The hourly rate is a strange unit when you look at it directly. It prices the supplier’s time, which is an input. It says nothing about what you receive, which is the only thing you wanted. Every other significant purchase a business makes is priced on the output.
That matters because software work has changed. Time billing survived in professional services because time was a reasonable proxy for effort, and effort was a reasonable proxy for value. Both links have weakened enough that the unit no longer describes anything useful. The awkwardness is now visible to buyers rather than only to suppliers.
Time billing solved a real problem at first
Software work is uncertain. Nobody knows exactly how long a thing will take, because the difficulty is discovered while doing it. Fixed prices under that uncertainty mean someone carries risk. Early in the industry, neither side had the information to price that risk.
Time-based billing allocates that risk to the buyer, in exchange for flexibility. It is a defensible arrangement when the supplier genuinely cannot estimate and the buyer genuinely wants to change direction. It is also, conveniently for the supplier, an arrangement in which being slow is not penalised.
Hourly billing creates three business problems
The problems with hourly billing are simple once you look at what the unit rewards.
- It misaligns incentives. A supplier billing by the hour earns more when work takes longer.
- It prices the wrong thing. A senior engineer who solves in two hours what a junior would take two days on is worth more and bills less.
- It gives the buyer nothing to hold. A time-and-materials contract has no definition of done, meaning no checkable finish line.
Nobody is deliberately slow, but nobody is under pressure to be fast either. The thousand small decisions about whether to do something the quick way or the thorough way all lean one direction.
The unit actively penalises expertise. That is why agencies staff projects with more people than the work requires and call it capacity.
The lack of a finish line matters commercially. There is no moment at which the supplier has met their obligation, so there is no moment at which the buyer can say this is not what we agreed. Overruns are not breaches; they are just more invoices.
AI made the hourly model much harder to defend
The tension existed before, and the recent shift made it acute.
If a task that took two days now takes four hours, a supplier billing by the hour has three options.
- Bill four hours and take an 80 percent revenue cut on that task.
- Bill two days and hope nobody notices.
- Stop billing by the hour.
Most of the industry has quietly chosen the second, which cannot last, because buyers are using the same tools and can estimate what things should take now. The gap between what is billed and what is plausible is closing from the buyer’s side.
We chose the third, and it was uncomfortable. Revenue per engagement fell. Volume and margin both rose, because a fixed price on faster delivery beats a day rate on slower delivery for everyone except the day rate.
A fixed price needs more than a different number
A fixed price is more than a different number on the same document. Four things have to be true, and each is work the supplier would rather avoid.
- A written scope with a definition of done, specific enough to be checked.
- An estimating discipline that survives being wrong.
- A change mechanism that is priced as a new increment, visibly, at the time.
- The willingness to decline work that cannot be fixed-priced honestly.
The written scope is the hard part and it is where the value is, because most projects fail on ambiguity rather than on execution.
If our estimate is poor, we absorb it. That means estimates have to be made carefully by someone accountable, and it means the firm has to be able to survive being wrong occasionally.
Scope changes are normal. What matters is that a change is priced as a new increment, visibly, at the time, rather than appearing as a surprise on an invoice.
Some work cannot be fixed-priced honestly, usually because the client’s own decisions are the unknown. Saying so is the corollary of the model.
Fixed price also costs the buyer something
Being fair, fixed price is not free.
You give up unlimited direction changes. Under time and materials you can pivot weekly and the meter simply runs. Under fixed scope, a pivot is a re-scope and it has a price. For genuinely exploratory work that constraint is wrong, and a retained team is the better arrangement.
You give up some flexibility on the margins. A supplier carrying delivery risk will scope conservatively and resist the small additions that would previously have been absorbed. That is the risk being priced, and it is the honest cost of the arrangement.
You also have to do more work up front. A fixed price requires a real specification, which requires your people to make decisions before building starts rather than during. Many organisations find this harder than they expect, and it is the most common reason a fixed-price engagement struggles.
Agencies object because real work is uncertain
The objection from every agency we have discussed this with comes in the same form: fixed price only works for simple, well-understood projects. Real work is uncertain, and pretending otherwise means either padding the estimate or losing money.
There is something in it, and the honest response has three parts.
