Usage-based pricing is not a tariff model, it is a bet on the correlation between consumption and perceived value. It works when value grows with volume and the billed unit is legible. It backfires when usage is a cost to the customer, or when your company needs predictability. The most solid models are hybrid: a subscription base, a usage component.
Usage-based pricing has become the default pricing reflex of the last few years. “You pay for what you consume” sounds fair, modern, aligned with value. Yet we see too many companies switch to it by imitation, without asking the only question that matters: does your value grow with usage?
This is not a question of fashion, it is a question of economic model. The same pricing mechanism accelerates one company and suffocates another, depending on what the customer comes for.
What usage-based does well
When it fits, this model is a powerful accelerator. It lowers the barrier to entry, lets the customer start small, and grows the invoice at the pace of the value they get. It also changes the commercial relationship: instead of renegotiating a flat fee every year, account growth happens on its own.
It excels in three specific situations.
- Value is proportional to volume. The more the customer consumes, the more they gain. Price follows naturally, and the increase is not up for debate.
- Adoption is progressive. A low entry ticket defuses perceived risk, shortens the decision, and opens the door to accounts that would have refused an annual commitment.
- The usage unit is legible. The customer understands what they are paying for and can anticipate it, which avoids the single most destructive conversation there is: the unexpected invoice.
Where it turns against you
The same model, misapplied, sabotages growth instead of carrying it. The symptoms always appear in the same order.
- Revenue becomes unpredictable. You no longer steer your numbers, you discover them at month end. So do your investors, and that is often how the question surfaces.
- The customer optimises against you. If usage is a cost rather than a gain for them, they will learn to reduce it. You have designed a model that encourages them to consume less of what you sell.
- The unit of value is illegible. Billing per technical call or per stored volume, when the customer thinks in business outcomes, creates anxiety at every statement and turns every renewal into a negotiation.
Usage-based pricing is not a model: it is a bet on the correlation between consumption and value. If that bet is wrong in your case, no pricing table will rescue it.
The three-question test
Before any switch, three questions settle the decision better than a competitor comparison.
Does your customer gain something by consuming more? If the answer is yes, usage aligns your interests. If consuming more means enduring more for them, you have just created a permanent conflict between their interest and yours.
Can the customer predict next month’s invoice? If they cannot, they will over-provision, cap their usage, or ask for a flat fee. All three reactions degrade your revenue.
Can your company absorb variable revenue? A company that has to present a trajectory to a board or to investors needs a floor. The question is not ideological: it is about the share of revenue you can let float without losing control of your steering.
Choosing the billing unit
This is the most structuring decision of the model, and it is too often taken for technical reasons, because the data happened to be available in the product. The right unit has three properties: the customer understands it without explanation, it rises when they get more results, and it is measurable without dispute.
A unit that meets all three is rare, and it is found on the customer’s business side, not on your architecture side. It also shapes how you will stack your offers, which follows from the segment you serve, as we set out in positioning: say who you serve. It also changes how you sell: it gives reps a shared language with the buyer, and it simplifies proving value at renewal.
That choice is not made in isolation: it follows from the segment you serve and the promise you carry, which makes it a go-to-market decision as much as a pricing one. We place the trade-off in context in building a go-to-market plan.
The units that hold, depending on what you sell
Four families of units come up, and each has its terrain.
The outcome unit bills what the customer gets: a case processed, an application received, an order placed. It is the sturdiest, because the invoice rises when the customer succeeds, and nobody argues with a cost that follows a gain. It does require the outcome to be attributable to your product without ambiguity, otherwise you will spend every renewal proving causation.
The business volume unit bills the material the customer entrusts to you: contracts managed, employees tracked, catalogue references. It is legible, predictable, and it grows with the customer’s own business without renegotiation. Its weakness is being decoupled from real usage: a customer who hands you a lot of volume and never opens the tool pays a lot for nothing, and leaves at renewal.
The active seat unit bills the people who actually use the product. It is understood instantly. It also has a serious flaw: it punishes the spread of usage inside the company, which is exactly what you want to encourage.
The technical unit bills what your infrastructure consumes: API calls, gigabytes, compute minutes. It is fair from your side and illegible from theirs. Keep it for products whose buyer is technical, and already thinks in orders of magnitude.
The practical test fits in one sentence: show the candidate unit to an existing customer and ask them to predict next month’s invoice. If they hesitate, it is the wrong unit, whatever its economic merit.
The guardrails that make usage bearable
A usage model without guardrails exposes both sides. Four mechanisms show up in the contracts that hold.
The minimum commitment sets an annual floor the customer agrees to consume. It hands you back part of your predictability and gives the customer a price break in return. It is the most common mechanism, and the easiest to defend in negotiation.
