Who this guide is for
The Real Question: Who Carries Which Risk
| Risk | Fixed price | Time and materials |
|---|---|---|
| Cost overrun on the agreed scope | Supplier (that is what you pay the premium for) | You, unless there is a ceiling |
| Scope change during the project | You, through priced change requests | You, but changes cost only the hours they take |
| Quality under pressure | You: a supplier losing money cuts testing and refactoring first | Shared: quality is visible and paid for, if you check it |
| Ambiguity in the requirements | Contested: each side reads the gap in its own favour | You: ambiguity becomes hours |
| Management effort | Lower during delivery, much higher before signing | Higher during delivery: someone must steer every week |
| Speed of adapting to new information | Slow: every change is a negotiation | Fast: you reprioritise the backlog |
The Five Contract Models You Can Actually Sign
| Model | You pay for | Cost risk sits with | Works best when | Typical failure |
|---|---|---|---|---|
| Fixed price (lump sum) | An agreed deliverable at an agreed price | Supplier | Scope is stable, specified and testable | Change-request war; quality squeezed at the end |
| Time and materials (T&M) | Hours or days worked at agreed rates | Client | Scope is discovered as you go and you can steer weekly | Open-ended burn with nobody watching |
| Capped T&M (not-to-exceed) | Hours worked, up to a ceiling | Shared: client up to the cap, supplier above it | Scope is mostly known but details will move | Cap set too low, work stops at 90% done |
| Fixed price per phase or sprint | A fixed price for each short, defined increment | Supplier, one phase at a time | A long project can be cut into releases of a few weeks | Phases defined too large, so it becomes plain fixed price |
| Dedicated team (retainer) | A team's capacity per month | Client | Ongoing product development with a stable backlog | Paying for capacity without measuring output |
What the Evidence Says About Software Budgets
| Finding | Number | Source (date) |
|---|---|---|
| Average cost overrun across 1,471 IT projects | 27% | Flyvbjerg and Budzier, Harvard Business Review (Sep 2011) |
| Share of those projects that were 'black swans' | 1 in 6, with 200% average cost overrun and almost 70% schedule overrun | Flyvbjerg and Budzier, HBR (Sep 2011) |
| Large IT projects: budget, time and value versus plan | 45% over budget, 7% over time, 56% less value than predicted | McKinsey with the University of Oxford, 5,400+ projects (2012) |
| IT projects that go so badly they threaten the company | 17% | McKinsey with the University of Oxford (2012) |
| Which projects carry the highest overrun risk | Software projects run the highest risk of cost and schedule overruns | McKinsey with the University of Oxford (2012) |
| When the US government allows time and materials | Only when it is not possible to estimate accurately the extent or duration of the work | FAR 16.601 (current text) |
| When the US government considers fixed price suitable | With reasonably definite functional or detailed specifications and a fair price establishable at the outset | FAR 16.202-2 (current text) |
What Fixed Price Really Buys You
- The risk premium. A competent supplier prices the tail into the quote. You pay for the average overrun whether it happens or not. Suppliers who do not add a premium are either very sure of the scope or have not understood it.
- The specification you must write. A fixed price is only as fixed as the document behind it. Producing a specification that a stranger can price and test takes weeks of your best people's time, and it is usually underestimated more than the software.
- The change-request economy. Every gap and every new idea becomes a priced change, negotiated one at a time from a position where you cannot switch supplier. Some agencies win fixed-price bids low and make their margin on changes.
- The quality squeeze. When the supplier starts losing money, the first things to go are the ones you cannot see in a demo: automated tests, documentation, security hardening, refactoring. You inherit the bill later as maintenance cost.
- Adversarial incentives. On every ambiguous line, the supplier's interest is the cheapest reading and yours is the richest. The relationship turns into contract interpretation at the moment you most need collaboration.
Michele's take
What Time and Materials Needs From You
- A product owner with authority on your side, who can say yes and no to scope within days, not weeks.
- A weekly burn report showing hours by person and by feature, compared with the estimate for that feature.
- A prioritised backlog so that if the money runs out, it runs out after the most valuable things are done.
- A shared definition of done, including tests, review and deployment, so that hours buy finished work and not demos.
- The right to see and change the team: who is billing, at which seniority, and the option to replace someone who is not productive.
- A ceiling with an early warning, so that the burn never becomes a surprise in the monthly invoice.
The Scope Certainty Test
- Testability: could you write acceptance criteria today for at least four fifths of the scope, precise enough that two people would agree whether each one is met?
- Stability: will the requirements survive contact with real users, or is part of the point of the project to learn what users need?
- Integrations: are the systems you must connect to known, documented and accessible, with test environments the supplier can use from week one?
- Comparables: has the supplier, or anyone you can ask, delivered something substantially similar that the price can be checked against?
