Time tracking for freelancers: what to track, what to bill, and how to keep it honest

Employees track time because someone asks them to. Freelancers track time because it is the number the invoice is built from, which makes it the difference between a good year and a confusing one.

That raises a question employees never have to answer: what actually counts?

Billable and non-billable are both worth tracking

The obvious split is work you charge for and work you do not. Client deliverables are billable. Your own bookkeeping, proposals, tax admin and marketing are not.

The mistake is concluding that non-billable time should therefore go unrecorded. It is the number that tells you your real hourly rate. If you bill twenty-two hours in a week and worked thirty-eight, your effective rate is not what your contract says it is. Freelancers who feel underpaid at a rate they negotiated themselves are usually looking at the wrong denominator.

Track both. Bill one.

Per-client separation is the whole point

A weekly total is nearly useless. Hours per client per day are what you need, for three separate reasons.

Invoicing. Obvious, and the one everybody thinks of.

Scope. The client you quoted forty hours for and have given sixty-five is not a billing problem, it is a conversation you should have had three weeks ago. You cannot have it without the number.

Choosing work. After two or three months of honest data you know which client pays well per hour actually worked, as opposed to which one pays well per hour quoted. Those are frequently not the same client, and the gap is usually invisible without measurement.

This is why classification matters more than precision. Being ten minutes off on a day is fine. Filing Acme’s hours under Northwind is not.

The genuinely ambiguous work

Some categories resist clean assignment, and pretending otherwise is how timesheets become fiction.

Email and Slack

An inbox is not a client. A thread is. The practical rule most freelancers converge on: substantive client communication is billable — reviewing requirements, answering technical questions, writing a status update someone asked for. Reading a newsletter in the same window is not.

Automatic tracking helps less here than anywhere else, because the window title says “Slack” for both. Two things reduce the damage: a workspace or channel per client, if you can get it, and treating communication as a category you review rather than one you assume.

Context switching

Getting back into a codebase after an interruption takes real minutes. They are billable — they are a cost of the work — but nobody itemises them, and they are exactly the minutes that vanish from a reconstructed timesheet.

Thinking away from the screen

The walk where you solve the problem is real work and no tool will see it. If it is a routine part of how you work, decide on a convention with yourself and apply it consistently. Consistency matters more than the specific answer.

The memory test

Try this once. Before opening any tool, write down what you think you worked on last Tuesday and for how long. Then look at the record.

Almost everyone discovers two things: the day was more fragmented than they remembered, and the total was lower than they would have invoiced. That gap is not a character flaw. Memory compresses. It keeps the story of the day — “I worked on Acme all Tuesday” — and discards the forty minutes on someone else’s bug report, the call, and the hour of admin.

Any system that depends on you passing the memory test will fail, which is the argument for recording as you go. Whether you do that with a disciplined timer or with automatic tracking is a preference. Doing it from memory is not a third option; it is just guessing with extra steps.

A weekly routine that takes ten minutes

1. Friday, review. Open the week per client. Look for anything obviously misfiled and anything you cannot explain.
2. Sort the ambiguous pile. Communication and unassigned time, decided while the week is still recent enough to remember.
3. Compare to the quote. Hours spent against hours estimated, per active project. This is where scope creep is caught early enough to be a conversation instead of a write-off.
4. Export the billing range. Straight into the invoice.

Nothing here requires a daily habit, which is the only reason it survives a busy month.

Choosing a tool

Three questions decide most of it.

Automatic or timer-based? Timers are accurate when you are disciplined and interrupted rarely. Automatic tracking is accurate when you are neither. Be realistic about which describes your last six months rather than which describes the freelancer you intend to become.

Local or cloud? A time tracker reads window titles and URLs, so it reads client names and project codenames. Cloud sync buys you continuity across machines and costs you control over that data. If you have signed an NDA, that trade-off deserves a real decision rather than a default.

One-time or subscription? A subscription is rent for a service that runs on someone else’s machine. If the software runs entirely on yours, there is a reasonable argument that it should be bought once — which is how Tracker is priced after early access.

Tracker is one answer to those three questions: automatic, local-only, one-time. It is not the only defensible set of answers, but it is worth knowing which ones you picked and why, instead of finding out at renewal.