Your customers have already done the math. They are paying for 50 seats while 12 people log in, and sooner or later someone on their side puts that slide in a renewal deck. When a payments giant like Adyen spends $335 million in cash to buy a usage-based billing platform, it is betting that the "pay for what you use" conversation stops being a negotiation tactic and becomes the default way software is sold. If you run a subscription business, this deal is about your pricing page — not someone else's M&A strategy.
What Adyen Actually Bought
In June 2026, Amsterdam-based payments company Adyen announced a definitive agreement to acquire Orb, a San Francisco billing platform founded in 2021, for $335 million in cash. The deal closed on July 1, 2026, with Orb operating as a subsidiary under an incubator model and its co-founders reinvesting part of their proceeds into Adyen shares.
Orb is not a generic invoicing tool. It is an infrastructure engine that ingests raw usage events at scale — API calls, compute minutes, AI tokens, seats touched, messages sent — and turns them into rated charges and invoices. Its architecture stores the full event stream instead of aggregating usage early, which lets merchants decouple metering from invoicing: you can change pricing logic, re-run history, and backtest a new price against last quarter's actual consumption before showing it to a single customer. Its client list reads like a who's who of usage-priced developer platforms, including Vercel, Replit, Supabase, and Glean.
Adyen's stated logic is worth reading closely, because it describes your problem too. Billing and payments currently live in isolated silos: the billing system decides what to charge while knowing nothing about whether the charge will succeed, and the payments system executes transactions while knowing nothing about the pricing contract behind them. Connecting the two creates what Adyen calls a two-way intelligence advantage — billing signals improve fraud and risk models, while real-time payment data improves billing execution. In plain English: fewer failed charges, fewer false fraud declines, and fewer customers lost to a billing hiccup that looked like a collections problem but was really a data problem.
Why This Deal Is a Signal, Not Just a Headline
Acquirers do not pay nine figures for a billing engine unless their largest customers keep asking for it. Adyen said exactly that: existing and prospective customers are actively asking for help with complex, high-volume usage models, particularly as AI reshapes how software is priced and consumed. Three structural shifts sit underneath that demand.
Per-seat pricing is losing its grip. Per-seat plans remain the most common model, but buyers spent the 2022–2024 correction scrutinizing unused licenses, and they never stopped. One 2026 pricing study found that 42% of SaaS products now offer a usage-based option, up from 27% in 2023. Another industry dataset puts overall SaaS adoption of usage-based pricing above 59% in 2025, versus roughly 40% two years earlier. Consumption pricing has crossed from experiment to mainstream in about two years.
Usage-based companies expand faster. This is the growth math that makes CFOs pay attention. OpenView's research on usage-based pricing found that companies with a largely usage-based model posted top-quartile net dollar retention of 122%, compared with about 110% for tiered usage plans and 109% for no usage-based pricing at all. Related analysis found usage-based SaaS businesses growing revenue near 30% year over year versus roughly 22% for their peers — and seven of the nine recent software IPOs with the best net revenue retention ran usage-based models. When the price scales with value received, expansion happens without a renegotiation.
AI forced everyone's hand. Token metering, GPU minutes, per-resolution agent pricing — AI products cannot be priced per seat without either leaving money on the table or charging customers for value they never received. The billing infrastructure to meter millions of events in real time simply did not exist in legacy subscription tools, which expect a static quantity times a static price. Adyen buying Orb is the payments layer admitting that the pricing layer now drives the architecture.
Should You Add Usage-Based Pricing? A Decision Framework
Not every business should flip to metered billing. The companies that get hurt are the ones that copy Snowflake's model onto a product whose value has nothing to do with consumption. Before you touch your pricing page, run through these questions.
1. Is there a metric your customers already associate with value?
The best usage metric meets five tests: it tracks value the customer recognizes, it flexes up and down with their success, it scales without weird cliffs, it is predictable enough to budget against, and you can actually meter it reliably. Datadog charges on hosts, HubSpot on marketing contacts, Zapier on tasks — each one maps to something the buyer already believes is worth paying for. If you cannot name your metric in one sentence that a customer would nod along to, you are not ready.
2. Does usage vary enough across customers to matter?
If every customer consumes roughly the same amount, metering adds billing complexity without pricing power. Usage-based pricing pays off when a power user consumes 100x what a casual user does. That spread is where flat pricing either overcharges small accounts (killing adoption) or undercharges large ones (killing margins).
3. Can you live with less predictable revenue?
This is the trade nobody markets to you. Seat subscriptions produce smooth, forecastable cash flow. Pure consumption revenue moves with your customers' seasons, budgets, and usage patterns. Most small businesses solve this with a hybrid: a platform fee or base subscription that covers your fixed costs, plus metered overage that captures upside. Hybrid is now the dominant pattern for growth-stage SaaS for exactly this reason — predictability for you, fairness for the customer.
