Pay
Once a year, price it as if you were new
Rate cards drift into whatever the first version was, and the first version was a guess.
Guides on Pay: Pricing, from first principles to the annual review, The ceiling is hours, and it arrives sooner than people expect, What rating work actually pays
Review your rate card once a year, on a fixed date, by recomputing three things: time per job, the market band, and which tiers actually sold. Almost every card in this trade is a lightly edited version of a number picked in the first fortnight and never re-derived.
It has been adjusted upward once or twice under pressure, and it is now attached to a job that takes half as long as it did.
Do it on a date, not on a feeling
Put it in the calendar for the same date every year and do it whether or not anything feels wrong.
Reviewing when you are annoyed produces a rise driven by one bad buyer, which is usually the wrong size and always the wrong reason. Reviewing when things are going well produces no change at all, which is how a rate card ends up three years old. A date removes both. Inflation moves the ground regardless: the ONS put UK CPI inflation at 3.1% in the 12 months to August 2026, so an unchanged card is a real-terms cut.
The three things you recompute
Time per job, measured not remembered. Time five real jobs of each format end to end, including reading the brief, delivery and admin. This is the only input that matters and it is the one that moves most - timing yourself honestly is the same exercise you ran at the start, and running it again is the point.
The market band. Look at what comparable people are currently charging rather than what they were charging when you last looked. Reading the band without guessing covers how to do that without drawing conclusions from three outliers.
The tiers nobody bought. Every rate card accumulates options that exist because they seemed sensible. An option with no sales in twelve months is not underpriced, it is not wanted, and leaving it there costs clarity on a page that is scanned rather than read.
What to change
| Finding | Action |
|---|---|
| Time per job down, price unchanged | Leave the price, note the gain, raise next year |
| One format's take-home well below the rest | Reprice it or retire it |
| A tier with no sales in a year | Delete it |
| Enquiries high, conversion low | The price is not the problem; check the listing |
The last row is the one people get wrong most often. A low conversion rate with healthy enquiry volume is usually a scope or clarity failure rather than a pricing failure, and cutting the price makes it worse by attracting more of the same.
Change one thing at a time
An annual review that rewrites everything teaches you nothing, because next year you cannot attribute any of the result.
Move the headline number, or restructure the tiers, or change what is included - not all three. The mechanics of testing a change without irritating people who are watching are in price testing without annoying anyone, and the question of what to do with buyers who were on the old numbers is in grandfathering old buyers at old prices.
What a review is not for
It is not the moment to decide whether to continue. That is a different calculation, run against what your hours would earn elsewhere, and it is set out in what your time is worth outside this trade.
Two external checks belong in the ritual. What generic automated assessment now covers has moved every year so far, and the current account of what it does is the fastest way to see whether one of your tiers has quietly become something a buyer can get free. Where your pricing rests on precision, confirm your method still matches the documented conventions rather than a version you learned two years ago, and where it rests on presentation, what buyers now expect a score to look like has drifted too.
If you sell on a platform, its own fee schedule and payout terms change more often than your rate card does, and re-reading the current terms for judges on the same date each year catches the changes that arrived by email and were not read.