Fixed price or day rate: the rates, the risk, and a worked comparison.

Under a day rate, a supplier who finds a faster way to do something earns less for doing it. Nobody sets out to exploit that, but incentives shape behaviour whether or not anyone is being cynical, and over a long engagement it shows. Under a fixed price the incentive reverses: finding the faster route is rewarded, and the buyer is indifferent, because they agreed the outcome rather than the hours.

That was the whole of the original version of this guide, and it is still true. What it lacked was evidence. So this revision adds the measured UK day rates from live contract postings, the rates agencies publish, what the government’s own commercial guidance says about which party should carry which risk, and one project priced under both models with a best and a worst case. The point is not that fixed price is always right. It is that you should know what a day is worth before anyone sells you one.

ITJobsWatch tracks the day rates attached to live UK contract postings. These are not what agencies charge; they are what individual contractors were offered, which is the floor under every agency day rate you will see. All figures are for the six months to 14 September 2026. Two things stand out. Developer rates have barely moved in two years, with the UK median at £520 in 2024, £500 in 2025 and £525 now, while the number of postings has tripled. And the London premium for developers is small at the median, £529 against £525, but wide at the top, where the ninetieth percentile is £709 against £613.

ITJobsWatch, six months to 14 September 2026, from live contract postings. Contractors carry their own gaps, insurance and equipment, so these are not salaries; multiply by roughly 220 working days for an annualised figure.

Agencies quote more than contractors because you are buying a team, a process and somebody to blame. The published UK guides from 2026 agree with each other closely enough that a buyer can use them as a sanity check on any quote. Read them against the measured rates above and the markup is visible: a senior contractor at the ninetieth percentile costs £613 a day; a London agency’s published senior day rate is £600 to £900 and its hourly rate £75 to £200. The difference pays for management, a bench, and the risk premium if the price is fixed, which is not unreasonable provided you know that is what you are paying for.

Read on the dates shown. Tulip Tech warns that any quote citing £20 to £40 an hour as a UK rate is mislabelling offshore pricing; we agree.

The Cabinet Office’s Digital, Data and Technology Playbook, which central government must follow on a comply-or-explain basis, puts the principle in one sentence: "Ensuring that risks sit with the party best able to manage them is key to delivering value for money and successful outcomes." On pricing it says the approach "should reflect the level of certainty or risk around the scope and requirement. Where the scope of a project is certain, then fixed pricing may be appropriate and where there is increased uncertainty in scope, a variable approach may be more suitable." It also warns against paying for the agile process itself, "e.g. completion of a specific number of sprints", rather than for delivered outcomes.

The accompanying Contracting for Agile guidance note adds the middle way most private buyers never hear about: "You could further manage the commercial risk by using fixed-price sprints (contracted for in phases) rather than contracting for the entirety of the requirement on a fixed-price basis at the outset." That is how we price larger builds: one fixed price per phase, each with its own scope, so that a new idea becomes a new phase rather than a delay with no explanation.

The question for a private buyer is therefore not "fixed or time and materials" but "who is best placed to manage the risk that this takes longer than expected". For a defined build, that is the supplier, who has done it before. For genuine research, it is nobody, which is why research is honestly billed by the day and scoped small.

Time and materials looks cheaper on the quote because the quote is a best case. Helium42’s April 2026 survey of UK AI consultancy pricing reports that "SMEs experience 15-25% average project overruns under T&M engagement" and that fixed-price engagements make up 45 to 55 per cent of SME work for that reason. Tulip Tech’s June 2026 guide puts the other side: "A fixed-price contract gives cost certainty but prices in a risk premium of 10 to 25 percent." Both are right, and together they tell you when each model wins.

A fixed price is the supplier charging you a premium to carry the overrun. If the supplier has built this kind of thing before, the premium is small, because they know the overrun will be small; that is what experience is for. If they have not, the premium is large or the price is a guess. So a high risk premium is a signal about the supplier, not about the model. A day rate is you carrying the overrun yourself, at whatever it turns out to be, with the incentive pointing the wrong way. The average overrun in the published data is the same size as the risk premium, which means that on average a buyer pays roughly the same either way, and the only difference is who was exposed.

On average is the problem. Nobody experiences the average; you experience your project. Under time and materials the distribution has a long tail, and the tail is where the disputes live. A 25 per cent overrun on a forty-day estimate is ten extra days, which is survivable; the projects that end up in correspondence are the ones that ran to double, and under a day rate nothing in the contract stops that except your willingness to cancel a half-finished system. A fixed price with a written scope moves that conversation to before the work starts, when it is a scoping question rather than a dispute.

