Fixed price or time and materials: who carries the risk
A billing model is not an administrative detail — it decides who loses money when the project slips. Read a proposal for where the overrun lands, not for the headline price.
The choice between a fixed price and time and materials is presented as a commercial preference. It is nothing of the sort. It is the clause that decides who loses money the day the project runs three months late. Until you read it that way, you are negotiating the headline price while believing you are negotiating the risk.
Two models, two places for the overrun to land
The mechanics are simple enough to fit in two sentences. On a fixed price, the supplier commits to a scope and a number; the gap between their estimate and reality comes out of their margin. On time and materials, you buy days; the same gap comes out of your budget, one invoice at a time. Everything else — the method, the rituals, the quality of the code, the seniority of the people — is independent of the model, whatever a proposal tells you.
Neither model is morally superior, and the choice turns on a single question: can the scope be described today, in writing, with its business rules and its edge cases? If yes, a fixed price protects you. If no, it traps you, and the trap closes through change orders rather than through invoices.
What each model rewards, regardless of the people involved
A fixed price rewards finishing early, which is good, and it rewards classifying every new request as out of scope, which is not. Above all, when the price was cut too fine to win the work, it rewards adjusting on what does not show in a demo: tests, data migration, documentation, performance under real volumes, error handling. That is the blind spot of the model — the adjustment variable is invisible at acceptance and expensive eighteen months later.
Time and materials rewards nothing. That is not an accusation, it is the whole point: no mechanism pushes anyone to decide that the job is done. The steering function moves to you — prioritising, arbitrating, saying no, deciding when a subject is finished. That work is real, it takes something like half a person permanently, and it is almost never in the budget.
The two mechanical drifts
A fixed price on a vague brief produces change orders
A supplier prices what they read. What they do not read, they either do not price, or they price as risk with a margin on top. On an eight-page brief there are two strategies: quote generously and lose to whoever quotes tight, or quote tight and recover through change orders. The market selects the second. This is not dishonesty, it is selection — the one who priced the work honestly is not there to explain themselves, they were eliminated at the bid opening.
The symptom is easy to spot: the first change order arrives before the end of month two. From that moment the conversation shifts from "how do we make this better" to "is this inside the scope", and the relationship degrades before the software exists. The fixed price did not fail because the supplier was bad. It failed because nobody paid for the work that would have made the commitment possible.
Time and materials without a strong internal owner never ends
Time and materials transfers the decision function to the client. Without someone who has both the time and the authority, priorities shift with whoever spoke last, developers build exactly what they were asked for, and the spend advances while the scope never closes. The drift is invisible on any single monthly invoice — it appears in the cumulative total, usually around month twelve, when somebody finally adds it up.
The reliable warning sign is an engagement that has been running for two years and where nobody can produce a written list of what remains to be done. Not because the list is hard to write, but because writing it makes the end date visible, and no one on either side has an interest in that.
The three hybrids that actually work
Paid scoping, then fixed price
The scoping phase is bought separately — workshops, a prototype on the priority use case, a written rules document — and produces the material that makes a firm price possible. It is normally deducted from the fixed price if the project goes ahead. This is the only arrangement where a committed price is not a bet.
Fixed price per batch
Only the next batch is firmly committed; later batches are estimated, not signed. You are not buying certainty over eighteen months, you are buying it over three, and you keep the right to redirect the rest once you have seen the first batch running.
Capped time and materials
Days billed as worked, with a contractual cap and an obligation to raise the alarm at, say, 70% of it. Watch the false comfort: a cap without a content commitment guarantees you a budget, not a result. You can reach the cap and hold an unusable application.
What the comparison actually says
| Fixed price | Time and materials | Fixed price per batch | |
|---|---|---|---|
| Who pays for the overrun | The supplier | The client, every month | The supplier, on the current batch only |
| What the client must bring | A written scope before day one | An owner able to run a technical team day to day | An owner who arbitrates the content of the next batch |
| Effect of a change of need | Priced change order, decided before it is built | Absorbed without discussion, visible in the cumulative total | Deferred to the next batch, usually with no change order |
| Budget visibility | Complete at signature, on a frozen scope | None beyond the current month | Firm on the batch, estimated beyond it |
| Adjustment variable under pressure | What does not show in a demo: tests, migration, documentation | The schedule first, then the budget | The content of the batch, negotiated in advance |
| How it ends badly | A change-order war from month two | An engagement that never ends: nobody writes down what is left | Batches shrinking to protect the headline price |
The change procedure is where the contract is really decided
A change order is not an incident, it is the normal life of a project. What separates a workable contract from a painful one is that the order of operations is written down before anyone needs it.
- Every new request is qualified in writing: does it fall inside the signed scope, does it clarify that scope, or is it a genuine addition? A meaningful share is settled here, at no cost.
- An addition is priced in days, with its effect on the schedule of the remaining batches, and the price is given before the decision — never after the work is done.
- The business owner picks one of four options: pay for it, remove an item of equivalent effort from the scope, defer it, or drop it. The choice and its reason are written down.
- The change order updates the contract and the scope document, which stays the single reference. Nothing out of scope starts before signature.
What should worry you is not the presence of such a procedure. It is its absence. A contract with no change procedure has not abolished change orders — it has abolished the rule that governs them, which leaves the strongest party to define the scope after the fact.
Reading a proposal for what you are actually buying
- A firm price with a scope described in three bullet points: you are buying disguised time and materials, whose adjustment variable will be the definition of the word "module".
- A day rate plus an "indicative" workload: you are buying time and materials, which is perfectly honest as long as it is stated as such.
- A fixed price with no paid scoping beforehand: either the supplier did that work for free and will recover it, or they did not do it and will recover it anyway.
- A payment schedule tied to calendar dates rather than acceptance milestones: you are paying for time passing, whatever the model on the cover page.
- A clause saying minor changes are included, with no definition of minor: that is the future dispute, already drafted.
- No mention of acceptance — no list of what will be tested, no reservation period, no statement of what signature triggers. At that point the billing model no longer matters: nothing triggers anything.
A fixed price is not a guarantee. It is a transfer of risk, and a transfer of risk is only worth something if the party taking it was given the means to assess it. A firm price announced with no paid scoping, no scope written line by line and no change procedure transfers nothing at all — it merely postpones the moment you discover the invoice. Ask for the fixed price, by all means. Buy what makes it possible first.
Continue reading
What rebuilding a buying group's system really costs
Nobody publishes figures for this kind of project, which leaves general managers walking into board meetings empty-handed. Here are defensible ranges, batch by batch, and the three cost items no quote ever contains.
Rebates and year-end bonuses: what software must compute
The accounting difference between a rebate and a year-end bonus takes two paragraphs. The mechanics your software has to model — tiers, bases, pro-rata, versioned scales — take a project.