Cost-Plus Pricing for Software: The Pass-Through Trap

TL;DR Cost-plus pricing for software attaches a markup to the cost of delivering a unit. Near-zero reproduction cost made the question look settled for two decades, and generative AI inference reopened it by putting a real per-use cost under the product. Passing a supplier’s token rate through with a margin on top is cost-plus, whatever the planning deck calls it: the unit came from the supplier and measures what the supplier sells, and every efficiency gain flows to the buyer as a lower price. Cost belongs in the architecture as the floor beneath the price; above the floor, the value metric decides the unit. One test settles which you are running: if your cost to serve halved next quarter, would your price move?

What Cost-Plus Pricing Means Once the Product Is Software

Cost-plus pricing attaches a markup to the cost of producing a unit. The markup is a ratio applied to cost; the margin is the resulting share of revenue. A target margin and a target markup are different numbers, and confusing them produces the wrong price: a 50% markup on a one-dollar cost gives a $1.50 price and a 33% margin.

What is an example of cost-plus pricing in software?

A vendor builds a document summarization tool on top of a third-party LLM. Each summarization request costs the vendor a fraction of a cent in inference fees. The team prices per document, sets a target gross margin, backs into a per-document price, and publishes a rate card. That is cost-plus pricing. The unit came from the supplier. The price derives from the cost. The customer buys summaries; the vendor prices tokens.

Traditional software costs almost nothing to serve one more user. With nothing under the markup, cost-plus had nothing to do, so most teams stopped asking whether their pricing is cost-derived.

What counts as cost in a software company

Infrastructure is the cost category most teams count. Support burden, implementation effort, and allocated development cost behave differently, and most cost models exclude them, so the floor understates what serving the customer requires. Discount approvals calibrated to that floor let deals close below what the account costs to serve.

Generative AI Put a Real Cost Back Under the Product

When a product depends on LLM inference, the vendor pays a per-use fee to a supplier, denominated in tokens, at a rate the supplier sets. That is a genuine marginal cost per unit of product output.

Why inference cost restored the condition cost-plus was built for

Classic machine learning had a training cost and a serving cost too small to force a pricing decision. Generative AI inference is different: the cost scales with use, it is visible on every invoice, and it is large enough to move gross margin at launch.

Supplier rates move, by large and unpredictable amounts; AI price cuts and deflation capture owns how model price cuts reach pricing decisions. Pin your price to a supplier’s rate and your pricing decision becomes a function of your supplier’s.

What changes when your marginal cost is somebody else’s price

A supplier’s rate is set to price the component the supplier sells, and it measures nothing about what your customer receives from your product. A token is not a business outcome. An inference call is not a job completed.

Passing a Supplier’s Cost Through Is Cost-Plus Pricing, Whatever the Deck Calls It

Generative AI cost is variable and hard to forecast, so the team protects margin: take the supplier’s unit, apply a margin, surface a price. That is cost pass-through, and cost pass-through is not value-based pricing.

The two differ in where the unit comes from. Value-based pricing emerges when licensing, packaging, and pricing decisions align around how Customer Groups extract value. Pass-through inherits its unit from the supplier. The vendor has adopted a metric chosen to price a component, on behalf of customers who are buying an outcome.

What a pass-through pricing surface resembles from the inside

It resembles margin discipline. The target gross margin is set, the cost is known and auditable, and the price follows mechanically. None of that is wrong; it describes a well-run cost-plus operation. The problem is what the arithmetic is measuring.

Why the supplier’s unit is the wrong unit for your customer

The value metric is the singular countable unit that denominates customer value. When it arrives from a supplier’s rate card, it is answering the supplier’s question: how much inference did you consume? That question is a proxy for your customer’s question, and the proxy holds only while supplier pricing stays still and usage patterns stay predictable.

Neither condition holds for long. Rates change at the supplier’s discretion, and usage patterns shift as customers settle on the features they value.

Why cost-plus and value-based pricing are not alternatives

Cost-plus and value-based pricing occupy different positions in the architecture: value-based pricing emerges from it, and cost-plus is a way of setting the number inside a pricing model. They are different categories of decision, so you cannot pick one in place of the other.

The vendor facing a generative AI cost structure has a real decision: keep the supplier’s unit and pass the cost through, or choose a unit that measures what your layer delivers. Pass-through or recast owns that decision in full.

Value-based pricing emerges from the architecture and cannot be applied to a finished price. The ordering is fixed: a value-based number applied to a cost-derived unit does not produce value-based pricing. Pricing model and value metric are separate decisions, and the unit is the one cost-plus gets wrong.

Nearmap, an aerial imagery company, priced its product in Australia on the data a customer consumed, a flat rate per tranche of imagery, on the reasoning that every customer extracted the same value from the same volume. The unit had come from the cost side: more data pulled meant more data cost to serve. In the United States that model stalled, because a roofing contractor wanted a few parcels and an insurer wanted whole regions, and a data unit could not tell them apart. The company replaced it with a value metric that reads what each Customer Group uses, so the parcel buyer and the region buyer each pay for the value they extract. Its ASX filings show revenue rising from A$97 million in the fiscal year SPP worked with the company to A$146 million two years later, and in October 2026 reports ahead of a planned relisting put its fiscal 2026 revenue expectation at US$350 million.