Take a business application a supplier estimates at forty developer days and eight project-management days. Price it three ways from the published figures: a contractor team at the measured UK medians, a London agency team at the published senior day rates, and a fixed price. For the worst case under time and materials, apply the 25 per cent overrun at the top of Helium42’s range; for the fixed price, apply Tulip’s 10 to 25 per cent risk premium to the contractor best case, which is what a fixed price is.

Read the table across, not down. The gap between the two columns of a time-and-materials row is the risk you carry; the fact that the fixed rows have no gap is the risk you paid a premium to remove. The gap between the rows is a different thing altogether: it is the number of days and the rate per day, which depend on who is doing the work and whether they have done it before. A buyer comparing quotes should separate the two questions, because a supplier can be cheap on one and expensive on the other, and the quote will not say which.

Arithmetic on the published figures above; not a client account. The fixed rows are the same in both columns because that is the point of a fixed price. Our band is lower than the market rows because a small senior team without agency overhead needs fewer days, not because a day is cheaper.

A fixed price only works if both sides accept a written scope with an explicit list of what is not included. Many buyers find that document uncomfortable, because it looks like the supplier limiting their obligations. It is doing the opposite: it is the only thing that makes the price meaningful. A fixed price without an exclusions list is a day rate with extra steps, because every ambiguity becomes a negotiation. Ours, and any good one, covers at least the following.

It also demands something of you. A fixed price assumes a certain pace of decisions, content and test users from the client, and the fair version of the contract says what happens to the timeline when they are late, which is usually that the date moves and the price does not. Buyers sometimes read that clause as the supplier protecting itself. It is, and it is also the clause that lets the supplier commit to a date at all, because the alternative is a date with a silent assumption behind it that nobody wrote down.

Genuine research, where nobody can know the shape of the answer in advance. A feasibility spike on whether a model can read your particular documents accurately enough is research; nobody should fix-price it, and nobody should let it run more than a few days without a written result. Ongoing support and maintenance, where the work is real but unpredictable in any given month, is honestly bought as a retainer with a stated response time or as days against a cap. And discovery, which is why discovery should be a separate small paid engagement that ends in a written scope you own, rather than free work absorbed into a quote.

The published market agrees on the shape. The AI Consultancy sells its day rate of £950 to £1,500 explicitly for "open-ended scopes" such as fractional advisory and remediation, and fixes everything else. Helium42’s figures put retainers at £3,000 to £15,000 a month for on-call advisory and hourly work at £150 to £400. Those are sensible ways to buy unpredictable work. They are poor ways to buy a defined build.

A supplier who insists everything can be fixed price is overselling. One who insists nothing can be is avoiding the scope conversation. The tell is whether they ask enough questions before the number, because the questions are where the risk gets moved. Capped time and materials, a day rate with a ceiling, is the compromise most often offered, and it is better than an open day rate. But the cap tends to become the price, because the incentive under it still points towards using the days, and the government’s agile note offers a cleaner middle: fix the price per phase, and let each phase be small enough that fixing it is honest.

Every project is a fixed price against a written scope with an exclusions list: a single-process tool £4,000 to £10,000, a business application £10,000 to £20,000, a first AI implementation £4,000 to £15,000, a custom AI system £8,000 to £25,000, excluding VAT. Larger builds are priced per phase in the way the government’s agile guidance describes, one fixed price per phase, so the risk stays with us and the decision to continue stays with you. We do not sell day rates, and the AI adoption audit at £2,500 to £6,000 is the paid discovery that produces the scope, credited in full against any build.

Where a piece of work is genuinely research, we say so and scope it as a short fixed feasibility phase with a written result, rather than pretending to fix a price we could not honestly stand behind. Support after go-live is agreed at go-live, with a stated response time, and never as a condition of handover: you hold the code, the accounts and the runbook whether or not you buy it.

Contract day rates are ITJobsWatch’s medians and percentiles from live UK and London postings for the six months to 14 September 2026, read on that date. Agency and freelance rates are from the 2026 UK guides listed below, read on 14 September 2026 with their publication dates recorded; where a guide cites ITJobsWatch or G-Cloud rate cards we say so. Overrun and risk-premium figures are Helium42’s and Tulip Tech’s, as published. Government guidance is quoted from the Cabinet Office’s Digital, Data and Technology Playbook and its Contracting for Agile guidance note, both on GOV.UK. The worked comparison is our arithmetic on those figures and is not a client account. All prices exclude VAT.