Pay
The right comparison is hourly and honest
Flexibility is worth something, and so is not having a manager, but neither pays rent.
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
The only comparison worth making between flexible online jobs is take-home per hour worked, including the unpaid hours. Everything else - autonomy, working from a sofa, no manager - is real but is not currency, and it gets used to excuse bad rates far more often than it gets counted honestly.
Rating work does well on some of these axes and badly on others, and which ones matter depends entirely on why you are doing it.
The four axes
Effective hourly rate, meaning fees received divided by every hour spent, including enquiries that went nowhere and admin. Not the advertised rate. The gap is measurable: the ILO's 2018 survey of 3,500 crowdworkers found they spent 20 minutes on unpaid activities for every hour of paid work, which cut average earnings from US$4.43 an hour to US$3.31.
Variance, meaning how different a good month is from a bad one. High variance is survivable with savings and brutal without them.
Ceiling, meaning the realistic annual maximum. Most flexible online work has a low one because it is priced per unit of your time.
Stability, meaning how likely the work is to still exist, at a similar rate, in two years. This is the axis people ignore and then get hit by.
Where rating work actually sits
Against the broad field of flexible online income, the pattern is consistent.
| Axis | Rating work | Typical microtask or survey work | Freelance skilled services |
|---|---|---|---|
| Effective hourly | mid, wide spread | low, narrow | high, very wide |
| Variance | high | low | high |
| Ceiling | low to mid | very low | high |
| Stability | uncertain | poor and worsening | mixed by discipline |
| Unpaid overhead | moderate | low | high |
Read across the rating column and the shape is a mid-rate, high-variance job with a modest ceiling and meaningful uncertainty. That is a good fit for supplementary income and a poor fit for anything you must be able to forecast.
The variance is the part newcomers underrate. A month can be double or half the one before it for reasons that have nothing to do with you, which is why treating a quiet month as data rather than a verdict is a practical skill rather than a consoling attitude.
What rating work is genuinely better at
Three things, and they are not small.
Low startup cost. No equipment, no course, no credential. The barrier is a profile and a fortnight of patience, which is a lower barrier than almost any other online work with a comparable rate.
Time granularity. Jobs are short. You can do one in a gap, and you cannot do that with anything that requires a project's worth of context.
Price control. You set the number. Most flexible online work hands you a rate and the only lever is volume - which is the whole reason a ceiling exists there and a partial one here. The reasoning behind actually using that control is in why two earners charge very different rates.
What it is worse at
Predictability. Almost anything with a shift pattern beats it here, and that includes ordinary part-time employment.
Compounding. Skilled freelancing gets better paid as you get better, because the buyer can tell. In this trade, experience mostly buys you speed rather than a higher rate, and speed has a limit.
Portability. Your reputation lives inside a marketplace and mostly does not travel. That is a real risk and the mitigations are limited, which is why not putting everything on one platform is a stability measure rather than a growth one.
The stability question, taken seriously
Every online task that a model can do has repriced downward, quickly, and there is no reason to think that process has finished.
The exposed part of this trade is generic assessment: a score with a short justification, produced from an image, to no particular standard. That is already free, instant and adequate for most casual purposes - the technical account of what automated scoring does describes exactly how far it reaches, and the tools packaging it are what a buyer tries before ever considering a person.
The unexposed part is anything where the value is that a specific named person actually looked, applied a stated method, and can be held to it. That is what buyers describe when they explain why they commissioned someone rather than running a tool - their own accounts of the reviews they received are mostly about specificity and attention rather than accuracy.
So the honest stability read is: the low end of this work is going away and the specific end is not. Which end you are at is a choice you make when you decide what to sell.
How to actually compare, for yourself
Run one month of your own numbers against one month of the alternative, on the same basis.
Total fees received, divided by total hours including admin, minus the tax set-aside. Do it for both, then adjust for variance by asking what your worst plausible month looks like in each.
If the alternative wins on the hourly and you still prefer this, that is fine - you are buying flexibility and you now know the price of it. Buying it knowingly is the whole point, and it is a different position from discovering the price two years later. What this work pays as a structure rather than a number is the input for your side of that calculation, and the platform-side figures you would reconcile it against are set out on Rate Cock's judges page.