ARR used to be the gold standard. For today’s AI startups, it might be fool’s gold.
When I started as a VC associate in 2017, my first and most powerful lesson was in SaaS metrics. In that era, the pricing was flat/recurring and the math was simple: if a high-growth startup checked certain metric boxes (90%+ annual retention, 80%+ gross margins, and >3x LTV/CAC), they got a term sheet valued at 10x ARR.
Once VCs embraced this math, ARR multiples became the standard for startup valuations. Founders leaned into SaaS pricing models (even if their product didn’t fit the true SaaS mold) to maximize their valuations, while VCs grew comfortable assigning ARR multiples to companies with weaker retention and gross margin profiles. Along the way, we lost sight of an important truth: not all revenue counts as ARR, and not all ARR is created equal. You’ve probably seen stories and posts highlighting this topic.
In the age of AI, startups are shifting away from SaaS pricing models and muddling the metric-based rubric used to guide and grade startups for nearly a decade. The time has come for investors and founders to go beyond surface-level ARR for budgeting, forecasting, and valuation purposes.
The AI Conundrum: Why ARR Is Losing Its Shine
Usage-Based Pricing Dominance: AI infrastructure and tools often rely on usage-based pricing (tokens, compute hours, and API calls) making “recurring” revenue inherently more volatile.
Retention Black Holes: Early AI adopters often experiment, then churn. Without years of retention data, projecting “annualized” anything becomes a leap of faith.
Easy On-Ramps Create Easy Outs: AI startups often minimize sign-up friction (free trials, early opt-outs, etc.) to capitalize on post-launch buzz. When their spotlight fades, curious experimenters move on and founders realize their “ARR” wasn’t truly bankable.
Rethinking “Quality” ARR Across All Startups
Not all ARR is created equal. It’s important for all startups (AI and otherwise) to develop a realistic view of how “recurring” their revenue truly is in order to budget accurately and set appropriate expectations.
To assess the true quality of ARR, consider these four criteria:
- Retention & NPS: The most useful indicator of future customer behavior is past customer behavior. Examine retention metrics—logo, gross dollar, net dollar—and customer satisfaction scores (NPS/CSAT) to gauge the stickiness of your revenue.
- Contract Terms & Billing Cadence: Contracts with longer durations, upfront payments, auto-renewals, price escalators, and cancellation penalties are more bankable. In contrast, offering month-to-month agreements (or a penalty-free opt out after 90 days) will likely translate to higher churn.
- Customer Size: Larger customers generally offer more stable budgets and longer lifecycles, making them more reliable sources of recurring revenue. Consumers sign up and cancel on a whim, while enterprises vet thoroughly and commit for the long-haul.
- Bone, Muscle, or Fat?: Determine whether your product is mission-critical (bone), important but not essential (muscle), or a nice-to-have (fat). When times get tough, customers will cut marketing tools and employee perks before they consider downgrading their ERP.
Final Take
The rise of AI demands a reset in how we value and contextualize growth. ARR still matters—but only when it passes the sniff test of true stickiness, contract durability, and economic clarity.
