Estimating from records, not imagination

Everyone is wrong in the same direction

Ask any freelancer how long a job took against what they thought, and the answer is almost always “longer”. Not sometimes longer: systematically longer, by a factor that stays roughly constant across a career. Experience makes the estimates faster to produce and does not make them much more accurate.

This is not a personal failing. It is a well-documented bias in how people forecast tasks, and it has a mechanism: when you estimate, you imagine the work going as intended. You picture the drawing, the layout, the export. You do not picture the two days waiting for the client’s copy, the file that would not open, the brief changing on Thursday, or the morning lost to a tax form. Those things happen on every project and are absent from every mental simulation of one.

The correction is not to try harder to imagine them. It is to stop estimating from imagination and start estimating from records.

Estimating from the outside

The reliable method is to ask: how long did the last three jobs like this actually take?

That single question outperforms any amount of careful task breakdown, because your past projects already contain all the interruptions, delays and repeated exports that your imagination omits. They happened, and they are in the number.

This requires a record, which almost nobody keeps, which is why almost everybody keeps estimating from imagination. The record needed is small: for every project, the type of work, what you quoted, and the hours it actually took. One row in a spreadsheet, filled in at handover, alongside the portfolio images and the testimonial ask.

After six or eight projects you have something that no amount of thinking produces: your own multiplier. Most people find theirs is somewhere between 1.5 and 2.5. Once you know it, estimating becomes arithmetic: imagine the work honestly, then multiply by your number.

Tracking hours without misery

You do not need to bill hourly to track them, and tracking them is worth it even on fixed-price work, because it is the only way to know your effective rate.

The lightest version that works: a note of start and stop times in a text file or a notebook, totalled at the end of the week. Free timer tools exist (Toggl’s free tier, Clockify, or a plain stopwatch) and the software barely matters. What matters is capturing the whole project, including the admin: the emails, the calls, the quote you wrote, the invoice, the file delivery, the trip to the printer.

That last point is where most surprises live. Designers habitually count only the hours with the software open. A project with twelve hours of drawing and nine hours of everything else has an effective rate half of what the designer believes, and they will price the next one identically.

The estimate you give the client

Two useful practices.

Quote a range, or a date with a condition. “Three to four weeks, assuming the photographs arrive by the 14th.” The condition is what makes the date real, and it names the dependency in the same breath.

Add contingency before you say the number, not after. Deciding privately that a job is fifteen hours and then telling the client twenty is not padding; it is including the hours you know from experience will exist and cannot itemise. What does not work is quoting fifteen and absorbing the other five, which is the default behaviour and is where a rate quietly halves.

The multipliers that are not your fault

Three things reliably extend a project and none of them are estimation errors:

Client response time. A one-week revision cycle with a client who replies in five days is not a one-week cycle. Calendar time and working time are different quantities and should be quoted separately: “about twenty hours of work, spread over four weeks”.

Number of stakeholders. Each additional person with an opinion adds coordination, contradiction and delay. A project with one decision-maker and the same project with a committee of five are not the same job, and they should not carry the same price.

Unfamiliar production. The first time you prepare artwork for embroidery, screen printing on fabric, a new platform’s specification or a format you have not used, budget extra and expect to learn something the hard way. The second one costs a third of the first.

When you are already over

It happens on a fixed price. Three options, in descending order of grace.

If it was your estimate that was wrong, absorb it, deliver, and record the real number so the next quote is correct. This is the cost of estimating, and asking a client to pay for your mistake damages more than it recovers.

If the scope grew, that is a different thing and there is a mechanism for it: a priced change note at the time, not a complaint at the end.

If it is going to be substantially late, say so early, with a new date and the reason, without waiting to see if you can catch up. Nearly every client can handle a two-week delay announced three weeks ahead. Almost none can handle it announced on the deadline, because by then they have made commitments to other people based on your silence.

The estimate you should refuse to give

When a client asks for a figure on the phone, before any scope exists, the honest answer is a range wide enough to be useless, or a refusal with a reason.

“It depends on how many variants and how many languages: for something like this it usually sits between X and Y. Let me send a proper quote tomorrow once I know which.”

A number said casually becomes the number they remember, and every subsequent quote is measured against it. This is one of the few places where saying less protects you.

The limitation

Better estimating does not make projects shorter, and it will not make you accurate. Even with good records, individual projects vary enormously: one client’s decisiveness against another’s, one printer’s competence against another’s.

What records give you is accuracy on average, which is the only kind that matters for a business. You will still lose money on individual jobs and make it back on others. The freelancer who tracks is not the one who never misjudges; it is the one whose annual income matches what they thought they were charging.