Fixed Price vs Hourly Billing for Software Projects

The two models, honestly described

Hourly (or time-and-materials) bills the hours worked at an agreed rate, with an estimate that carries no ceiling. It is flexible, starts fast, and puts every risk — slow work, wrong turns, scope drift — on the buyer's invoice. Its defenders call this fairness: you pay for exactly what happens. The catch is that you also pay for everything that happens slowly, and you find out the total only at the end.

Fixed price commits to a number for a defined scope before work starts. The vendor absorbs overruns, which concentrates their attention wonderfully on efficiency and on getting the scope clear up front. Its known weakness is rigidity: change costs friction, and a vendor losing money on a fixed job is tempted to cut corners. Both models work; they just fail differently, and the failures are predictable.

Check which model a contract actually is, not which it claims to be. A 'fixed price' that lists hourly rates for anything outside a vaguely drawn scope is an hourly contract with a deposit. A 'time-and-materials' agreement with a hard cap and a committed deliverable is fixed price in all but name. Read for who pays when the estimate is wrong; that single question classifies any contract in front of you.

Side by side

The comparison that matters is not philosophical — it is who pays when reality diverges from the plan, because it always diverges somewhere. Run down the table asking one question per row: in our project, which cell would we rather live in when things wobble? Most defined projects answer the same way in every row.

How big is the risk being allocated? The research answer

The billing-model choice allocates a risk whose size is published. McKinsey research with the University of Oxford, across more than 5,400 IT projects, found large projects running 45% over budget and 7% over time on average, delivering 56% less value than predicted. The Standish Group's CHAOS research found challenged projects averaging 189% of their original cost estimates, and its recent figures rate only 31% of projects successful, with 50% challenged and 19% failed.

Under hourly billing, those distributions are your invoice. Under fixed price, they are the vendor's margin problem. The same research carries the mitigations, too: small, tightly scoped projects succeed at far higher rates than large ones in the CHAOS data — which is why phasing big ideas into small fixed deliverables beats both a mega-project and an open meter.

Where hourly genuinely wins

Fairness demands the honest cases for hourly. True research work — 'we do not know if this is possible' — cannot be fixed-priced without the vendor padding the number to cover unknowable risk, so you would pay for the uncertainty either way and hourly is at least transparent about it.

Long-running embedded engagements are the other case. If you want a developer effectively on your team for a year, shipping whatever each month needs, a monthly rate or hourly arrangement matches the reality better than pretending the year is one definable project. The mistake is letting these two legitimate cases justify hourly billing for a booking system with a feature list — which is a defined project wearing an open-ended contract.

Even inside the legitimate cases, structure helps. Research work can be boxed into a fixed-price spike — 'two weeks to prove the approach, written findings either way' — which buys the exploration without the open meter. Embedded engagements benefit from quarterly reviews of what the hours actually shipped. Hourly is a tool, and like any tool it works best inside a frame.

A worked example: the same project under both models

Price a defined project — a small client portal — both ways, with every input shown. Hourly: the vendor estimates 120 hours at $85 an hour, so the anchor is $10,200. Apply the McKinsey-average 45% drift and the invoice lands at 174 hours × $85 = $14,790. Nothing went wrong; the estimate just met reality, and the buyer paid the difference.

Fixed, same traditional vendor: they quote $12,500 — the estimate plus a risk premium — and absorb any drift. The buyer pays more than the hourly anchor but caps the downside, and the vendor now profits by finishing efficiently rather than slowly. Fixed, AI-assisted studio: generation collapses the production hours, so the quote is $2,400 with the same cap and a warranty. The models allocate risk; the production method sets the level. Both choices matter, and they are separate decisions.

Notice what the worked numbers do to the usual objection that fixed prices carry padding. They do — the $12,500 quote includes roughly $2,300 of it — and the padding is still cheaper than the average drift it replaces. The buyer who chose the $10,200 anchor did not avoid the risk premium; they paid it retrospectively, with interest, as the overrun.

The failure modes of each — and their prevention

Fixed price fails through the scope document. Vague scope makes everything arguable: the vendor says the report screen was never included, you say obviously it was, and the relationship sours. It also fails through quality: a vendor underwater on the number ships the letter of the scope as cheaply as possible. Prevention is the same for both — scope precise enough that 'in or out' is rarely debatable, a warranty that survives delivery, and acceptance criteria written down.

Hourly fails quietly, which is worse. There is no single moment of dispute; the estimate of 120 hours just becomes 190, meeting time appears on invoices, and every small decision is deferred to you because deciding costs the vendor nothing. Prevention is caps and checkpoints: a not-to-exceed number, weekly hour reports, and a standing question — what did the hours buy this week?

The deeper prevention for both models is the same: shrink the thing being priced. A 10-day fixed phase can only fail by 10 days' worth; an open-ended six-month engagement can fail by any amount. The Standish data showing small projects succeeding at far higher rates than large ones is not a curiosity — it is the operating instruction.

How to keep a fixed quote honest

Fixed price only works when the paperwork underneath it does. Four checks cover nearly everything, and all four can be verified in an afternoon before any contract is signed — which is precisely when they are cheap.

• A scope document with counts and names — user types, screens, integrations — plus explicit exclusions.
• A written change process: changes welcome, each priced as its own small fixed quote before work proceeds.
• A definition of done that includes deployment to production and a stated warranty period.
• Phasing for anything large: each phase fixed-priced and shipping something usable, so you can stop, redirect or continue with full information.

Why AI-assisted studios can offer fixed prices at all

There is a structural reason fixed pricing clusters at the AI-assisted end of the market. Fixed price requires the vendor to predict effort accurately, and ai based software development makes effort dramatically more predictable: boilerplate, tests and first-draft interfaces are generated rather than estimated, leaving senior engineers' time — architecture, review, release — as the main variable, and that is the part experienced teams can forecast.

The underlying shift is well documented: Stack Overflow's 2025 Developer Survey found 84% of developers using or planning to use AI tools in their work, and GitHub's controlled experiment measured developers completing a standard task 55.8% faster with an AI assistant. When the unpredictable majority of the hours disappears, committing to a number stops being brave and starts being ordinary.

What 79 fixed-price projects actually look like

The model is testable against data rather than argument. Across PINCLER's 79 documented projects — every one a fixed price between $500 and $2,500, published at /research/what-you-can-build — the median project closed at $1,450 with a median delivery of 13 days. Four projects shipped within 7 days, 33 within 14, and all 79 within 30. Budget reach in the same dataset is worth knowing before any negotiation: $1,000 covers 55 of the 79 documented projects, $1,500 covers 75, and $2,000 covers all of them.

The distribution is the point. Hourly billing's defenders argue that fixed prices must hide either padding or corner-cutting; a published dataset where every project landed inside a $2,000 spread, on schedule, is what the alternative looks like when the production method genuinely supports the commitment. It is why every PINCLER project is fixed price, quoted in writing within one working day of a free 30-minute call, with larger ideas split into fixed-price phases — and why that quote is a useful control to weigh any hourly estimate against.