Keep five things from every job

Nobody starts this work planning to keep records, and nearly everybody wishes they had, usually on the day a payment reverses or a tax deadline arrives.

The good news is that the useful set is small and takes about a minute per job.

The five things

The brief, as agreed. Not your memory of it - the actual message where scope was confirmed. This is the document that settles a dispute about what was owed, and it is the one people most often cannot produce.

What you delivered, and when. A copy of the work, or at minimum a record that it was sent and the timestamp. Delivery evidence is what turns "they never sent it" into a resolved case.

What you were paid, gross. The headline figure the buyer paid, before anybody took anything.

What you actually received, net. After commission, processing, currency and withdrawal. These two numbers being different by more than people expect is the subject of the post on effective rates, and you cannot manage the gap without recording it.

Time spent, including the unbilled parts. Reading, clarifying, delivering, chasing. This is the number that tells you whether a tier on your rate card is worth keeping, and it is invisible unless written down.

The minimal system

A spreadsheet with one row per job and those five columns.

That is the whole recommendation. Not software, not an accounting package, not a system with categories - a sheet you can fill in during the minute after you deliver, because a system that takes longer than that will be abandoned in week three and the gap will be invisible until you need it.

Add a sixth column for the platform if you work on more than one, and a seventh for the date the money became actually withdrawable rather than merely earned. That second date is the one that matters for holds.

Keep copies of delivered work somewhere durable and organised by date. Platform message histories are not a record you control: accounts get closed, disputes get archived, and services shut down.

Why disputes need this

A chargeback or a dispute can arrive months after a job, at which point the entire question is what was agreed and what was delivered.

The party with a timestamped brief and a timestamped delivery generally wins. The party reconstructing events from memory generally does not, regardless of who was actually right.

This is also the argument for confirming scope in writing before starting, even when the buyer is obviously fine and the job is obviously small. It costs twenty seconds and it is the entire basis of your case in the one case out of fifty where you need one. The post on getting paid covers how those cases actually unfold.

Why tax needs this

Rules differ by country and change, so this is orientation rather than advice, and the specific answer for your situation belongs to someone qualified in your jurisdiction.

The orientation is: income from this work is generally taxable income, most places let you deduct genuine costs of earning it, and both halves of that require records that exist before the deadline rather than after.

Costs worth recording as you go include platform commission, processing and payout fees, equipment used for the work, and any subscription that exists because of it. The commission in particular catches people out - it is usually deducted before the money reaches you, which makes it easy to forget that you earned the gross figure and paid a cost, rather than simply earning the net one.

Set aside a percentage of every payment as it arrives rather than facing an annual reckoning. The percentage will be wrong; the habit will not be.

What the records tell you besides compliance

After thirty rows the sheet starts answering questions you did not build it for.

Which format actually earns the most per hour, once the unbilled minutes are counted. Whether your entry tier is profitable at all. Whether the expensive-looking jobs are worth their scope. Whether your time per job is falling, which is the single best indicator that this is going to work.

That last one is why the reasons two earners charge very different rates can be diagnosed on your own numbers rather than guessed at, and why the choice of pricing unit stops being theoretical once you can see what each format actually returns.

The part that is genuinely optional

Elaborate categorisation, monthly reports, and any tooling more complicated than a sheet.

Add those if the volume ever justifies them. For most people it does not, and the elaborate version is mostly a way of feeling organised while recording less.

Two outside notes. Where your assessments cite a measurement, record which convention you used along with the job - the conventions differ, and a dispute about a number is unwinnable if you cannot say which method produced it. Where a job was scored to a rubric, keep the rubric version too; how scored assessments are constructed changes over time, including yours, and "this is what my five-axis score meant in August" is occasionally the whole answer. If any part of your workflow passes material through an automated tool, know what that tool retains - the account of how uploads are handled is the general version, and your obligation to the buyer is specific.

Platform-side records for the service this site is published by - payout history, job history, the dates that matter for holds - are on Rate Cock's rewards page, and exporting them periodically is a better habit than trusting they will always be there.