Why Cost-Justified Prices Feel Fair, and What That Costs You

Cost-plus has a genuine advantage that its critics rarely concede: buyers accept it.

Peer-reviewed research on price fairness finds that buyers judge a price increase tied to a rise in the seller’s cost as legitimate. They judge a demand-driven increase as unfair. The effect explains why cost-plus persists in rooms where everyone present knows the theory.

Why buyers accept a cost-justified price increase

Peer-reviewed research on price fairness finds buyers hold an existing relationship to a stricter standard than a new one, and sales training does not move that standard.

What happens once a buyer knows your price tracks your cost

A price explained by cost teaches the buyer that your price is a function of your cost. A buyer who has learned that will negotiate your cost. In the renewals we see, procurement asks for the cost basis by name, and how B2B buyers judge a price change was settled before the meeting.

The advantage at signature becomes the constraint at every renewal. Peer-reviewed research on business markets finds a positive relationship between a firm’s ability to quantify customer value and its performance, with the benefit accruing at the firm level rather than to individual sellers. A value story has to be built into the architecture; if the architecture is cost-derived, there is none available when the buyer asks the cost question.

What are the downsides of using cost-plus pricing in software?

The standard criticism, that cost-plus ignores market demand, is correct and shallow. The structural problem is that efficiency stops paying you. A guaranteed markup and a captured share of delivered value are different measurements of different quantities. A company can hit its markup on every deal and still transfer most of what it created to the buyer, with no signal in the margin line that it happened.

We saw this across our corpus in the years after the pandemic. Vendors raised prices on existing customers and justified the increase by an outside mechanism, rising interest rates and inflation first, tariffs later. Buyers accepted the increase on those terms. When the mechanism faded, with inflation easing or a tariff withdrawn, the same buyers came back at renewal and asked for the price to return to where it had been, and they were within their rights: the vendor had tied the price to the input, and the input had moved the other way.

Who Keeps the Efficiency Gains Under Cost-Plus Pricing

When price is a function of cost, every reduction in cost to serve is on its way to the buyer, whether or not the vendor announces it.

The engineering team that halves inference spend through better caching, a smaller model, or a rewritten retrieval path has, under a cost-plus architecture, created a saving the vendor cannot keep for long. Some vendors say nothing and hold the margin for a while. But a buyer who was sold a price on its cost basis learns when that cost falls, and starts pushing the price point down from the next renewal on. The efficient company ends up pricing below the one that operates badly, with no mechanism to capture what it earned.

A supplier’s price cut travels the same road: the buyer learns the cost fell, and the vendor keeps none of the improvement.

Peer-reviewed qualitative research on business-to-business pricing finds that effective pricing depends on dedicated routines and cross-functional coordination, and that unresolved internal conflict over what list prices are for produces systematic discounting. A list price derived from cost gives a seller nothing to defend in the room. Why 90 percent of value-based pricing is a hoax owns the diagnosis of what happens when a seller cannot anchor a number to value.

One Test, and the Questions Behind It

If your cost to serve halved next quarter, would your price move?

A yes means the architecture is cost-plus, whatever the strategy deck calls it.

If the answer is yes and you want to know what your unit should be instead, talk to a pricing expert.

Where did your billing unit come from?

If the unit arrived from a supplier’s rate card, it is answering the supplier’s question. Ask whether any customer-facing output could serve instead; if no one can name one, that absence is diagnostic.

Can you explain your price to a customer without referring to your costs?

If the best explanation for your price is “this is what it costs us to deliver,” a cost-plus architecture is what you are running, and the buyer now knows it too.

If your supplier changed its rates tomorrow, which part of your architecture responds, and how quickly?

The answer names the dependency. A pricing architecture that responds automatically to supplier rate changes is pass-through by construction, whatever your capacity to absorb a change.

What does your pricebook say you realized, and does the cost story explain it?

Realized price is the price achieved on a transaction. If the spread between list and realized price is large and the cost story cannot explain it, the unit is not capturing the value the sale found. Pricebook deviation measures that spread.

Where Cost Belongs in the Architecture

Cost has a legitimate and permanent place. It establishes what the company cannot go below. It is what discount discipline is calibrated against. And it is a necessary input to every packaging decision about what a distinct version costs to operate.

SPP’s margin-calibrated discounting is a cost-aware practice, and a different practice from deriving a list price from cost: the distinction is between cost as a floor and cost as a source.

Above that floor, the value metric decides what unit the price attaches to, and the pricing decision attaches a number to the result. How to price software owns the full ordering of those decisions.

Cost-plus makes genuine sense for professional services and implementation work, where effort is real and auditable. For the software product itself, treating cost as the origin of a list price is the one move that forecloses everything that follows.

For the software monetization decisions that sit above the cost floor, start with the software monetization pillar.

If you are not sure which side of the test your architecture sits on, bring your pricebook and talk to a pricing expert. We read it against the cost floor and the value metric, and we tell you which one your price is tracking.

FAQs