clean. Revenue was recurring and contracted, so "growth" was a stable, forward-looking quantity. Margins were high and structural, once you'd built the software, the marginal cost of another customer was close to zero, so "profit margin" mostly reflected how much you chose to reinvest in sales and R&D, not the cost of delivering the product. Both terms behaved. Bessemer and others popularized the metric precisely because, in that world, the Rule of 40 reliably separated efficient growth from unsustainable burn. The benchmark worked because the business model underneath it was uniform.

Agentic AI is not that world.

Why GaaS Breaks the Two Hidden Assumptions

Strip the Rule of 40 down and it rests on two quiet premises. First, that revenue growth is a meaningful, stable signal, that this quarter's growth tells you something durable about next quarter's. Second, that gross margin is high enough and stable enough to be treated as a near-constant, so the "margin" term is really a proxy for spending discipline rather than for cost of goods.

GaaS companies, agents sold per task, per outcome, or per run rather than per seat, violate both. Revenue tracks customer usage, which swings with their workload, their internal adoption curve, and even their own seasonality. And a chunk of every dollar earned flows immediately out the door to model providers, vector databases, and tool-call APIs. The cost of delivering an agent run is not near-zero. It can be 30, 40, sometimes 60 cents on the dollar before a single human at the company gets paid. That's not a SaaS cost structure. It's closer to a cloud reseller's, which is exactly the comparison some operators are starting to make as they realize compute is becoming the new cost of goods sold. When both of the rule's load-bearing assumptions wobble, the output number wobbles with them.

The Growth Term Is Noisier Than SaaS Ever Was

In a subscription business, a customer who signs a $50K annual contract contributes $50K of ARR whether they log in daily or never. The revenue is decoupled from behavior, which is what made it forecastable. GaaS deliberately recouples them. If a support-automation agent resolves 12,000 tickets one month and 4,000 the next because the customer cleared a backlog, revenue drops by two-thirds, and nothing about the relationship has actually deteriorated.

This makes the growth term in the Rule of 40 jumpy in a way SaaS growth almost never was. A single large customer ramping a new workflow can spike your quarter; the same customer pausing a project can crater it. Quarter-over-quarter annualized growth, a common way to compute the rule, becomes nearly meaningless because you're annualizing noise. The result is a score that looks spectacular one quarter and alarming the next, describing volatility more than health. This is the same forecasting headache that makes per-task pricing notoriously hard to model, and it lands squarely inside the one metric the Rule of 40 cares most about.

There's a subtler problem too. Usage-based revenue can grow for reasons that aren't good news. If your agent starts retrying failed tasks more aggressively, or a workflow gets chattier and burns more runs to do the same job, billed usage rises. Revenue went up. Value delivered did not. A naive Rule of 40 reads that as growth and rewards it. The metric can't tell expansion from inefficiency, because it was never designed to.

The Margin Term Has a Floor SaaS Never Had

Here's the part that quietly breaks the rule. In classic SaaS, you could choose a negative margin. Burning cash to grow was a strategy, not an accident, you spent on sales and marketing because the underlying gross margin was 80% and you knew it would be there waiting once you eased off the gas. The Rule of 40 traded growth against margin precisely because margin was elastic and recoverable.

GaaS gross margin is neither as high nor as freely chosen. Every agent run incurs a hard, external cost. You cannot grow your way past it, running more agents means paying more inference, not less. So the margin you plug into the Rule of 40 is anchored to a cost of goods you don't fully control, set by a handful of frontier model providers whose token pricing you can read off a published rate card but cannot negotiate at modest volume. When OpenAI or Anthropic or Google adjusts prices, or your agent's reasoning behavior shifts and it burns more thinking tokens per task, your gross margin moves without you touching a thing.

This means the rule's two terms are no longer independent, and that's the deeper flaw. In SaaS, growth and margin were genuinely separate levers. In GaaS they're coupled: pushing growth (more usage) directly pushes up COGS, which pushes down margin. You can't add two numbers as if they trade off cleanly when one mechanically drags the other. The arithmetic still works. The economic story it's supposed to tell falls apart.

A GaaS-Adjusted Rule of 40

The rule doesn't have to be thrown out. It has to be re-specified. Three changes make it usable.

Use net revenue retention, not raw growth

Raw period-over-period growth is too noisy for usage-based businesses. Swap it for net revenue retention measured on a cohort, how much a set of customers acquired in a given period is spending now versus then. NRR smooths out the lumpiness of individual usage spikes and, crucially, distinguishes real expansion (customers genuinely doing more) from one-off surges. A GaaS company with 130% NRR is durably growing inside its base; one showing 200% quarter-over-quarter growth off three big customers may not be. McKinsey's work on what separates the fastest-scaling software companies leans hard on retention quality for exactly this reason, and the logic transfers cleanly to agents.

