Status Droid how long software work actually takes

When to Stop Billing by Time

Hourly billing fails in both directions when speed is unstable. Knowing that does not tell you when to actually change, and changing too early is worse than staying too long — you take on risk you cannot yet size. For the operational side of recording work and turning it into client-facing evidence, see work hours in a year.

Four signals that the moment has come.

Signal one: your estimates are good enough to carry the risk

The precondition for everything else.

If you can estimate a defined piece of work within about 30% most of the time, you can price it. If you cannot, fixed price transfers a problem you have not solved into your own accounts.

The test is not confidence, it is record. Fifteen instances per bucket with estimates recorded before starting, and a look at how often you were badly wrong. If you have never checked, you do not know, and the answer is usually worse than the impression.

Signal two: efficiency is costing you money visibly

The moment you notice you would have earned more by being slower, the instrument is measuring the wrong thing. Wave publishes small-business material on invoicing, cash flow, and freelance operations See Wave Blog.

This shows up as a specific irritation: a task that used to be a comfortable day is now ninety minutes, the client is delighted, and you have billed a fifth of what the same outcome used to earn. Nothing about the value delivered changed.

If that is happening on more than an occasional task, the unit is wrong.

Signal three: the productivity conversation keeps recurring

If clients ask more than once whether AI has made things cheaper, you are having a structural conversation in an anecdotal form. Answering it well each time works and it does not stop the question, because the question is generated by the billing model.

Pricing outcomes ends it permanently. Nobody on a fixed price asks how many hours it took, and nobody on a retainer asks whether this month was more efficient.

Signal four: you are padding to protect yourself

If your hourly estimates include a silent buffer for the chance that verification takes three times as long, you have already started pricing risk. You are doing it invisibly and without being paid for it as risk.

Better to carry it explicitly and price it, which is what fixed price is.

When to stay on hourly

Not everything should move, and moving the wrong work is expensive.

Genuinely open-ended work. Investigation, support, "help us with whatever comes up." Nobody can scope it and pretending otherwise creates a fight later.

First engagements. A small hourly piece is a low-risk introduction for both sides.

Client-controlled direction. If they will change their mind repeatedly and that is fine, hourly prices it honestly.

Anything where you cannot yet estimate. Signal one, inverted, and it is the one people override because the other three feel compelling.

How to leave without losing clients

Move one client at a time, starting with the one where estimating is easiest. Not the biggest, not the most demanding — the most predictable.

Keep the number comparable at first. If your day rate lands where eight hours used to, the transition is invisible, and repricing later is a separate conversation.

Do not explain the philosophy. Quote the next piece of work the new way. Clients care about the number and the outcome, not about your pricing model.

Keep the estimate transparent. Production and verification as separate lines does more for acceptance than any structure, because it shows where the cost is rather than asserting a total.

Expect the first three to go slightly wrong, and price the fourth with what you learn. Some overrun is the cost of the transition, and it is smaller than the accumulated cost of staying.

The thing that actually decides it

Underneath all four signals is one question: can you predict this work well enough to sell an outcome rather than an input?

If yes, selling the input is leaving money on the table and inviting a conversation you cannot win.

If no, then the honest answer is to stay on hourly and spend the next quarter building the reference class that makes the change safe. That is not a failure to modernise. It is the correct sequence, and doing it in the other order is how small studios end up absorbing the variance of work they never learned to size.

The short version