What You Charge Per: The Pricing Decision You Can't Undo
The price you set is easy to change. What you charge per is not.
Summary
- What you charge per decides whether revenue can grow inside an account without a renegotiation, whether margin survives success, and whether the customer can predict their own bill.
- The metrics that track value best are the ones customers find hardest to forecast. The usual resolution is a minimum commitment plus usage above it.
- Per-seat pricing is the most common mistake in Indian SaaS, and it is actively self-defeating for a product whose promise is doing more work with fewer people.
Most pricing conversations are about the number. How much should we charge?
There's a bigger question sitting underneath it that hardly anyone asks: what are we charging per?
Per seat. Per transaction. Per tonne. Per site. Per GB. Per patient. Per order processed.
That choice gets made casually, usually in your first year, usually by copying whatever a competitor does. And it turns out to be one of the few pricing decisions that's genuinely expensive to reverse, because changing it means rebuilding your contracts, your billing, your product tracking and your sales incentives all at once.
Getting the price level wrong is survivable. Getting this wrong quietly caps how big your accounts can become.
Why it decides more than the number
Three things follow from what you charge per.
Whether your revenue can grow without a renegotiation. If the thing you charge for grows naturally as your customer gets more value, your revenue grows with them and nobody has to have a conversation about it. If it doesn't, every rupee of growth needs a new negotiation, which means most of it never happens.
Whether your margin survives success. If your costs go up with usage but your price doesn't, your heaviest users are your worst customers. That's a problem that gets worse the better you do, and it shows up first in your contribution margin.
Whether the customer can predict their own bill. This matters far more in Indian enterprise than most founders expect. A finance team that can't forecast what they'll owe you next quarter will push back hard, and procurement will treat unpredictability as a risk to be priced.
Five tests
Run any candidate metric through these.
Does it track the value they get? As the customer gets more out of your product, do they pay more? If the answer is no, you'll watch accounts grow enormously without your revenue moving.
Can they predict it? Can their finance team put a number in next year's budget with confidence?
Does it grow on its own? Without you selling anything new.
Can they game it? Sharing logins, batching transactions, routing around the meter. If they can, some of them will.
Can both sides check it? They have to be able to verify the invoice, or every renewal turns into an .
Almost no metric passes all five. The job isn't finding a perfect one, it's choosing your trade-off deliberately instead of by accident.
The honest comparison
Seven value metrics against the five tests
| What you charge per | Tracks value? | Predictable? | Grows on its own? | Can be gamed? |
|---|---|---|---|---|
| Flat fee | No | Very | No | No |
| Per seat | Weakly | Yes | Sometimes | Yes, shared logins |
| Per active user | Fairly | Somewhat | Yes | Harder |
| Per transaction | Strongly | Less so | Strongly | No |
| Share of value (GMV, volume, recovery) | Strongest | Least | Strongest | No |
| Per unit of compute or storage | Weakly, it tracks your cost | No | Yes | No |
| Per outcome | Strongest | Least | Strongly | Arguments over credit |
Notice the pattern. The metrics that track value best are the ones customers find hardest to predict. That tension is the real design problem, and the usual resolution is a minimum commitment plus usage above it: the customer gets a floor they can budget, you get growth when they grow.
The per-seat trap
This deserves its own section, because it's the most common mistake we see in Indian SaaS.
Per-seat pricing feels safe. It's easy to explain, easy to forecast, and everyone else does it.
But think about what it means. You charge based on how many people at your customer's company use the software. If their headcount is flat, and in a lot of Indian businesses it is, deliberately, your revenue from them is flat. Forever. However much more value they're getting. However much more they're doing with the product.
You've put a ceiling on your own growth, at the architecture level, and no amount of customer success work will lift it. It is also the ceiling that caps net revenue retention.
It's worse for AI-heavy products. The whole promise of those products is that the software does the work instead of a person. If you sell something that lets a customer do more with fewer people, and you charge per person, you are charging less precisely when you deliver more. Your best outcome is your worst commercial result.
When your cost and your price disagree
Check one thing, and check it per customer.
Does what you charge for move in the same direction as what it costs you to serve them?
Where cost and price pull in different directions
| Business | What drives your cost | What people usually charge for | Do they match? |
|---|---|---|---|
| Classic software | Seats, light compute | Per seat | Yes |
| Data-heavy software | Storage, queries | Per seat | No |
| AI-native product | Model usage | Per seat, or flat | No |
| Marketplace | Transactions, support, fraud | Share of value | Mostly |
| Logistics tech | Shipments | Per shipment | Yes |
| Services | Delivery hours | Fixed project fee | No, unless scope is controlled |
Where the answer is no, you have three choices. Change what you charge per. Put a fair-use cap in with a rate beyond it. Or accept the shape knowingly and price the average high enough to carry the heavy users.
All three are fine. Not choosing is not.
How to change it without losing your base
If yours is wrong, here's the sequence that works.
New customers only, first. Run the new metric on new logos for two or three quarters. Prove it works before you touch anyone existing.
Build a calculator. Any existing customer should be able to see, in one screen, what their bill becomes under the new model. Surprise is what causes , not the change itself.
Grandfather with an end date. Twelve to eighteen months, stated clearly. Open-ended grandfathering creates a permanent group of customers on terms nobody can explain, and it never gets fixed.
Move people at renewal, never mid-contract.
Give a real reason to move early. Extra features, a price lock, a longer term at a better rate.
Expect to lose some. Five to fifteen percent of the affected base is normal. Model it before you start rather than discovering it in month three.
The important point: this is much cheaper at 60 customers than at 2,000. If you think your metric is wrong, the cheapest day to change it is today.
By business model
SaaS. Move away from pure seats if your value doesn't scale with headcount. Most companies land on a base plan plus a usage element.
AI-native. Your cost moves with usage, so your price must too. Platform fee plus usage above an included allowance is the usual answer.
Marketplace. Your metric is the take rate, and the design questions are who pays, how much, and whether it tapers as a seller grows.
Services. Your metric is effectively hours, which is why moving to deliverables or a retainer is a change of metric, not just a change of format.
Manufacturing. Per unit is natural. The question is whether you're pricing by weight, by piece or by the value the part delivers in your customer's product, and the third of those is usually worth far more.
Where to start
Answer two questions about your largest customer.
If their business doubled next year, would what you charge them double? And would what it costs you to serve them double?
If the answers don't match, you've found the thing that's going to limit you, and you've found it while it's still cheap to fix. Where this sits against everything else is set out in the five layers of pricing.
Our Pricing Maturity Assessment covers this alongside the other layers, so you can see whether your metric is your binding problem or whether something else is costing you more right now.
Frequently asked questions
What is a value metric in pricing?
It's the unit you charge per: a seat, a transaction, a tonne, a gigabyte. It matters more than the price itself, because it decides whether your revenue grows as your customer grows and whether your margin holds as they use more.
Is per-seat pricing bad?
Not always. It works when the value your customer gets genuinely scales with how many people use the product. It caps you when it doesn't, and for AI products that replace work rather than assist people, it can actively work against you.
How do I change my pricing model without losing customers?
Run it on new customers first, build a calculator so existing customers can see their own number, grandfather with a stated end date, move people at renewal rather than mid-contract, and give a genuine incentive to switch early. Expect to lose a small share, and model that before you start.
How useful was this article?
One tap. It tells us what to write more of.
The next one
Get what we publish next, by email.
Working notes on raising, borrowing, protecting, growing and structuring capital in India. One email a week at most, and you can leave any time.
We use your address only to send this. See our privacy policy.
We store your address to send you these emails and nothing else. See our privacy policy.
Related reading
