Software-as-a-service metrics are the common language that founders, investors, and operators use to evaluate the health, growth trajectory, and value of subscription-based technology companies. For AI startups specifically, these metrics take on additional significance because AI businesses often have unique cost structures — particularly around compute and model training — that can dramatically affect unit economics. Understanding and tracking the right SaaS metrics is essential for any AI founder seeking to raise capital, manage growth efficiently, and build a durable business.
ARR (Annual Recurring Revenue)
ARR is the single most important metric for any SaaS company. It represents the annualized value of recurring subscription revenue, excluding one-time fees, professional services, and variable usage charges. ARR provides a normalized view of the company's revenue run rate.
How to calculate: Take your current monthly recurring revenue (MRR) and multiply by 12. If a customer pays $10,000 per month on a subscription, that represents $120,000 in ARR. Only include committed, recurring revenue — not projected revenue or revenue from one-time engagements.
For AI companies, ARR calculation can be complicated by usage-based pricing models. Companies like Databricks and OpenAI charge based on compute consumption, which can fluctuate. In these cases, investors often look at "committed ARR" (contractually guaranteed minimums) separately from "consumption-based revenue" (variable usage above minimums).
Cursor, the AI-powered code editor, provides an interesting example of rapid ARR growth in AI. By offering a subscription product that developers use daily, Cursor has been able to build predictable recurring revenue while also benefiting from organic growth as developers recommend the tool to colleagues.
MRR (Monthly Recurring Revenue)
MRR is the monthly equivalent of ARR and provides a more granular view of revenue trends. It is particularly useful for tracking month-over-month growth and for decomposing revenue changes into their components:
- New MRR: Revenue from new customers acquired during the month
- Expansion MRR: Additional revenue from existing customers who upgraded or increased usage
- Contraction MRR: Lost revenue from existing customers who downgraded
- Churned MRR: Revenue lost from customers who canceled entirely
This decomposition reveals the underlying dynamics of a company's revenue base. A company with strong expansion MRR and low churn is in a very different position than one that relies entirely on new customer acquisition to grow.
NRR (Net Revenue Retention)
Net revenue retention, also called net dollar retention, measures how much revenue a company retains from its existing customer base over a period, including expansion and contraction. NRR is calculated as:
NRR = (Starting MRR + Expansion - Contraction - Churn) / Starting MRR x 100
An NRR above 100% means that existing customers are spending more over time, even after accounting for customers who leave. The best enterprise SaaS companies achieve NRR of 120-150% or higher, meaning their existing customer base grows by 20-50% annually without any new customer acquisition.
For AI companies, NRR is often exceptionally high because AI products tend to become more valuable as customers integrate them deeper into their workflows. Glean, the AI-powered enterprise search company, exemplifies this dynamic — as more employees in an organization adopt the platform and more data sources are connected, the product becomes increasingly indispensable, driving natural expansion.
Investors consider NRR above 130% to be exceptional, 110-130% to be good, and below 100% to be a warning sign that the product may not be delivering sustained value.
Churn Rate
Churn measures the rate at which customers or revenue is lost over a given period. There are two important types:
Customer churn (logo churn) — The percentage of customers who cancel their subscriptions in a period. Calculated as: customers lost / total customers at start of period.
Revenue churn (gross revenue churn) — The percentage of MRR lost to cancellations and downgrades. Calculated as: (churned MRR + contraction MRR) / starting MRR.
Revenue churn is generally more important than customer churn because losing a large enterprise customer has a much bigger impact than losing a small individual subscriber. The best enterprise SaaS companies maintain annual gross revenue churn below 10%, and often below 5%.
AI startups should be particularly attentive to churn because the AI market is evolving rapidly. Customers may churn if a competitor releases a significantly better model, if the underlying foundation model APIs they depend on change pricing, or if they decide to build AI capabilities in-house.
CAC (Customer Acquisition Cost)
CAC measures how much it costs to acquire a new customer. It is calculated as:
CAC = Total sales and marketing spend / Number of new customers acquired
A more precise calculation separates sales costs (salesperson salaries, commissions) from marketing costs (advertising, content, events) and may also distinguish between different customer segments. Enterprise customers typically have much higher CAC than self-serve customers.