Prepaid credits charge an envelope up front, drawn down over time. They improve your cash position and remove the monthly surprise. Watch the expiry date: credits that lapse create lasting resentment for a short-term accounting gain.
The cap bounds the customer’s invoice over a period. It costs revenue in peak months, and it buys something rarer: the sign-off of a finance director who will not approve an open-ended spend.
The consumption alert warns the customer as they approach a threshold. It is not a contract clause, it is the product feature that prevents the most cancellations. An invoice that doubles without warning is paid for in trust, not in euros.
What usage does to your metrics
Moving to usage shifts a problem from the customer to your own dashboards, and nobody sees it coming.
Your annual recurring revenue stops being a fact and becomes a projection. Two companies reporting the same figure are no longer talking about the same thing: one has signed contracts, the other an extrapolation of past consumption. If you are raising or reporting to a board, settle in advance what you call recurring, and hold the same definition quarter after quarter.
Sales forecasting changes nature too. It no longer covers only the deals to be signed but the consumption of accounts already signed, which calls for a different method and different signals. Above all it assumes consumption is legible in your systems, which points back to building your revenue operations foundation.
In exchange, one indicator becomes far more legible: net revenue retention. In a usage model, expansion shows up immediately, without waiting for an annual renegotiation. It is the best early warning you will have on the real health of your base.
The real debate is not usage versus subscription
The question is not binary. The most solid models we design are often hybrid: a subscription base guaranteeing predictability, and a usage component capturing growth when the customer succeeds. You secure the floor and share the upside.
Sizing that floor is the delicate part. Too high, and it cancels the easy entry that justified moving to usage in the first place. Too low, and it secures nothing, leaving you with the unpredictability and none of the benefits. We usually set it against the fixed cost of serving an account, which gives a defensible floor in negotiation.
The choice turns on your reality, not on the trend: what is your unit of value, how do your best customers actually consume, and how much predictability does your company need to hold its trajectory?
Switching without breaking existing revenue
A pricing model change touches the installed base, and that is where most migrations go wrong. Three precautions limit the damage.
Test on a cohort of new customers before touching existing accounts. In one quarter you will have real consumption data, which beats any model.
Leave the installed base on its current model for as long as necessary, and migrate at renewal, with a clear compensation. A forced migration mid-contract produces departures the new grid will not offset for two years.
Finally, measure the effect on perceived value before measuring the effect on revenue. A pricing model that improves the first quarter while degrading the relationship gets paid for at the next renewal.
The Arkimoz method
We start by measuring how value is actually created among your existing customers, not how we would like it to be created. From there we derive the billing unit that aligns your interests with theirs, then model the impact on revenue predictability before any switch. Finally we test on a cohort before generalising.
The right pricing model is not the most fashionable one. It is the one where the more your customer gains, the more you gain. Never the reverse.
Frequently asked questions
- What is usage-based pricing?
- It is a model where the customer pays according to what they consume (volume processed, transactions, active users, technical calls) rather than a fixed amount per period. It sits opposite flat subscription, and moves closer to it as soon as you add a minimum commitment or tiers.
- When is usage-based pricing a bad choice?
- In three cases. When usage is a cost rather than a gain for the customer: they will learn to reduce it, and therefore to pay you less. When the billed unit is technical while the customer thinks in business outcomes: every invoice becomes a source of anxiety. And when your company needs to announce predictable revenue, which this model alone does not allow.
- How do you choose the billing unit?
- By observing how value is actually created among the customers who succeed, not how you wish it were created. The right unit has three properties: the customer understands it without explanation, it rises when they get more results, and it is measurable without dispute.
- Which billing unit should you choose?
- Four families come up: the outcome unit (a case processed, an order placed), the business volume unit (contracts managed, employees tracked), the active seat unit, and the technical unit (API calls, gigabytes). The first is the sturdiest when the outcome is attributable to you without ambiguity. The last only suits technical buyers. The practical test: ask an existing customer to predict next month's invoice with the candidate unit. If they hesitate, it is the wrong one.
- How do you avoid billing surprises for the customer?
- Through four mechanisms that combine: a minimum commitment that sets a floor, prepaid credits that remove the monthly surprise, a cap that bounds the invoice over a period, and a consumption alert as the customer approaches a threshold. The alert is not a contract clause but a product feature, and it is the one that prevents the most cancellations.
- Can you combine subscription and usage?
- That is the most common setup among companies that hold their trajectory: a subscription base covering fixed costs and providing predictability, plus a variable share capturing growth in accounts that succeed. The delicate point is sizing the base: too high and it cancels the easy entry, too low and it secures nothing.