Per-Token, Per-Seat, Per-Outcome: Why Your Next Pricing Model Change Will Break Your Spreadsheet

Jamin Ball wrote something in a recent Clouded Judgement post on per-token pricing that's worth sitting with: "We're entering a world where your pricing model is your business model."

He was writing about the shift from GPU hours to token dollars — how AI inference costs are making the gap between your pricing model and your cost structure existential. Price too low and you're paying customers to use your product. Price too high and you lose to whoever figured it out first.

The whole piece is worth reading if you're thinking about AI pricing strategy. But the takeaway that matters most for pricing operations teams isn't about tokens specifically. It's about velocity. The speed at which SaaS companies need to change their pricing models has fundamentally accelerated — and almost nobody's operational infrastructure is ready for it.

The old world: set it and forget it

In traditional SaaS, pricing changes happened maybe once a year. Maybe twice if you were aggressive. The process looked something like this: leadership spends a quarter debating the new model, someone updates the rate card in a spreadsheet, sales ops sends out a "new pricing effective January 1" email, and reps start using the new numbers. The whole cycle from decision to execution took weeks or months, and that was fine because pricing was forgiving. With 80%+ gross margins, an imperfect pricing model was a missed optimization, not a crisis.

That era is ending.

The new world: pricing as a living system

Three forces are converging to make pricing changes more frequent, more complex, and higher-stakes:

AI inference costs create variable COGS. When every "magic" feature in a product triggers real compute costs on the backend, pricing has to account for consumption in a way that per-seat pricing never did. Ball's point about credit-based pricing models is exactly right — credits become an abstraction layer that lets companies manage token economics internally while presenting a clean pricing interface to customers. But somebody has to manage that abstraction. Somebody has to model what a credit is worth, how it maps to different product features, and how the math changes when the inference provider drops prices by 40% overnight.

Hybrid models are the norm, not the exception. Pure per-seat pricing is giving way to combinations: a platform fee plus per-seat charges plus usage-based components plus add-on modules. Each of those dimensions has its own pricing logic, its own tier breaks, its own discount rules. A company selling three products with tiered pricing across two dimensions each has more pricing permutations than most spreadsheets can sanely manage.

Competitive pressure demands faster iteration. When a competitor ships a new pricing model in a quarter, spending six months modeling a response isn't viable. The companies that Ball says will "capture way more value per customer than was ever possible in the old SaaS world" are the ones who can iterate on pricing quickly — testing new tiers, adjusting credit values, launching promotional structures, and rolling back what doesn't work.

What actually breaks

The pain pricing managers and RevOps leaders describe isn't usually "we can't decide on a pricing model." It's "we decided, and now we can't execute the change." Here's what that looks like in practice:

The spreadsheet versioning nightmare. The master rate card gets updated. But there are three other versions floating around: the one the East Coast team uses, the one that got copy-pasted into the sales enablement deck, and the one a top AE modified six months ago and never told anyone about. Which one is truth? Nobody's sure, and the new pricing model hasn't even rolled out yet. (More on this pattern in our post on signs you've outgrown spreadsheet pricing.)

Discount logic that doesn't scale. The old model had simple volume discounts: 10+ seats get 10% off, 50+ get 15%. The new model has tiered usage pricing with graduated rates, a platform fee that doesn't discount, per-seat charges that do, and a discretionary discount band that varies by deal size. Try encoding that in a spreadsheet formula. Now try explaining it to a new rep in their first week.

The "what should I charge?" lag. A rep is on a call. The prospect asks about pricing for a custom configuration. Under the old model, the rep could look at the rate card and do the math in their head. Under the new model with tiered components and bundled credits, the rep says "let me get back to you" — and the deal loses momentum while someone in RevOps reverse-engineers the right number from a spreadsheet.

Stale quotes in the pipeline. Pricing changes on April 1. There are 30 open deals in the pipeline with quotes generated under the old model. Which quotes should be honored? Which need to be re-generated? Who's tracking this? In a spreadsheet world, the answer is usually "nobody until something goes wrong."

The operational gap nobody talks about

There's a strange disconnect in SaaS right now. Companies are investing heavily in pricing strategy — hiring pricing managers, engaging consultants, running conjoint analyses, modeling willingness-to-pay curves. The strategy side has never been more sophisticated.

But the operational side — the actual infrastructure for managing and executing pricing changes — hasn't evolved in a decade. It's still spreadsheets. It's still emailing updated rate cards. It's still reps looking up prices in a shared Google Sheet and doing discount math in their heads.

Strategy without operational infrastructure is just a good idea that doesn't ship. A company can design the perfect hybrid pricing model with usage tiers and credit-based AI features. But if there's no system that holds that model as structured data, enforces discount rules at the point of sale, and gives reps real-time access to accurate pricing, the model lives in a deck somewhere and reps keep quoting from the old spreadsheet.

This gap is where revenue leakage lives. Industry research consistently finds that B2B companies lose 1–5% of revenue to pricing execution failures — wrong discounts applied, stale rate cards used, manual calculation errors, pricing changes that didn't propagate. At a $50M ARR company, that's $500K–$2.5M annually. Not from bad strategy. From bad infrastructure.

What "pricing infrastructure" actually means

The fix isn't another spreadsheet with better formulas. It's a pricing system of record — a single source of truth that holds pricing models as structured, queryable data. Practically, that means:

Price books, not rate cards. A structured representation of products, pricing dimensions, tiers, and rules that a system can compute against — not a flat table a human reads and interprets.

Discount governance, not discount guidelines. Rules that are enforced at the point of quote generation — term discounts, volume breaks, multi-product incentives, discretionary bands with role-based limits — not a policy doc that reps are supposed to follow.

Real-time pricing, not "let me check." When a rep adjusts quantities or changes a configuration, the price updates instantly and correctly. No manual recalculation, no "I'll send you a revised quote tomorrow."

Pricing changes as deployments, not announcements. When leadership approves a new pricing model, it's updated in one place and immediately reflected everywhere — in every quote, for every rep, with proper handling of in-flight deals.

This isn't a new category. It's infrastructure that should have existed before anyone bought a CPQ, engaged a pricing consultant, or built a spreadsheet with nested IF statements. It's the foundation that makes pricing strategy executable — and the foundation that AI agents will need when they start generating quotes autonomously.

The Clouded Judgement connection

Ball's observation that "the companies who figure out pricing and packaging the fastest will have a big edge" is about AI-native companies, but the principle applies more broadly. In a world where pricing models change quarterly instead of annually, where hybrid structures replace simple per-seat pricing, and where AI features introduce variable costs that shift with every model generation — the companies with operational pricing infrastructure will iterate faster than the ones still managing rate cards in spreadsheets.

The token economy is making this urgent for AI-native companies right now. But every B2B SaaS company with more than one pricing dimension is heading toward the same operational complexity. The question isn't whether the pricing model will change. It's whether the infrastructure exists to change it without breaking everything.

---

Valmetric is price book management and quoting for B2B SaaS teams who outgrew spreadsheets. Start a free trial →