Specific beats glowing

The review that gets you the next booking is not the warmest one, it is the one that tells a stranger what they would receive. "Great, thanks" is a five-star review that persuades nobody, and "took longer than I expected but the written breakdown covered every question I asked and then some" is a three-star review that sells the job.

This matters because you have some control over which kind you get, and almost none of that control is in how you ask.

What the next buyer is doing when they read

They are not evaluating your quality. They cannot evaluate your quality, because they have not seen the work and would not know how to grade it if they had.

They are estimating risk. Specifically: will this person deliver, will it be what I asked for, and will it be awkward. Every review they read is being mined for evidence on those three, and nothing else in it registers.

That reframes the whole thing. A review saying you are wonderful answers none of the three questions. A review saying you asked a clarifying question, delivered a day early, and covered four named things answers all three, and it does so in a way that cannot be faked convincingly, which is why readers weight it heavily.

Buyers describe this in almost the same terms from the other side, and the commissioning-side account of what people write and why is worth reading once as a corrective, because earners systematically assume warmth is the currency when detail is.

The four things a persuasive review contains

Not all four, but the useful ones contain at least two.

A named deliverable. "The scored assessment", "the ten-minute audio", "the written report". This tells the reader what they are buying, which your listing already said and which they did not believe until someone else said it.

A specific observation the earner made. Anything that could only have come from the work actually being done. This is the single strongest signal available, because it proves attention in a way no adjective can.

A stated turnaround. Objective, checkable, and the one dimension where being good is unambiguous. It shows up disproportionately in reviews for exactly that reason, which is most of the argument for delivering early.

Something mildly negative. Counterintuitively load-bearing. A review with a small qualification reads as written by a real person who thought about it, and it makes the positive half credible. Reviews that are unbroken praise are discounted by readers, consciously or not. Northwestern's Spiegel Research Center found in 2017 that purchase likelihood typically peaks at ratings between 4.0 and 4.7, then falls as ratings approach 5.0.

Reviews are produced by deliveries, not by requests

This is the part earners get wrong for months.

The wording of your review request has a small effect on whether a review appears and essentially no effect on what it says. Paying for a better one is off the table anyway: the FTC's final rule on reviews and testimonials, announced in August 2024, bars incentives "conditioned on the writing of consumer reviews expressing a particular sentiment". What it says is determined by what the buyer has in front of them when they write it, which is your deliverable and your handover message.

So the levers are earlier than you think.

Structure the deliverable so it can be described. A buyer who received three headed sections can say "covered X, Y and Z". A buyer who received four paragraphs of prose can only say "very thorough". Defined axes hand the reviewer their vocabulary, and that is a real reason to have them beyond the speed gain in a reusable rubric.

Put at least one genuinely specific observation in every job. Something noticed rather than assessed. This is the sentence that ends up quoted, and it is also the thing that distinguishes you from an automated score. What a model already produces in two seconds is the generic layer, and a review praising you for the generic layer is a review arguing you are replaceable.

Say what you covered in the handover. One sentence listing the parts. Buyers reviewing a job three days later are reconstructing it from memory, and the handover message is the memory. The handover is worth writing deliberately for this reason alone.

Do not pad. Volume reads as thoroughness for about a day and as filler thereafter, and a reviewer who felt filled is not going to name anything, because nothing was nameable. The tension between being specific and being long is real, and it resolves toward specific every time.

Ratings and text do different jobs

The star rating is a filter, the text is the persuasion.

A buyer scanning a list uses the number to eliminate, then reads text on the two or three that survive. This means an average slightly below a competitor's costs you less than most earners fear, provided the text underneath is doing work, and it is why the early reviews matter so much more than the later ones.

It also means chasing the number is the wrong optimisation. Delivering a smaller, safer job to protect an average produces a wall of unremarkable fives that says nothing, and a listing that says nothing loses to one with a described three.

There is a broader point here about what any single score can carry, and the people who work with rating data professionally are consistently more cautious about it than the people reading it; the notes on what aggregate numbers do and do not support apply as directly to your star average as to anything else.

What to do with a review that is warm and empty

Nothing. Do not reply asking for more, do not reply at all beyond a brief thanks if the platform surfaces replies.

Empty reviews are not a problem to fix, they are a symptom, and the fix applies to the next job rather than the last one. Change what you deliver and what you say when you deliver it, and the texture of the reviews changes within about five jobs.

If a buyer's actual question was what a rating means about them rather than about you, that is answered directly on Rate Cock, and pointing there is a better use of the exchange than fishing for a better write-up.