The four shapes
Four ways to pay for engineering, and what each one buys
Read the rows as trades, not rankings. Every cell that looks like a weakness is the price of a strength in the same column.
| Option | What it optimises for | Where it goes wrong | How change is controlled | Best for |
|---|---|---|---|---|
| Time and materials | Flexibility. Scope can move weekly because you pay for effort, not for a promise about scope. | Drift. Without an engaged buyer setting priorities, effort flows honestly into work that no longer matters. | By steering: a prioritised backlog, a budget cap and a review cadence replace formal change requests. | Evolving products where learning changes the plan. |
| Fixed price | Budget certainty on a defined scope — the vendor absorbs estimation error and prices that risk in. | Rigidity. Every discovery becomes a negotiation, and a squeezed vendor recovers margin invisibly, from quality. | By paperwork: formal change requests, each priced and signed — protection that costs speed at every turn. | Well-specified work with genuinely stable scope. |
| Dedicated-capacity retainer | Continuity. The same people, month after month, accumulating context that transactional models discard. | Complacency. Guaranteed revenue without outcome review lets a quiet team coast on a busy invoice. | By governance: capacity is fixed, so control lives in what the capacity is pointed at each cycle. | A product or platform with a continuous roadmap. |
| Outcome-based | Alignment. Payment follows a measured result, so the vendor's incentive points at your goal, not your hours. | Measurement disputes. A result the vendor does not fully control, or cannot be cleanly measured, poisons the relationship. | By definition: change means renegotiating the outcome itself, which is why the model suits few, well-bounded results. | A bounded result both sides can measure the same way. |
The matching rule
The work has a shape before the contract does
The reliable way to choose is to describe the work first and read the model off the description. How well is the scope known — specified in detail, sketched, or genuinely emergent? Is the engagement a project with an end or a product with a life? And how much steering can your side actually supply, week by week? Well-known scope with an end date points at fixed price. Emergent scope with an engaged owner points at time and materials. A living product with a roadmap points at dedicated capacity. A single measurable result that the vendor mostly controls — rarely, but genuinely — points at outcome-based.
Choosing by preference instead of by shape produces the classic failures: fixed price wrapped around discovery work, where every learning becomes a change order; time and materials handed to a buyer with no time to steer, where the metre runs honestly toward nothing; a retainer for a programme with a natural end, paying for capacity after the need has passed. The model does not fail in these stories. The matching does.
Hybrids
One programme, three prices, each set by what is known
Illustrative engagement shape
A GCC operator wants a customer platform rebuilt: the destination is clear, the detail is not, and the platform will need a team long after launch. No single model prices all three truths well.
- 01A short fixed-price discovery, because its scope — interviews, architecture options, a costed plan — is knowable in advance and the deliverable is defined.
- 02The build on time and materials against the plan, with a budget cap and monthly steering, because discovery reduced the unknowns without eliminating them.
- 03A dedicated-capacity retainer from launch, sized to the roadmap, because the platform's life is continuous and the context is worth keeping in one team.
Each phase carries the contract its uncertainty deserves, and the transitions are decision points where the buyer can stop, change course or change vendor with clean edges.
Our role
We sell under every model here, so the bias is named
A vendor page about contract models is not neutral, and pretending otherwise would be. Our interest is a signed engagement; the page's discipline is that a mismatched model costs us the relationship soon after it costs you the budget.
- 01
The model argued from your work
Our proposal names the model and shows the reasoning — the scope certainty, the steering you can supply, the life of the system — so procurement reviews an argument, not a preference.
- 02
Change control written for the model
The mechanism that governs change — steering cadence, change requests, or re-pointing capacity — is in the contract from the start, matched to the model rather than borrowed from another one.
- 03
Transitions with clean edges
Where a programme spans models, each phase ends at a decision point with its deliverables and knowledge handed over, so continuing with us stays a choice rather than a default.
The risk appetite and the procurement rules are yours; our part is a contract shape that matches the work honestly, and a stated willingness to lose a deal we would win under the wrong model.
Common questions
What buyers ask
Which contract model should a first offshore engagement start on?+
Usually with something small and bounded enough that the model matters less than the evidence it produces. A short, well-defined first piece — often workable as fixed price precisely because it is small — lets both sides observe the other's real behaviour: how estimates hold, how questions are asked, how change is handled. What we would not recommend first is a long retainer or an outcome-based structure, both of which assume a working relationship that does not exist yet. Choose the first contract for what it teaches you; choose the second for the shape of the work.
How does change control actually differ across the four models?+
It moves from paperwork to governance as the model gets more relational. Fixed price controls change with documents: a formal request, a price, a signature — slow, but the slowness is the protection. Time and materials controls it with steering: a prioritised backlog and a budget cap mean change is absorbed weekly, and the control is whether you actually attend to the steering. A retainer controls it by direction — the capacity is constant, so the only change that matters is what it points at. Outcome-based barely tolerates change at all, because the outcome is the contract. A vendor who cannot describe the mechanism for the model they are proposing has not run that model before.
Are outcome-based software contracts realistic, or a procurement fashion?+
Realistic in a narrow band, and a fashion outside it. The model needs three conditions at once: an outcome both sides can measure the same way, a result the vendor substantially controls, and stability long enough for the measurement to stay meaningful. A defined migration completed to agreed acceptance criteria can qualify. "Increase our revenue" does not — too many hands besides the vendor's touch that number, and the attribution argument arrives with the first invoice. Where the conditions hold, the alignment is genuine; where they do not, you have bought a future dispute with a bonus clause attached.
In this cluster
Test your work against the models
A structured conversation that maps your programme's phases to models, ending with a contract shape you can defend to procurement.
Start the conversation
