Author
TL;DR When a vendor and a buyer price on outcomes, they sign a definition, and the question of who defines success holds three decisions with three owners. Naming the outcome belongs to whoever holds the licensing model, where the value metric and the billable event are written. Confirming the outcome fired is a separate decision, because a counter records a closed session and never the customer’s result. Absorbing the gap between the proxy and the result is an agreement question, rarely stated at signature and reliably tested at renewal. The market has settled it four ways, and handing the buyer the pen can protect the vendor. Four tests at the end come before the definition carries a price.
When a vendor and a buyer agree to price on outcomes, they sign a definition. They rarely agree on who owns it.
“Who defines success in outcome-based pricing” reads as a single question. It holds three separate decisions, each with a different owner, each moving money on a different schedule. Most treatments never separate them.
Three Definers Hide Inside One Word
Outcome-based pricing is a metric selection tied to the business result the customer achieves. That description sounds clean. The commercial reality is that “success” inside an outcome-priced contract carries three distinct roles, and collapsing them into one word is where disputed invoices start.
Who Names the Outcome
The naming role belongs by default to whoever holds the licensing model, within whatever limits the agreement puts on that authority. The licensing model is where the value metric lives, where entitlement rules are written, and where the definition of a billable event is encoded. Which metric that is belongs to a prior decision, covered in who should decide your value metric. Definitional authority is a property of the licensing model, and where it sits, and what the agreement lets its holder do with it, decides who wins the renewal argument.
A vendor who names the outcome picks a proxy metric that its own system can count. A buyer who names the outcome picks criteria drawn from their own experience of the result. Those are different metrics, often measuring adjacent results rather than the same one.
Who Confirms the Outcome Fired
Confirmation is a separate decision from naming. A counter records that a session closed. It does not record whether the customer’s problem stayed solved, whether another system finished the work, or whether a human intervened after the agent’s response. The gap between what the counter sees and what the customer experienced is a proxy-metric property rather than an engineering defect.
Who Absorbs the Gap Between the Proxy and the Result
Every outcome metric is a proxy. The vendor cannot observe the customer’s realized result directly. Someone absorbs the gap between the countable event and the business result. Which unit counts as the outcome is a separate question, covered in whose outcome a task-completion price measures. Who holds the pen and who confirms the count are the questions here. Where that absorption lands is an agreement question, and it is almost never stated explicitly at signature.
A shared definition at signature describes the negotiation and leaves the architecture undescribed. The definition is tested later, when the customer’s people have changed how they work and the product has changed what it closes on its own.
The link to value-based pricing strategy is useful context here: the upstream decision about what to charge for shapes every downstream dispute about whether the metric fired correctly.
A vendor we worked with billed on pageviews and carried a definitional argument into every renewal, because nobody could settle how many of the views were human. Moving the customer base onto a unit both sides could observe took the argument out of the room. At the next renewal with a large news organization, the same event on the same account, the dispute was gone and the renewal closed as a large expansion.
The Market Has Already Answered This Four Ways
The industry has no single answer to who holds definitional authority; it has at least four settlements already in production, and the differences between them are architectural rather than cosmetic.
Pattern 1: The Vendor Names the Outcome and the Vendor Counts It
The vendor writes the definition into the licensing model and operates the counter. The buyer’s only input is at contract time. If the buyer’s experience of the result diverges from the vendor’s count, the agreement either has an escalation path or it does not. Most do not. See the resolution rule in Agentforce pricing: Salesforce charges per resolution, and a conversation counts as resolved when it ends without a handoff to a human. The customer pays for an outcome that the vendor defines by what did not happen. The wider family of that rule is in pricing per resolution.
Pattern 2: The Vendor Names the Outcome and a Separate Check Has to Agree
A second system, operated independently of the primary counter, has to agree before the event bills. The buyer gains a layer of verification. The vendor absorbs the cost of the second check and the latency it introduces. This pattern is common where the buyer is large enough to negotiate it and skeptical of the vendor’s count.
Pattern 3: The Buyer Names the Outcome
The buyer writes the success criteria. The vendor prices against those criteria. This is the contrarian case. Atlassian credits versus HubSpot resolutions treats consumption and outcome as opposite bets on who carries the risk that usage turns into value: under consumption the buyer controls the count and carries that risk, under outcomes the vendor carries it and controls the definition. A buyer-authored definition keeps the price on outcomes but moves control of the count to the buyer’s side, which is where consumption pricing already puts it. The vendor is then pricing against a result it neither defines nor counts.
Pattern 4: Billing Eligibility Tied to Measured Quality
Eligibility to bill depends on a quality threshold as well as a count of closed events. A resolution that triggered a follow-up contact, or that received negative feedback, may not bill. Success is defined as the absence of a negative signal rather than the presence of a positive one. A metric built on what did not happen is measurable, and it differs from the customer’s outcome.
The four patterns differ in what each costs the vendor. Pattern 1 maximizes billable event volume and creates renewal exposure when the customer’s experience diverges. Pattern 2 reduces that exposure and reduces volume. Pattern 3 trades volume for durability. Pattern 4 prices on absence, which is measurable and contestable in different ways than a positive count.
Where Does Your Pricing Architecture Actually Stand?
A few questions return your pricing architecture score and show which of your licensing, packaging, and pricing decisions needs attention first. Real diagnosis, not a mailing-list toll.
