Skip to content
Back to articles
Enterprise15分

Usage-Based vs Subscription: SaaS Pricing Migration in the AI Era

藤井 恵美Principal Pricing Strategist
2026-04-2115分
SaaS PricingUsage-BasedSubscriptionAI APISnowflakeDatadog

The Decline of Fixed Subscriptions and the Rise of Consumption-Based Pricing

In 2026, the B2B SaaS industry is in the middle of a tectonic shift in pricing. OpenView Partners' SaaS Benchmarks report, published in February 2026, found that 61% of SaaS companies with ARR above $10 million have either already moved away from pure seat-based pricing or have a migration plan in progress. In 2023, that figure was 34% — the industry has essentially flipped in three years. Three structural forces are behind it. First, embedding AI features has caused "cost per user" to fluctuate dramatically with usage. Second, buyers have intensified pressure for "pay for what you use" billing. Third, investor-side pressure: fixed seat counts make it impossible to grow Net Revenue Retention (NRR).

Snowflake is the textbook example of usage-based pricing (UBP) — it has used a credit consumption model since its founding in 2012. Billing by compute warehouse uptime multiplied by size makes budget prediction difficult for customers, but allows revenue to grow naturally as workloads expand. Snowflake ended its most recent fiscal year with an NRR of 131%, combining pay-as-you-go with long-term storage price commitment discounts in a hybrid structure that balances predictability with elasticity.

Datadog and Twilio: Mature Patterns for UBP

Datadog illustrates another mature form of UBP. The company uses a multi-dimensional approach: a base charge by host count and monitoring module, layered with multiple consumption metrics — log ingest volume (GB/month), APM span count, synthetic test execution count. The most recent fiscal year's investor materials show 83% of existing customers using two or more products, and 31% using four or more. Multi-product times multi-dimensional is the structural driver of NRR.

Twilio's evolution is even more instructive. The company built its reputation on pure per-API-call UBP, but since 2024 has significantly strengthened Commitment-Based Pricing for enterprise customers. Committing to an annual spend amount in advance earns a 15–40% discount on unit pricing. This is a structural answer to the buyer-side frustration with UBP's unpredictability — it speaks directly to procurement teams and CFOs who need fixed numbers. By 2026, committed customers account for more than 78% of Twilio's revenue, and the old understanding of "UBP = variable costs" is already outdated.

AI API Era: "Tokens as the Unit of Consumption"

The central question of 2025–2026 is how to incorporate token-based pricing into a SaaS pricing model. The foundation model APIs from OpenAI, Anthropic, and Google all bill on a matrix of input tokens × output tokens × model unit price. If a SaaS product passes this raw consumption unit through to customers, customers facing uncertain estimates start suppressing usage of the AI features. The result is a perverse incentive where the highest-value workflows are the least-used ones.

Notion AI, Linear's AI features, and Vercel v0 each take different approaches. Notion AI uses a monthly flat-fee AI add-on with a usage cap — when the cap is hit, features are rate-limited but no overage charge is triggered. Predictable budgeting for buyers is the top priority. Linear bakes AI into its Plus plan and above without exposing token consumption directly. Vercel introduces a proprietary currency — v0 Credits — that are consumed based on the complexity of screen generation or code output, abstracting away raw tokens and making it easier to absorb unit price changes when swapping underlying models.

The emerging best practice in 2026 is a three-layer structure: credits system + tiers + overage discounts. Credits are distributed within tiers; overages are sold at progressively discounted rates. The internal accounting for credits is tied to actual cost (tokens × model unit price + inference infrastructure amortization + safety tooling API calls), but presented to customers as a single currency. This makes it possible to swap models behind the scenes without breaking either the customer experience or the accounting.

Impedance Matching with Japan's Annual Budget Cycle

This is where the Japan-specific challenges begin. Large Japanese enterprises — particularly listed companies — operate under an annual budget system where the standard practice is "reserve a fixed envelope at fiscal year start, and reconcile actuals at year end." Pure UBP is in direct conflict with this budget culture. A monthly invoice whose amount is uncertain is a friction point for purchase approval workflows, payment forecasting, and consolidated accounting.

Three solutions exist. First, the Prepaid Credit model: at contract signing, the customer purchases credits covering their estimated annual consumption in a lump sum, and draws down freely through the year. Unused credits are non-rollable, which allows the SaaS provider to recognize revenue aligned to the fiscal year. Second, True-Up / True-Down: actual consumption is reconciled quarterly, with shortfalls billed as additional charges. Unit prices for overages are fixed in the contract in advance, which aligns more easily with CFO budget management. Third, Commit + Flex (Twilio-style): an annual commitment with overage billed monthly in a flexible extension.

Domestic SaaS companies — MoneyForward, freee, Sansan, SmartHR — have each moved toward their own hybrid of seat + consumption pricing. When Sansan introduced AI business card analysis on a credit model in 2025, the original plan was pure UBP, but enterprise customer feedback led to a redesign as "annual commitment + automatic upsell on overage." In Japan's market, buyers place extreme value on the "no-surprise invoice" — that cultural preference must be built into the design.

The Execution Sequence for Migration Projects

Migrating an existing seat-based SaaS to UBP or a hybrid carries predictable pitfalls. The biggest risk is that existing customers' effective unit economics deteriorate — multiple cases have been reported where post-UBP launch invoices averaged 12–18% below prior fixed amounts. Preventing this requires analyzing each customer's consumption profile before migration and working backwards from "current fixed amount approximately equals expected consumption at new pricing" to design tier structures.

The recommended execution sequence: Phase 1, deploy a billing engine (Stripe Billing, Metronome, Orb, or m3ter) and build the metering event pipeline. Phase 2, offer the UBP plan to new customers only, and accumulate three to six months of production data. Phase 3, maintain grandfathered plans for existing customers while presenting migration incentives (first-year discounts, usage reports) at contract renewal time. Compressing these three phases leads to sales teams exhausted from explaining changes to customers, and an accelerating churn rate.

Another critical point: the right metric to track during a UBP migration is not Gross Margin in isolation but the combination of Net Dollar Retention × Gross Margin. Even if Gross Margin temporarily compresses, NDR exceeding 130% means aggregate cash flow over three years is a net increase. An organization that cannot explain this economic logic in an executive meeting will see the migration project killed for political reasons.

Closing: Pricing Is Part of the Product

The conclusion that 2026's pricing discussions have reached is that "pricing is not a sales negotiation — it is part of the product and the infrastructure." The billing system, metering pipeline, usage dashboard, and customer-facing cost optimization recommendations all need to be designed and treated as first-class product features. Snowflake and Datadog are strong because they have polished the "customer's financial experience" as a product. That is what sets them apart.

For Japanese SaaS companies competing on the global stage, the path forward is not importing Western pure UBP wholesale. It is developing the mathematical skill to design hybrid structures grounded in a deep understanding of how your customers' budget systems actually work. Pricing is the face of the product and the heart of the financials. The engineering that satisfies both constraints simultaneously is the essence of pricing strategy in 2026.

Let's solve your technical challenges together.

KGA IT Solutions delivers AI, cloud, and DevOps expertise to address your specific challenges.

Contact Us