Time-and-materials contracting has one great virtue: it is easy to administer. You pay for hours, hours are easy to count, and everyone goes home. Its defect is equally simple: it pays for presence, not progress. The person billing you is compensated identically whether the feature ships this month or next quarter, and the structural incentive, however professional the individual, points toward the engagement continuing rather than concluding. As AI leverage makes output per person more variable than ever, the gap between what an hour costs and what an hour produces is widening, and buyers are noticing. The response is a slow market shift toward outcome-based structures: paying for the thing shipped rather than the seat filled. The shift is real, but it is harder than the pitch decks suggest, and it fails in predictable places. Here is the honest version.
The spectrum: T&M to milestones to outcomes
It helps to see the models as points on one axis, how tightly payment is coupled to result, rather than as rival ideologies. Each step along the axis transfers risk from buyer to provider, and priced correctly, the provider charges for carrying it.
| Model | You pay for | Risk sits with | Where it fits |
|---|---|---|---|
| Pure time & materials | Hours or days worked | Buyer, entirely | Exploratory work, ongoing ops, undefined scope |
| Capped T&M | Hours, up to a ceiling | Shared, crudely | Buyer wants T&M flexibility with a budget backstop |
| Milestone-based | Defined intermediate deliverables | Shared, per milestone | Buildable work with checkable stages |
| Outcome-based | A verified end result | Provider, mostly | Well-specified outcomes with trusted measurement |
| Outcome plus performance bonus | Result, plus quality above a bar | Provider, with upside | Mature relationships where the metric is robust |
What outcome pricing actually requires
Outcome-based deals do not fail because the idea is wrong. They fail because one of three prerequisites was missing at signing, and everyone discovered it during the dispute. First, a defined outcome: 'improve the data pipeline' is not an outcome; 'events from these four sources land in the warehouse within five minutes, with a documented schema and alerting, verified over two weeks of production traffic' is. Second, measurement both sides trust: the acceptance test must be written down before work starts, runnable by either party, and insulated from moving goalposts, if verification is a meeting rather than a check, the contract is a handshake. Third, scope discipline: an outcome price is a price for a scoped thing, and every 'while you're in there' request is a renegotiation, not a favor. Buyers who cannot hold that line should not buy outcome-priced work, because they will convert a clean contract into a resentful T&M engagement with extra steps.
- Write the acceptance criteria into the contract as executable checks wherever possible, test suites, measurable thresholds, demo scripts, not adjectives.
- Agree the verification procedure and who runs it before signing, including what happens when the check is ambiguous.
- Establish a formal change path: scope additions are re-priced, never absorbed silently by either side.
- Price the risk transfer honestly: a fair outcome price is higher than the T&M-equivalent estimate, because the provider now carries the overrun.
Where outcome pricing fails: research-shaped work
There is a class of work where outcome pricing is structurally wrong, not just hard. Call it research-shaped work: the deliverable cannot be specified in advance because discovering what is feasible is the substance of the job. 'Get the model's accuracy above 95% on our data' sounds like an outcome, but nobody knows before doing the work whether 95% is achievable at all, so an outcome contract either prices in a huge risk premium, collapses into disputes, or quietly incentivizes the provider to game the eval. The same applies to early product exploration, novel integrations against undocumented systems, and most genuine R&D. The tell is simple: if a competent expert cannot estimate the work within, say, a factor of two, it is research-shaped, and it should be bought as time-boxed investigation with a defined question, not as a promised result. The mature pattern is sequencing: buy a short T&M discovery phase whose deliverable is a confident scope, then buy the now-specifiable build as milestones or an outcome.
The hybrid structures that work today
Pure outcome deals remain rare because the prerequisites are demanding. What is actually spreading are hybrids that capture most of the alignment without betting the whole engagement on a perfect spec. The common thread: a base component acknowledges that skilled attention has a market price, while a contingent component ensures progress is what gets rewarded.
- Base plus milestone releases: a reduced ongoing rate (say 60-70% of market) with the balance released at defined milestones. The provider is never working free; the buyer is never paying full price for stall.
- Discovery-then-fixed: a short paid discovery sprint producing a scoped spec, followed by a fixed price on the now-defined build. Sequencing solves the specification problem instead of pretending it away.
- Outcome bonus on top of T&M: standard rates with a meaningful bonus for hitting a date or a quality bar. Weakest alignment, easiest adoption, often the right first step for a new relationship.
- Retainer with outcome gates: for ongoing embedded work, a quarterly retainer that renews against a small set of agreed outcomes rather than against sentiment.
The honest incentive analysis
Advocates of outcome pricing tend to narrate only one moral hazard: T&M providers stretching engagements. That hazard is real, but the honest ledger has entries on both sides. Under outcome pricing, providers are incentivized to cut invisible corners, skip the tests, the docs, the error handling that the acceptance check doesn't cover, which is why acceptance criteria must include the invisible work explicitly. Buyers, meanwhile, are incentivized to smuggle scope ('surely that was implied') and to slow-walk verification when cash is tight, which is why the verification procedure and payment timing must be mechanical. And under T&M, buyers are not innocent either: unlimited hours invite unlimited indecision, with the meter running on the buyer's own meandering. The point of a payment structure is not to make anyone virtuous. It is to make the profitable behavior and the desired behavior the same thing, on both sides of the table, and hybrids currently do that better than either pure model.
