The Fixed-Bid Autopsy: Why Our Most Profitable Projects Started as T&M

Every agency founder eventually has the same argument with themselves at 1am: why did we quote this as a fixed price? We've been shipping client work for over a decade across web, mobile, and AI builds, and we finally sat down and traced the P&L of every project back to its pricing model. The pattern was ugly, honest, and useful.

This is the autopsy.

The Three Models, Honestly Described

Before anyone writes to tell us we missed a model: yes, there are hybrids, caps, retainers, and outcome-based deals. We'll get to those. But the base cases most clients actually ask for are three.

Fixed-bid means you quote a number for a defined scope. The client likes it because the number can go into a budget cell. You like it because, in theory, if you deliver efficiently, the margin is yours to keep.

Time & Materials (T&M) means you bill for hours or days worked, usually against an estimate and a cap on cadence (weekly or monthly). The client takes on the risk of overrun. You take on the risk of looking expensive on a Friday afternoon when a story blew up.

Capped T&M or not-to-exceed is the awkward middle child. You bill actuals, but you promise not to cross a ceiling without a signed change. In practice it behaves like fixed-bid for the client and like T&M for your bookkeeping, which is to say it combines the downsides of both unless you run it tightly.

Why clients ask for fixed-bid even when it's wrong

Non-technical buyers don't want a pricing model. They want a number that makes a procurement form go green. Fixed-bid feels safe to them because the risk of overrun appears to sit with you. What they often don't realise is that fixed-bid pricing taxes them — because you've priced in the risk, and if the project goes smoothly, they overpaid.

A good discovery conversation makes this trade-off explicit. A bad one lets both sides pretend risk doesn't exist until the first scope dispute.

What the Autopsy Actually Showed

We pulled roughly five years of projects, bucketed them by pricing model, and compared contribution margin after delivery costs (engineering, PM, QA, design — not overhead). We're not going to publish exact numbers because every agency's cost base is different, but the shape of the finding was consistent and, in our experience, generalisable.

• Fixed-bid projects had the highest variance. The good ones were our best margins. The bad ones were our worst — a few actually lost money once rework was counted.
• T&M projects had the tightest variance and a healthy, boring median margin. Almost none lost money. Almost none were spectacular.
• Capped T&M was the surprise: it behaved like T&M on upside and like fixed-bid on downside. The cap only ever bit us; it never helped us.

The uncomfortable conclusion: fixed-bid is a bet on how well you understood the scope at kickoff. If your discovery was strong, you win. If it wasn't, you subsidise the client's learning.

The project phase that actually determines margin

When we sliced the data differently — by when the scope stabilised rather than by contract type — the picture got clearer. Projects where the client knew what they wanted by week two had great margins regardless of model. Projects where scope was still moving in week six were losses waiting to happen, especially under fixed-bid.

In other words: pricing model is downstream of scope clarity. If you can't tell which one you're in during the sales conversation, you shouldn't be quoting yet.

A Decision Rule That Actually Works

After the autopsy, we rewrote our internal rule for which model to propose. It fits on a napkin:

If scope is well-defined AND the client has shipped
software before     -> Fixed-bid (with change control)

If scope is well-defined AND the client has NOT shipped
software before     -> Capped T&M, with a mandatory
                       discovery phase billed separately

If scope is fuzzy OR the project is R&D-flavoured
(AI, integrations with undocumented systems)
                    -> Pure T&M, weekly billing,
                       two-week kill switch

The kill switch matters. Clients will accept T&M if they feel they can stop at any time. We bake in a clause that either side can end the engagement with two weeks' notice, no penalty. In five years we've had it invoked twice, both times mutually. It is the single most effective trust-builder we have.

What "well-defined scope" actually means

We used to say scope was defined when the client could describe the feature list. That was wrong. Our working definition now:

1. The user flows are drawn, not just described.
2. The data model has been sketched to at least second-level entities.
3. The three riskiest technical questions have been named out loud.
4. The client can tell us which feature they'd cut if we ran out of time.

If points 1 – 4 aren't true, we're in discovery, not delivery, and we price accordingly. We wrote more about how we structure that phase on our blog.

The Contracts That Prevent 1am Arguments

Pricing model is only half the story. The contract language around change, acceptance, and dispute is where projects actually die.

Change control. Every fixed-bid contract we sign now has a line that says: any change to the agreed scope will be quoted as a separate change order and must be signed before work begins. No ambiguity, no "we'll figure it out at the end." Clients respect it when you enforce it from week one and resent it when you spring it in week ten.

Acceptance criteria. We tie payment milestones to acceptance of specific deliverables, not to calendar dates. A date-based milestone invites the client to delay feedback. A deliverable-based one makes them a participant.

Assumptions log. Every fixed-bid proposal we send now has an assumptions section — things like client will provide design assets by week two, third-party API documentation is accurate, no more than two rounds of stakeholder review per milestone. If an assumption breaks, the price can move. This single page has saved us more arguments than any legal clause.

The one clause we stopped using

We used to include an "out-of-scope hourly rate" in fixed-bid contracts, so if the client asked for extras, we'd bill at a defined hourly rate. It sounds reasonable. In practice it created perverse incentives: the client treated it as an all-you-can-eat buffet, and our team got dragged into unscoped work instead of finishing the committed work. Now we require a change order and a timeline impact statement for anything beyond the original scope. Smaller invoices, happier teams.

When to Walk Away From the Deal

The autopsy also surfaced a hard truth: our worst fixed-bid losses weren't bad luck. They were deals we should have declined. The signals were visible in the sales process.

• The buyer insists on fixed-bid but can't answer our discovery questions.
• Procurement is driving the conversation, not the product owner.
• The timeline is set by an external event (conference, funding round) that nobody will admit is immovable.
• There are more than three stakeholders with veto power.

Any two of those and we propose T&M with a discovery phase. If the client refuses, we pass. We used to think passing on revenue was expensive. Running the autopsy showed us that losing on a fixed-bid is far more expensive — not just in margin, but in team morale, which has a cost that doesn't show up in any spreadsheet.

Where We'd Start

If you're reading this and your agency's margin is wobbling, don't start by renegotiating rates. Start by pulling your last ten projects and bucketing them by pricing model and by when scope stabilised. The pattern will probably look like ours. Then rewrite your sales qualification to detect fuzzy-scope projects earlier, and build a discovery offering you can sell as its own small engagement.

Fixed-bid isn't the enemy. Fixed-bid without discovery is. Price the thinking separately, and the building gets a lot more predictable.