- The padding is real and it is the price of the arrangement.
- Uncertainty is not uniform.
- Much of the claimed uncertainty is the supplier not having done the work of specification.
A fixed quote contains a risk premium. The relevant question is whether that premium is smaller than what the buyer would have paid in overruns. Across a portfolio it usually is, because the supplier can pool risk across projects and the buyer cannot.
Most projects are a large well-understood part and a small genuinely uncertain part. Fixing the first and structuring the second as a bounded discovery increment is more honest than treating the whole thing as unknowable.
Ambiguity feels like uncertainty and is actually an unwritten document. Firms that fix prices get much better at scoping quickly, because there is a direct financial consequence to being vague.
When we estimate badly, we absorb the difference
A model like this is only credible if you say what happens when it fails.
We have underestimated. The recent one that stings involved an integration with a third-party system whose documented behaviour and actual behaviour diverged substantially, in ways that only appeared under production data volumes. It took roughly 60 percent more effort than quoted.
The client paid the quoted price. We absorbed the difference, which came out of that engagement’s margin entirely and then some. Nobody raised a change request, because it was not a change; it was us being wrong about difficulty.
Two things came out of it.
- A pre-contract integration probe on anything involving an unfamiliar third-party system, which is now standard and costs a day.
- A clearer internal rule about when to price a discovery increment separately rather than folding it into a fixed scope.
That is what the model costs when it goes wrong, and it is the mechanism by which estimates improve. A firm billing by the hour would have invoiced the extra 60 percent and learned nothing.
Within five years, hourly billing will be unusual for defined software work
Our expectation is that time-based billing for well-defined software work becomes unusual within five years, for reasons that have nothing to do with anybody’s principles.
The cost base moved. Buyers can now estimate independently, because the same tools that made delivery faster made estimation more accessible. A unit of measurement that both parties know overstates the work cannot survive that transparency indefinitely.
What survives is time-based billing where uncertainty is genuine and shared: research, discovery, and work where the client is the one who cannot specify. That is a real and permanent market, and considerably smaller than the current one.
Outcome pricing changes how the firm runs
The effects on how we operate turned out to be larger than the effects on what we charge, and mostly unanticipated.
Utilisation stopped being a metric. When revenue is per outcome rather than per hour, measuring whether people are busy optimises against automating the work that makes them busy. We track whether scopes ship and whether clients would sign again, and nothing else about individual time.
Estimating became a senior discipline rather than an administrative step. Somebody owns the number and is measured on it over a portfolio, not on any single quote. That role attracts a different kind of person than a project manager filling in a template.
Adopting new tooling became uncomplicated. Any firm billing by the hour has to reconcile efficiency gains with revenue loss, and the reconciliation is always awkward. We do not have that conversation, which means we can adopt anything that works the week it works.
Buyers can move toward fixed outcomes in stages
Practical steps, in order of how much they change.
- Ask for the comparison.
- Insist on a definition of done you can check yourself.
- Start with something bounded.
- Watch what happens when something goes wrong.
Any supplier proposing a fixed price should tell you what conventional delivery would cost and take. If they will not, the fixed price is a number without a reference point.
The definition of done should be a list of outcomes instead of a list of activities. If you cannot verify it without the supplier’s help, it falls short of a specification.
A discrete piece of work with a clear finish line, priced fixed, tells you more about a supplier in six weeks than a reference call will.
The behaviour of a supplier who has underestimated is the most informative thing you will observe. A firm that absorbs it and says so is telling you the model is real.
Our proposals price the finished thing
Every proposal carries a written scope, a fixed price, and the conventional comparison alongside. Payment is on milestones you can verify rather than on elapsed time. Change is priced as an increment at the point it is agreed. If we cannot beat the conventional number meaningfully, we say so and decline, which we do regularly enough that it is a real constraint rather than a slogan.
We would rather state this clearly than have it discovered later: this makes us worse at some things. We are slower to start, because the scope has to exist. We are more resistant to mid-flight changes than a time-and-materials supplier. And we will occasionally decline work that another firm would happily take.
Those are the costs of the arrangement, and they are the same costs that make the price mean something. The full structure is on engagement models, and the arithmetic behind the comparison figure is in our piece on how we price work the model mostly does.
