For Starters #59: On Credit-based Pricing

Or how to confuse and discourage the most enthusiastic customers

I came up in an age where phone service, both landline and cellular, was sold by the minute - and every minute cost something different depending on where it fell in the day and which day it was. Despite nights and weekends being cheaper they were still limited. The end of the month inevitably meant bracing for just how egregious the overage charges were (again).

For a short time in the late 1990s, I lived in Germany where Deutsche Telekom sold not minutes, but an abstractions of minutes called Blips or Units or Zeitteils or Punkts. It’s been a few years and my German’s pretty rusty. Whatever the branding, the number of Zeitteils was fixed during the day - however the number of minutes each Zeitteil contained fluctuated throughout the day and week. Almost if Swatch Time beat slower at work and faster at the club.

Both examples resulted in the some negative consumer outcome; rationed usage and being surprised by the bill.

The Deutsche Telekom example also highlights a key supplier feature of a credit-based pricing system, the credits can be defined however the suppliers want. Sure, AT&T made minutes fluctuate in value, but a minute is a known thing. Buyers have no reference for the value of a fluctuating Zeitteil.

B2B buyers have a range, budgeted or not, they’re willing to spend on any given solution/JobToBeDone/etc. It doesn’t really matter how the price is constructed it just needs to fit in within the range. Of course, the amount changes based on the size and strategic goals of the buyer’s organization, though in general the larger the annual revenues, the larger the range for just about everything, because that’s how percentages work. I talk more about this in For Starters #57.

The best way to construct the price is against a value metric meaningful to the buyer as their business grows.

For example, a project management tool might sell a bucket of concurrent projects per year, a tool for real estate brokerages might price against estimated annual transactions, a financial management tool might price against assets under management. In all cases there’s a maximum that’s unlikely to be met without an outsized change to the customer’s organization. The annual license fee fits the business and its growth goals for the year. Win-win.

The second worst way to construct the price is user-based or seat-based.

Employee count is simply a poor proxy for business performance and user-based pricing make employees incrementally more expensive by equating access and value. Enterprise buyers hedge user-based entitlements today by rationing access across their organization through tools like SailPoint and SolarWinds. User-based pricing itself is an odd relic of on-premise software era where we still needed to manually enter license codes as we installed software machine by machine and the software supplier couldn’t tell how the software was used. (BTW: Helping clients transition from user-based pricing to the above customer value-aligned metric is a key focus of my work.)

The worst way to construct pricing is credit-based.

In situations where; it’s difficult to quantify customer value, a small number of users across segments are driving excessive use, and meaningful costs are incurred with every use - credit-based systems are pretty tempting. At best, they’re a transition until a customer value-based pricing, packaging, and segmentation can be designed. However until then, they’re terribly anti-customer. Requiring customers to make a value assessment every time they engage. Thus discouraging innovation, experimentation, and identification of wider organizational value.

“The main roadblock to faster adoption has been getting buyers to accept it. Credit-based pricing can be hard to predict, hard to manage, and feel like a black-box. And nothing scares procurement more than runaway costs with no visibility....each company defines a “credit” differently, which can make it extremely frustrating for folks to figure out what exactly they’re buying.” -

Kyle Poyar

, Growth Unhinged

All of that, plus users lack both understanding and control over how those credits are spent during usage. Can the admin limit all requests to a 500 credit maximum? Will that limit simply chew up more credits while the LLM figures out how to constrain itself? Or is the LLM free to respond however it wants…INSERT COIN TO CONTINUE.

From OpenAI’s Help page on the subject

“Business users receive per-seat limits for all advanced features. If a user exceeds their limit and the workspace has purchased credits, they can continue drawing from the shared pool. Purchasing credit packs is optional and only necessary if a Workspace Owner wants to unblock users who exceed their usage limits.”

If I’m understanding correctly, a single user can lock out their entire organization’s access to advanced features by asking GPT-5 to think too much.

Strange game Professor Faulken. Strange game.

The most understandable credit system I’m aware of in SaaS is SaasyDB’s “One credit unlocks every employee at a company.” The value assessment of “is this company worth a credit“ seems like a meaningfully low bar in planning prospecting. Plus half of SaasyDB’s plans are unlimited credits. Calling back to For Starters #52 - Price is Mostly Packaging. For some customers, not being bothered managing credit limits is worth a bit more. Which suggests suppliers of generative AI haven’t yet penciled out what non-credit pricing level looks like or they have….and it’s astronomically outside everyone’s range.