- Freeze tolerance: can the business live with the scope frozen for the duration of the contract without losing something important?
| Yes answers | Contract model | What to do first |
|---|---|---|
| 5 of 5 | Fixed price for the whole deliverable | Invest in the specification and the acceptance tests; check the supplier can absorb a bad overrun |
| 3 or 4 of 5 | Fixed price per phase, or capped time and materials | Cut the work into releases of a few weeks; fix price or cap each one, and fix the 'no' answers in the first phase |
| 0 to 2 of 5 | Time and materials, after a paid discovery | Buy a 2-to-6-week discovery at a fixed fee; repeat the test at the end of it |
The Hybrid That Usually Wins: Discovery, Then Phased Commitments
- Paid discovery at a fixed fee (2 to 6 weeks). The output is a backlog with acceptance criteria, an architecture, an integration map with real access tested, and an estimate with a range. You own all of it and can take it to another supplier.
- First release on fixed price per phase or capped time and materials. Short phases, each with its own price or ceiling and its own acceptance. The risk of each phase is small enough for the supplier to carry honestly.
- Evolution on time and materials or a dedicated team. Once the product is live, the scope is decided by users and data every sprint; a fixed price at this stage only adds friction.
“Customer collaboration over contract negotiation.”
— Manifesto for Agile Software Development · 2001
Which Model for Which Situation
| Situation | Model I would choose | Why |
|---|---|---|
| MVP for a funded startup | Fixed-fee discovery, then capped T&M | Learning is the purpose; the cap protects the runway |
| Replacing a legacy core system in a bank or insurer | Fixed price per phase with acceptance per phase | Budget approval needs numbers; hidden legacy rules make a single fixed price fiction |
| Project with a hard regulatory deadline | Fixed price on a narrowly specified scope, T&M for the rest | The deadline scope can be specified; everything else should not hold it hostage |
| Integration against documented third-party APIs | Fixed price | Scores 5 of 5 when access and test environments are real |
| Ongoing product development | Dedicated team or T&M with a monthly ceiling | Scope changes every sprint by design |
| Data migration or bounded internal tool | Fixed price | Built many times before; source, target and rules are known |
Contract Clauses That Protect Your Budget in Either Model
- A ceiling with notice: the supplier must warn you in writing when 75% of the cap or phase budget is consumed, with a forecast to completion.
- A change-control procedure: every change request priced within a fixed number of working days, with its impact on the timeline, and nothing billable without written approval.
- Acceptance criteria and a test period: what 'done' means for each deliverable, how long you have to test it, and what happens if it fails.
- Payments tied to working software: milestones paid on accepted releases, not on elapsed time or delivered documents.
- Repository and environment access from day one: the code lives in a repository you own or can read at any time.
- Termination for convenience with handover: you can stop at the end of any phase, paying for accepted work, and receive code, documentation and credentials.
- Named key people and a rate card: who will do the work, at which seniority, at rates fixed for at least twelve months.
- Intellectual property on payment: rights to the code you paid for transfer to you, with a list of third-party components and their licences.
Red flags in a proposal
What the Law Says When Things Go Wrong
| Jurisdiction | Rule | What it means for you |
|---|---|---|
| United States (federal procurement) | FAR 16.202-1: firm fixed price not subject to any adjustment. FAR 16.601: T&M needs a ceiling price the contractor exceeds at its own risk | Even the public sector caps time and materials; you should too |
| Italy | Civil Code art. 1661: the client may order variations up to one sixth of the agreed total price. Art. 1664: unforeseeable cost changes above one tenth allow a revision of the part exceeding one tenth | A fixed price is not frozen by default; agree in writing how changes and revisions work |
| Spain | Civil Code art. 1593: a contractor on a lump-sum price (ajuste alzado) cannot ask for more because wages or materials rose, only for plan changes the owner authorised. LCSP art. 309: public service contracts may be priced by units of time or a lump sum | Changes you approve informally can reopen the price; document every approval |
| Saudi Arabia | Civil Transactions Law (in force December 2023) art. 470: on measured work the contractor must immediately notify a significant increase in quantities. Art. 471: no extra payment for work needed to deliver the agreed design | Require early written notice of overruns in the contract, whatever the model |
Mistakes I See Repeatedly
- Choosing the model before the scope. The model is a consequence of the Scope Certainty Test, not a policy.
- Comparing a fixed price with a T&M estimate as if they were the same number. One includes a risk premium and excludes changes; the other does the opposite.
- Signing time and materials without an owner. Nobody steers, the burn becomes the plan.
- Buying a fixed price to avoid managing the supplier. You will manage it anyway, through change requests and disputes.
- Skipping discovery to save a few weeks. The weeks come back as months of rework.
- Setting a cap with no warning threshold. Work stops at the cap, usually just before the release.
A 30-Day Plan Before You Sign
- Days 1–3: name the owner. One person on your side with authority over scope and budget, and the time to use it.
- Days 4–7: run the Scope Certainty Test with that owner and write the answers down with the evidence for each yes.
- Days 8–12: choose the structure. Five yes: fixed price. Three or four: phases or a cap. Fewer: discovery first.
- Days 13–20: ask two or three suppliers to quote the same structure, with assumptions, exclusions, team names and the change-request rate.
- Days 21–26: compare like with like. Normalise the quotes for risk premium, exclusions and changes; check references on a project of similar size.
- Days 27–30: negotiate the clauses, not just the price. Ceiling with notice, change control, acceptance, repository access, termination with handover.
Conclusion
If you want a second opinion