If your answers are yes, yes, and "yes, with a base fee," usage-based pricing deserves a serious look. If any answer is no, improve your tiering instead and revisit in a year.
The 5 Mistakes That Sink Usage-Based Launches
Billing vendors sell the upside. The failure modes are yours to manage.
Mistake 1: Picking a metric customers can't predict or control
Nothing destroys trust faster than an invoice nobody saw coming. A metric tied to your infrastructure costs (say, raw compute events) rather than customer value produces bills that feel random. Worse, if customers cannot reduce the metric by changing their behavior, every invoice feels like a tax. Choose metrics buyers can forecast from their own dashboards, and publish a pricing calculator before you launch.
Mistake 2: No guardrails against bill shock
Even a fair metric produces outliers — a runaway script, a viral week, a misconfigured integration. Launches without spending caps, anomaly alerts, and hard ceilings generate the horror stories that end up as chargebacks and churned logos. Set default caps for new accounts, send threshold alerts at 50%, 80%, and 100% of typical spend, and require explicit opt-in to exceed them. Guardrails cost you a little revenue; bill shock costs you the customer.
Mistake 3: Invoices nobody can reconcile
When customers cannot see what they consumed, support tickets multiply and renewals stall. Every usage invoice should break down consumption by metric, period, and rate — line items a customer can tie back to their own usage dashboard. This is also where finance teams quietly drown: a 2026 survey of more than 350 software leaders found that 63% lack full confidence in their billing operations. If your invoice lines do not reconcile to your metering data to the penny, fix that before you scale the model.
Mistake 4: Treating revenue recognition as an afterthought
Variable consideration changes how and when you can recognize revenue. Under ASC 606, usage-based fees are generally recognized as the usage occurs — which sounds simple until you layer on minimum commits, prepaid credits, tiered rates, and overage true-ups, each with its own recognition pattern. Define your revenue recognition policy for every pricing component before launch, not during your first audit. Classify and track usage revenue separately from subscription revenue in your books from day one; blending them into one "SaaS revenue" line makes NRR, margin, and forecast analysis nearly impossible later.
Mistake 5: Running metered billing on spreadsheets and hope
Legacy billing systems assume static quantity times static price, so teams bolt on custom middleware, hand-built rating scripts, and month-end spreadsheet true-ups. That stack breaks exactly when volume spikes — the moment you can least afford billing errors. You do not need enterprise infrastructure on day one, but you do need automated metering, rating, and reconciliation with an audit trail. If your month-end close involves exporting CSVs and eyeballing totals, your billing stack is already the bottleneck.
A Practical Playbook for Adding Usage to Your Pricing
Here is a sequence that works for small subscription businesses, in order:
Start hybrid, not pure usage. Keep your existing subscription as the base tier and add one metered component on top — overage units, API calls beyond an included quota, or per-outcome fees. This bounds the change for existing customers and gives you real consumption data to price against.
Instrument metering before you charge for it. Run your usage tracking in shadow mode for one to two billing cycles: meter everything, show customers their would-be usage on a dashboard, and invoice nothing. You will discover metering bugs, surprising consumption patterns, and the right unit economics before any dollar depends on them.
Backtest the price against history. Orb's headline feature — replaying historical usage through a proposed price — is a discipline you can copy with any tooling. Take three months of real usage data, apply your candidate rates, and check the distribution: what would your top 10 customers have paid versus today? If anyone's bill doubles, your rate is wrong or your grand-fathering plan is missing.
Grandfather generously and communicate early. Give existing customers 60 to 90 days' notice, honor current pricing through their contract term, and offer a choice between old and new plans where feasible. Pricing migrations fail on surprise, not on math.
Tighten dunning for variable invoices. A $49 subscription retry is routine; a $4,900 usage invoice that fails needs a different playbook — partitioned retries, proactive outreach before the retry, and backup payment methods on file. This is precisely the billing-meets-payments loop Adyen is buying: use payment signals (risk scores, decline codes, card-expiry dates) to drive collection behavior instead of blindly retrying every failed charge the same way.
Keep your books usage-aware. Record metered revenue in its own accounts, reconcile processor payouts gross-to-net every cycle (usage invoices dispute at higher rates, so fee and refund tracking matters more), and keep the raw usage-to-invoice mapping exportable. When a customer disputes a $12,000 overage line or your accountant asks how deferred credits moved this quarter, the answer should be a report, not an archaeology project. Clean, granular records are also what make usage revenue forecastable — and forecastable revenue is what lenders and acquirers pay for.
Keep Your Billing Records Audit-Ready
As you experiment with hybrid and usage-based pricing, the unglamorous work — separate revenue accounts per pricing component, gross-to-net payout reconciliation, and an exportable trail from metered event to invoice line — is what keeps billing disputes, tax filings, and due diligence from turning into fire drills. Beancount.io provides plain-text accounting that gives you complete transparency and control over your financial data — no black boxes, no vendor lock-in. Get started for free and see why developers and finance professionals are switching to plain-text accounting.
