Credits Are Camouflage

If I need a spreadsheet to understand what your software costs, your pricing isn't sophisticated. It's camouflage.

Here's the game as it's played right now. You want to enrich a contact. That's 2 credits. But if the record needs a phone number, that's 8 credits. Oh, and you turned on the AI summary, so that's an AI credit — priced differently from a regular credit, naturally. And the data came from a third party, so there's a provider credit stacked on top. Four different currencies to answer one question: what did that contact cost me?

Nobody can answer it in the room. That's not an accident.

We solved a problem and bought a worse one

Seat-based pricing had real flaws. You paid for licenses people didn't use, procurement fought over true-ups, and the vendor's incentive was to sell you shelfware. Fine. Legitimate complaints.

So the industry moved to consumption. And somewhere in that move, "pay for what you use" quietly became "pay in a currency only we control, at an exchange rate we set, that we can change every quarter."

A seat was blunt, but I could budget it. I knew what 40 seats cost next year. Tell me what 40,000 credits cost next year when the credit-per-action table gets "optimized" in Q2 and the enrichment I do a thousand times a week silently reprices from 2 credits to 5. You can't. Neither can I. That uncertainty isn't a rounding error — it's the whole product of the pricing model.

Credits are a float, and you're short the currency

Think about what a credit actually is. It's a token you prepay for, at a bulk "discount," that expires, whose purchasing power the vendor adjusts unilaterally. You are holding a currency you can't hedge, can't audit, and can't spend anywhere else.

Every design choice runs one direction:

• Bundling so you can't unbundle the expensive action from the cheap one.
• Expiry so unused balance is pure margin.
• Reprice-by-action so your effective cost drifts up while the sticker "$X per credit" never moves.
• Multiple credit types so no single number ever describes your spend.

None of that helps you forecast. All of it helps them.

"But usage is genuinely variable"

Here's the honest objection, and I'll take it head-on: consumption pricing exists because value is variable now. An AI action that drafts a sequence costs the vendor real compute. Charging every customer a flat seat for wildly different usage is its own kind of lie. That's true.

But variable cost to the vendor is not an argument for opaque cost to the buyer. Electricity is consumption-priced and I can still tell you what a kilowatt-hour costs and roughly what my bill will be. Cloud is consumption-priced and, painful as it is, AWS will hand me a dollar figure per GB. The problem was never that GTM software went consumption. The problem is that it went consumption and refused to denominate it in dollars I can plan around.

The whole category is mid-transition — credits, pure consumption, hybrid, outcome-based, all of them clawing for dominance at once. That flux is exactly why buyers are getting rolled. When the model itself is unsettled, "it's complicated" becomes the vendor's favorite sentence. Complexity is where margin hides.

So do this, on the next call

Before you sign anything, make the vendor convert their pricing into a number your CFO already thinks in. Not credits. Dollars, over the unit that actually matters to your motion:

• $ per account enriched or worked
• $ per contact
• $ per meeting booked
• $ per opportunity created

Take your real annual volume, hand it to them, and make them do the credit math in front of you. Then write the resulting dollar-per-unit number into the contract with a cap on how far it can move in a year. If they can produce that number cleanly, great — you've got a vendor who understands their own economics. If they stall, deflect, or tell you it "depends on too many factors," you've learned something more valuable than a price.

You've learned the complexity is the point. And you should never buy the thing whose cost the seller can't state out loud.

— C.B.