For AI startups, CAC can be uniquely favorable due to product-led growth dynamics. When an AI product delivers immediate, visible value — like Cursor's code completion or Glean's enterprise search — users adopt it organically and champion it within their organizations. This bottom-up adoption can dramatically reduce CAC compared to traditional top-down enterprise sales.
LTV (Lifetime Value)
Customer lifetime value estimates the total revenue a company will earn from a customer over the entire duration of their relationship. A simplified calculation is:
LTV = Average revenue per account (ARPA) x Gross margin / Customer churn rate
The LTV-to-CAC ratio is one of the most important efficiency metrics investors evaluate. A ratio of 3:1 or higher is considered healthy — meaning you earn at least three dollars in lifetime value for every dollar spent acquiring a customer. Below 3:1 suggests the company is spending too much on acquisition relative to the value it captures. Above 5:1 may indicate the company is under-investing in growth and could afford to spend more aggressively on customer acquisition.
Gross Margin
Gross margin measures the percentage of revenue remaining after subtracting the direct costs of delivering the product. For traditional SaaS companies, gross margins typically range from 70-85%, primarily reflecting hosting and infrastructure costs.
AI companies often face significantly different gross margin profiles due to the cost of compute. Running inference on large language models — responding to user queries in real-time — requires substantial GPU resources, and these costs scale with usage. Some AI companies have reported gross margins as low as 30-50% for their AI-powered features, well below traditional SaaS benchmarks.
Databricks has navigated this challenge by building an efficient data and AI platform that generates strong gross margins despite significant compute requirements. The company's ability to optimize infrastructure costs while delivering high-value analytics and AI capabilities has been key to its growth toward a potential IPO.
Improving gross margins is a critical priority for AI startups. Strategies include optimizing model inference costs, using smaller specialized models instead of large general-purpose ones, caching common queries, and negotiating better compute pricing. Investors pay close attention to gross margin trends, viewing improvement as a sign that the company is maturing its cost structure.
Burn Rate and Runway
Burn rate is the rate at which a company spends cash, typically measured monthly. Net burn accounts for revenue: if a company spends $2 million per month and earns $500,000, the net burn is $1.5 million per month.
Runway is how long the company can operate at its current burn rate before running out of cash: cash on hand divided by monthly net burn. A company with $30 million in the bank and $1.5 million monthly net burn has 20 months of runway.
AI companies often have higher burn rates than traditional SaaS startups due to compute costs, expensive ML engineering talent, and the need to invest heavily in model development before generating revenue. Investors generally expect companies to have 18-24 months of runway after a funding round.
The Magic Number
The magic number measures sales efficiency — how much new ARR is generated for every dollar spent on sales and marketing:
Magic number = Net new ARR in the quarter / Sales and marketing spend in the prior quarter
A magic number above 1.0 indicates highly efficient growth. Between 0.5 and 1.0 is acceptable. Below 0.5 suggests the company needs to improve its sales efficiency before scaling further.
The Rule of 40
The Rule of 40 is a heuristic that combines revenue growth rate and profit margin (or free cash flow margin). The sum of the two should be at least 40%:
Revenue growth rate (%) + Profit margin (%) >= 40%
A company growing at 60% with a -15% profit margin scores 45 — above the threshold. A company growing at 20% with a 10% profit margin scores 30 — below the threshold.
The Rule of 40 helps investors evaluate the trade-off between growth and profitability. High-growth AI companies can justify negative margins if their growth rate is high enough, but as growth decelerates, profitability must improve to maintain investor confidence.
Why These Metrics Matter for AI Startups
For AI founders, SaaS metrics are the vocabulary of investor conversations. When Databricks, Glean, or Cursor discuss their business with investors, these are the numbers that determine valuations, funding round sizes, and the terms of investment. Founders who understand and track these metrics rigorously can tell a compelling growth story, identify problems early, and make data-driven decisions about where to invest resources. In the competitive AI landscape, operational discipline around metrics is often the difference between companies that scale successfully and those that flame out despite having strong technology.