"Signs You Should Move from SaaS to Usage Based Pricing" — CJ Gustafson (Mostly Metrics)

Why this is in the vault

CJ synthesizes three Run the Numbers CFO interviews (Confluent's Rohan Sivaram, Figma's Praveer Melwani, Couchbase's Greg Henry) into a single diagnostic: the trigger for moving off subscription pricing isn't founder preference, it's gross-margin decay from AI-inference or per-usage COGS creeping up until "you can't afford not to." Directly on-topic for RDCO's Snowflake/consumption-pricing thesis work.

Mapping against Ray Data Co

The load-bearing line for RDCO: CJ's three-item diagnostic checklist (variable cost to serve, cost-to-serve smallest vs. largest customers, gross-margin trend ex-CS headcount) is a reusable lens for the founder's Snowflake GenAI cert study and the broader phData consumption-pricing positioning — it gives a concrete "when does a vendor's usage-based pricing story hold water" test rather than treating UBP as a self-evidently superior model. It also reinforces the "one-way door" framing already logged from the Databricks sales-comp piece: switching pricing models restructures sales comp, rev rec, and forecasting org-wide, which is exactly the kind of irreversible-decision caution that should temper any RDCO move toward usage-based pricing for MAC or Squarely before the cost-to-serve data justifies it.

The core argument

Subscription-to-usage pricing is a one-way door, not a pricing tweak — it rewrites sales comp, rev rec, and forecasting cadence. Confluent's CFO described discovering the need for consumption pricing directly in COGS (the marginal cost of a cloud customer interaction is no longer ~zero, unlike on-prem software), and noted the pattern was nearly comp-less five years ago (only Snowflake and MongoDB were doing it publicly at scale). Figma's CFO quantified the AI-inference hit directly: gross margin went from ~90-91% pre-AI-features to ~86% by Q4 as inference costs to power new features landed in COGS, which is why Figma pre-loaded free "AI credits" into all seats before monetizing heavy usage — avoiding the trap where flat pricing makes a company root against its own product being used. Couchbase's CFO described the operational rebuild required: tracking per-customer consumption curves, over/under-consumption alerts, and comp plans, replacing the old subscription rhythm where the signed contract was the end of the story with a consumption rhythm where signing is just the start. CJ's closing diagnostic: watch variable cost-to-serve, cost-to-serve-smallest-vs-largest-customer, and gross margin trend (net of CS headcount changes) — when all three degrade together, the pricing-model decision has effectively already been made.

⚠️ Sponsorship

Two disclosed relationships, both previously logged for this sender:

• Brex (confirmed recurring sponsor) — top-of-email paid block, "Mostly Media runs on Brex" framing pitching agentic-finance/AI-expense-automation positioning. Standard clean third-party sponsor slot, consistent with prior Brex placements for this sender.
• Mostly Talent (CJ's own recruiting arm) — "Job Stuff" self-promo block pitching finance-hiring placement services. This is self-consulting, not a third-party sponsor; flagged per the standing README note for this sender.

Neither relationship touches the substantive content (the Confluent/Figma/Couchbase CFO synthesis), which is podcast-sourced reporting rather than vendor-influenced framing.

Related

• [[2026-05-11-mostlymetrics-usage-vs-subscription-pricing]] — CJ's earlier contrarian re-frame that UBP amplifies rather than causes winning; this issue's "when do you actually need to switch" diagnostic is the operational complement to that thesis piece.
• [[2026-06-04-mostly-metrics-consumption-based-arr]] — CFO Mailbag on how to talk about ARR under consumption billing (Mercury, Notion CFOs); same terrain, reader-Q format vs. this issue's interview-synthesis format.
• [[2026-07-14-mostly-metrics-databricks-consumption-sales-comp]] — the sales-comp mechanics of a consumption pricing model in practice (Databricks), which this issue's Couchbase/Confluent quotes describe from the CFO-strategy side rather than the rep-payout side.