Use gross margin, not a softer profit line

Many Rule of 40 calculations use EBITDA margin or free cash flow margin. For GaaS, insist on a gross-margin-inclusive view, because the whole point is that COGS is no longer negligible. A GaaS company posting strong "operating margin" while running 35% gross margins is hiding its single most important constraint. If the margin term doesn't reflect inference and tool-call costs, the Rule of 40 is measuring a SaaS company that doesn't exist.

Set the threshold higher, not at 40

If your gross margin ceiling is structurally lower than SaaS, say it tops out in the 55-70% range rather than 80%+, then a "passing" combined score should arguably be higher to represent equivalent quality, or you should benchmark against a GaaS-specific cohort rather than the SaaS canon. A GaaS company hitting 40 with thin margins is in a materially worse position than a SaaS company hitting 40, because it has less room to convert growth into profit later. The number is the same; the future it implies is not.

Worked Example: Two Agent Companies, Same Score

Picture two GaaS startups, both scoring a tidy 45 on the classic rule.

Company A grew billed usage 70% year over year and runs at a negative 25% operating margin. Underneath, its gross margin is 68%, its NRR is 125%, and its growth comes from existing customers expanding into new workflows. The negative operating margin is sales and engineering spend, recoverable, deliberate. This is a healthy 45.

Company B also nets to 45: 85% usage growth against a negative 40% margin. But its gross margin is 38% because it passes through three vendors' models with no caching discipline, its NRR is 95% (the base is quietly shrinking), and a third of that growth is one enterprise pilot plus an uptick in agent retries inflating run counts. Same score. Completely different company. The classic Rule of 40 rates them identically and the adjusted version, NRR-based growth, gross-margin term, higher bar, instantly pulls them apart. That gap is the entire argument for not taking the raw number at face value in this category.

When the Rule Still Earns Its Keep

None of this means the Rule of 40 is dead weight for agent companies. As a rough, late-stage triage tool, is this business anywhere near balancing growth and efficiency?, it's still a fast gut check, and investors will keep applying it whether or not it fits, so operators should know their number. It's also genuinely useful internally as a forcing function: computing it honestly makes a GaaS team confront its real gross margin instead of waving at "we'll pass through model costs," which is its own well-documented margin trap.

The mistake is treating it as a verdict. For SaaS it was close to one because the model was uniform. For GaaS it's the start of a conversation, a number that should immediately prompt three follow-ups: What's the gross margin underneath this? Is the growth retention or a spike? And how exposed is that margin to the next model price change? Answer those, and you're no longer applying a borrowed yardstick. You're measuring the business you actually have.

Insights Most People Overlook

• A rising Rule of 40 can signal a deteriorating business. Because usage-based revenue grows when agents retry, loop, or get chattier, the growth term can climb on pure inefficiency. A GaaS company can watch its score improve while it quietly destroys gross margin per task. No SaaS company ever had to worry that its growth metric and its cost discipline were the same lever pulled in opposite directions.

• The two terms aren't independent anymore, which is the real break. Everyone debates whether 40 is the right threshold. The deeper issue is that the rule assumes growth and margin trade off cleanly. In GaaS, growth mechanically consumes margin through inference cost. Adding two coupled variables produces a number, but not an interpretable one, the arithmetic survives, the meaning doesn't.

• Lower margins demand a higher passing score, not the same one. A GaaS 45 and a SaaS 45 are not equivalent achievements. The SaaS company can throttle spend and harvest 80% margins; the GaaS company hits a hard COGS floor it can't grow past. Holding both to "40 is the bar" silently grades the harder business on an easier curve.

• NRR is the metric the Rule of 40 was secretly relying on all along. SaaS could use raw growth because contracts made revenue sticky, retention was baked in. GaaS strips that out, so you have to add retention back explicitly. The adjusted rule isn't really a new metric; it's the old one with the assumption it quietly depended on made visible.

• The most dangerous GaaS companies are the ones that score well early. A startup with a few big usage spikes and deferred infrastructure costs can post a gorgeous Rule of 40 right before its margin reality and invisible usage-based churn catch up to it. The metric's blind spots are most flattering precisely when the underlying risk is highest.

References

• Bessemer Venture Partners, Scaling to $100 Million and the Rule of 40
• Andreessen Horowitz, The Cost of Compute
• Anthropic, Model Pricing
• McKinsey & Company, SaaS and the Rule of 40: Keeping the Balance as You Scale