Do the classic SaaS metrics still work if we are an AI native company?

Names precisely which classic metrics stop working when revenue is consumption based and inference costs eat gross margin, then proposes replacements. The free preview alone reframes the problem.

Rethinking SaaS metrics for AI

From Growth Unhinged by Kyle Poyar 14 min read

• Poyar's argument: ARR is becoming untrustworthy for AI-native companies, and DAU/MAU break when the product runs as digital labour or inside another product via MCP.
• He does not trust LTV for any AI product right now, given experimentation budgets and the shipping pace of Anthropic and OpenAI.
• Mixed monetisation splits revenue into high-margin platform and low-margin tokens, so a single gross margin line hides the business.

Survey data from over eight hundred private companies, with an efficient growth matrix that plots CAC payback against NRR so you can locate yourself rather than just read averages.

What's Really Going on in Software: The 2025 SaaS Benchmarks Report

From Growth Unhinged by Kyle Poyar 20 min read

• Across 800 plus participants, AI-native companies grew 110 percent at 1 to 5M ARR against 40 percent for other B2B SaaS.
• At 5 to 20M ARR the gap is 90 percent median growth for AI-native against 30 percent.
• ARR per FTE jumped 42 percent for 20 to 50M companies and 50 percent above 50M.
• Early-stage gross margin compressed nearly 10 points year on year.
• 70 percent of companies have shipped AI features, and 36 percent say AI is core to the product.

The annual dataset on the hundred best private cloud companies, including growth rates, multiples, and how long it now takes to reach a hundred million dollars ARR. This is the definition of best in class, with numbers attached.

The Cloud 100 Benchmarks Report 2025

From Bessemer Venture Partners by Byron Deeter, Elliott Robinson, Sameer Dholakia and the Bessemer Atlas team 30 min read

• The 2025 Cloud 100 averaged 75 percent year-over-year revenue growth and an 11.2 billion dollar valuation.
• The average honoree took 7.5 years to reach 100M ARR; AI companies got there in 5.7.
• Average multiple is 20x ARR, down from 23x, with AI companies at 24x against 19x for non-AI peers.
• Public cloud comparables sit near 8x ARR, down 41 percent from the 2021 peak.

This question is about which classic metrics break when revenue swings with consumption, and this is a short, concrete fix for one of them: how to compute payback when there is no stable subscription number to divide by.

Adapting CAC Payback Period for Usage-Based SaaS Models | SaaS Metrics School | CAC Payback Period

On The SaaS CFO by Ben Murray 4 min