Gross margin defines how much revenue is left to fund growth.
Gross margin — revenue minus cost of revenue, divided by revenue — is one of the most consequential and most inconsistently calculated metrics in B2B SaaS. Traditional SaaS benchmarks (75-85%) don't apply cleanly to AI-heavy products, high-touch enterprise deployments, or vertical SaaS with meaningful implementation costs. Getting the calculation right — and knowing your realistic benchmark — is prerequisite to running any other unit economics math.
Hosting and infrastructure (AWS, GCP, Azure). Third-party APIs consumed on behalf of customers (LLM inference, payment processing, SMS/email, data providers). Customer support headcount (tier 1 + tier 2 support, not solutions engineering). Customer success headcount tied to a specific delivery obligation (implementation, onboarding, technical account managers — this is debated). Software licenses for tools embedded in the product delivery (not internal-use software). Data storage and CDN costs. Not COGS: R&D, sales, marketing, G&A, executive team, general CS not tied to specific customer delivery.
Horizontal SaaS with self-serve product: 80-85% (Slack, Notion, HubSpot). Vertical SaaS with moderate implementation: 70-80% (Toast, Procore, Veeva). Infrastructure SaaS (databases, observability): 70-78% (Snowflake, Datadog). Marketplace or transactional models: 40-60% (Shopify Payments, Stripe). Below-benchmark gross margins are fixable but require deliberate work — infrastructure optimization, pricing power, and reducing human-touch delivery.
AI-heavy products routinely see 50-70% gross margins because LLM inference costs are meaningful and often not fully passed through to customers. A product where each user query costs $0.05-0.50 in inference and is priced at $0.10-1.00 has gross margin below traditional SaaS. Levers to improve: use smaller/cheaper models where quality allows, cache repeated queries, batch inference, negotiate committed-use pricing with providers, price with usage tiers that pass cost through at scale, invest in in-house models for high-volume workloads.
Gross margin is not just a cost problem — it's a pricing problem. Products with strong pricing power (few substitutes, high switching cost, high value) can raise prices to absorb COGS increases. Products with weak pricing power (commodity features, easy switching, unclear value) can't. If your gross margin is below benchmark, first ask whether the product is priced correctly for the value delivered before optimizing costs. Under-priced products chronically look like margin problems but are actually pricing problems.
Startups often win early enterprise deals by promising heavy implementation and dedicated support. This produces revenue but compresses gross margin to 55-70%. The trap: this feels fine at $5M ARR ("we're building relationships") but becomes structural at $20M ARR because customer expectations are set and reducing service is impossible without churn. Fix early: productize implementation, self-serve onboarding paths, tiered support levels. Adding these at scale is 10x harder than building them from the start.
Excluding customer success or support from COGS (inflates gross margin 5-15 points). Excluding third-party APIs (understates true COGS). Not tracking gross margin by product line (masks unhealthy products). Reporting best-case gross margin from a specific segment as the company number. Not passing through variable costs (AI inference, usage-based charges) in pricing.
Investor directory · Fundraising library · Articles A–Z · Company funding database