Why Handing the Buyer the Pen Can Protect the Vendor
The prevailing position treats vendor-held definitional authority as the natural default and buyer input as a concession extracted under negotiation pressure. The durable arrangement often runs the other way.
What the Vendor Gives Up When the Buyer Writes the Definition
A buyer-authored definition is usually a narrower one. The buyer writes criteria drawn from their own experience of the result, which excludes events the vendor’s counter would have included. The vendor trades billable event volume for a meter that the buyer cannot dispute at renewal, because the argument would be with their own criteria.
That trade is per-vendor. It depends on the product, the Customer Group, and how much of the work the vendor controls. Where the vendor controls most of the execution, the trade is often favorable. Where the result is commingled with the customer’s people and other systems, a buyer-authored definition may exclude so much that the model becomes unprofitable.
The designing outcome-based pricing analysis covers the margin-protection design decisions in detail. Those decisions belong in that piece. What belongs here is the authority question: who holds the pen when the definition is re-opened.
What the Vendor Gets Back at Renewal
A vendor who holds the definitional pen enters the renewal argument in a weaker position than it appears. The buyer can contest the definition itself. A vendor whose definition the buyer wrote enters the renewal with a record the buyer cannot easily repudiate.
Renewal friction is a realized price problem. The invoice that survives the renewal argument is the one built on a definition nobody can walk back. See the enterprise SaaS pricing framing on renewal architecture for the structural argument.
The trade-off is real. A narrow definition that nobody disputes at renewal can also capture less of the value the product creates; a broad definition captures more and is argued over every cycle. Narrowing the definition to end the argument can cost more revenue than the argument did. Where that trade lands is a per-vendor judgment, and there is no universal rule for which to choose.
We made this move on our own private equity offer. A diligence read used to count per letter of intent, which is our event. It now counts only when the sponsor’s investment closes, which is the sponsor’s own definition of a deal, and the argument about what counted stopped at the first conversation.
Confirmation Is a Decision of Its Own
Naming and confirmation are separate decisions with separate owners. Most outcome-priced agreements conflate them, which is where disputed invoices live.
What a Resolution Counter Can See, and What It Cannot
A resolution counter records that a session closed without escalation, or that a ticket reached a closed state, or that a conversation ended without a follow-up contact within a defined window. It counts the absence of a negative signal, or the presence of a terminal state.
It does not record whether the customer’s underlying problem was resolved. It does not record whether a human resolved the issue after the agent’s response. It does not record whether the customer returned the next day with the same problem under a different ticket. Each of those gaps is a proxy-metric property rather than a measurement failure. Some proxy is always required.
Controlled experimental work on expert-service markets finds that a guarantee attached to the result changes seller behavior, while a more legible invoice does not. Transparency of the bill reduces information asymmetry without changing the incentive structure. The guarantee is what moves behavior. That is the strongest evidence we have that a confirmation pattern has to carry contractual weight. Showing the customer the count changes nothing by itself.
Research on performance-based advertising contracts shows that moving the charge onto a later event shifts risk and creates an incentive problem at the definitional boundary. The vendor optimizes the counted event. What sits just outside the counted event degrades. In plain terms, when only one part of the job is paid for, the vendor’s effort moves to that part and away from everything else. The same thing happens with any outcome metric where the proxy is narrower than the result.
For the cost of a run that produces nothing billable, when the agent is wrong covers who absorbs that exposure.
Who Arbitrates When the Counter and the Customer Disagree
Four confirmation patterns are live in the market. The vendor’s counter alone; a separate automated check that has to agree before the event bills; measurement against a baseline drawn from the customer’s prior record; and the customer’s own confirmation.
Each pattern moves the arbitration to a different place. The vendor’s counter puts arbitration in the contract escalation clause, if one exists. A separate automated check puts it in the check’s own methodology. A baseline comparison puts it in the baseline definition. Customer confirmation puts it in the customer’s operations team.
Where the arbitration lives is a design decision rather than a technical default. If you need to talk through where your confirmation architecture puts the arbitration burden, talk to an expert before the definition is encoded.
One we have seen: a product priced per application user served both commercial apps with outside users and apps the customer’s own employees had built for internal use. The vendor’s counter billed every user. The customer argued the employees out of the count, and one metric split into two populations the contract had never separated.
The Questions to Settle Before the Definition Carries a Price
These are tests; no scoring framework sits behind them. Apply them to any outcome metric before the definition is encoded in the licensing model.
The counterfactual test. If the result had arrived without the software, would the meter still have fired? A definition that cannot answer this question has an attribution gap. Empirical procurement contracting research finds that the choice between fixed-price and outcome-contingent contract forms tracks how complex and how verifiable the work is. Where attribution is ambiguous, outcome-contingent forms create disputes.
The commingling test. How many other actors touched this result, and does the agreement state what the vendor’s share is? A commingled outcome is one produced by the software together with the customer’s people and other systems. Buyers argue commingled outcomes downward. The agreement either settles the vendor’s share upstream or the renewal settles it downstream.
The divergence test. When the counter and the customer’s own view of what happened disagree, what does the agreement say happens next? If the agreement does not say, the renewal does. Renewal, well after signature, is the moment the definition is tested.
The trajectory test. If the customer’s people improve at using the product, or the product improves on its own, does the definition still mean what it meant at signature? As the share of work the system closes on its own rises, a per-success unit and a per-attempt unit converge on the same monthly bill. A definition that does not account for that trajectory is a definition that will be re-opened.
Each vendor answers these questions for its own product and its own customers. There is no general answer.
