Pay
Sample properly before concluding anything
Three profiles is not a market, and the three you looked at were the ones ranked highest.
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
Read the market band by sampling at least thirty listings across the whole ranking, inside one format, split by tenure. Almost every new earner instead glances at a few listings and picks something slightly below them, which produces a number with two known errors, both pushing the same way.
The listings you saw were selected by a ranking system that favours tenure and completed work, and you looked at too few of them to know anything about the spread.
The two biases in a casual glance
Ranking bias. The default sort on any marketplace is not random. It surfaces established sellers, which means the prices you saw are drawn from the upper part of the distribution and are attached to people with evidence you do not have. Concluding "the market is around this figure" from page one is reading the top decile as the average, and the first figures you see then act as an anchor - the adjustment heuristic that Tversky and Kahneman showed, in their 1974 paper in Science, leads to "systematic and predictable errors".
Small-sample bias. With five observations you cannot distinguish a band from noise. Prices in this trade are not tightly clustered, and a sample of five will regularly produce a range twice or half the real one purely by chance.
The two combine badly. You conclude the band is high, feel you cannot charge that, and set a price below a floor you have mismeasured in the first place.
A method that takes about an hour
Sample at least thirty listings, and get them from more than the first page. Move through the ranking deliberately - some from the top, some from the middle, some from wherever new listings appear - because the point is to observe the distribution, not the winners.
Record four fields per listing and nothing else: the price, the format, the stated turnaround, and a proxy for tenure such as completed job count or review count.
Then throw away the observations you cannot compare. A recorded video at ten minutes and a written note at 200 words are not the same product, and averaging them produces a number describing nothing. Compare inside a format or do not compare.
Discard the extremes at both ends before you conclude anything. The very cheapest listings are frequently inactive, mispriced or fronting something else, and the very highest are usually attached to a name doing something you are not. The bottom of any such sample is also distorted by the fact that generic assessment is now available instantly for nothing, so those listings are priced against zero rather than against each other - the account of what automated scoring covers explains which part of the band that pressure reaches. Trimming the top and bottom tenth of your sample gets you a band rather than a range, and a band is the thing you can actually price against.
What to do with the result
You now have three numbers per format: a lower quartile, a median, and an upper quartile. Read them as follows.
| Where you sit | What it says | When it is right |
|---|---|---|
| Below the lower quartile | Signals inexperience, selects for price-sensitive buyers | Almost never |
| Lower quartile to median | Accessible without being suspicious | First two or three months |
| Median to upper quartile | Normal for someone with visible evidence | Once you have reviews |
| Above the upper quartile | Requires a visible reason | When the format is genuinely different |
The important finding is usually not the median. It is the width of the band, because a wide band means buyers are not comparing on price alone and there is room to position; a narrow one means the product is legible as a commodity and your only levers are format and turnaround.
Why two earners with similar work charge very different rates explains what fills a wide band, and it is worth reading alongside your own sample rather than instead of it.
Control for tenure or the number is useless
The single most common error after sampling too little is comparing your price to the price of someone with 400 completed jobs.
Split your sample by the tenure proxy. The band among listings with few completed jobs is the band you are actually in this month, and it usually sits meaningfully below the headline one. Setting your entry price against the low-tenure band and your target price against the whole-market band gives you both a defensible opening number and a destination.
That destination matters more than the opening number, because your first rise is a scheduled event rather than a reaction - when to make it and by how much is a separate decision that should already be made before your first booking.
Resample on a schedule, not on a feeling
Prices move, and the direction is not always down. Do the exercise again every quarter, keep the sheet, and compare it to the last one.
A band that has narrowed and shifted down is crowding. A band that has widened is differentiation appearing, which is usually good news for anyone offering a distinct format. A band that has not moved while your own conversion has fallen is a problem with your listing rather than with the market.
Sampling is also how you notice things a headline price hides. Two listings at the same figure can imply very different take-home once fees, currency and payout mechanics are counted, which is why comparing platforms on take-home rather than commission is a different exercise from this one and needs doing separately.
Two boundaries worth naming
What buyers are weighing when they look at those same listings is the buyer-side subject, and the account of how someone chooses who to commission is the reference for it - the useful correction it offers is that price is often not the deciding variable at all, which changes how you read a band.
If part of your offer is objective measurement rather than judgement, the conventions it should follow are documented on the measurement property; a sample of prices tells you nothing about whether your method is sound, and buyers who care about precision check the method, not the price.
For the platform-side view of what is currently listed and at what, Rate Cock's judges directory is one of the samples worth including rather than the only one - a band drawn from a single marketplace describes that marketplace, not the trade.