Starting out
Completion and responsiveness beat everything you write
Marketplace ranking rewards behaviour, not copy, because behaviour is what it can measure.
Guides on Starting out: A rubric is a production tool before it is a quality claim, The brief is rarely the request, Getting the first booking, in order
Marketplace ranking is not a judgement about quality, and it is not influenced by how well your listing is written. It is a prediction about whether a buyer sent to you will complete a paid transaction and not complain afterwards, built from the handful of things the platform can measure directly.
Once you see it as a prediction rather than a reward, the list of things worth doing gets short and slightly boring.
What is being measured
No platform publishes its formula, but in the EU the Platform-to-Business Regulation requires marketplaces to explain the main ranking parameters in their terms - including, as Ireland's Competition and Consumer Protection Commission summarises it, whether payment to the platform can influence ranking. The platforms that describe their approach describe the same broad inputs.
Completion rate. The proportion of accepted jobs you finished. This is usually the heaviest input, because a cancelled job is the outcome the platform is most trying to avoid sending buyers into.
Response time. How long between an enquiry arriving and you replying. Measured in minutes, not hours, and typically as a median over a recent window.
Recency of activity. When you last completed work. This decays fast, which is why a break costs visibility - what happens if you stop for a month covers the recovery curve.
Review score and review volume. Both, weighted together, with volume mattering more than earners expect. A 5.0 across three jobs typically ranks below a 4.8 across sixty.
Cancellation and dispute history. A separate and heavier signal than completion, because a dispute costs the platform money and staff time.
Why copy does not move it
Search matching uses your text; ranking mostly does not.
The distinction is worth being precise about because it is where most effort gets wasted. Your words decide whether you are in the candidate set for a query. The behavioural signals decide where you sit within it. Rewriting your listing improves the first and does nothing to the second, and the second is what decides whether anyone reaches the screen you are on.
The routes buyers take to you in the first place are set out in where demand actually comes from, and search is only one of them.
The three things you actually control
Reply fast, consistently. This is the highest-leverage behaviour available to a new earner because it is the only signal you can max out on day one, before you have any record at all. Fast means minutes during the hours you are present, and the practical version that does not consume your life is in responding fast without living online.
Do not accept work you might not finish. A declined enquiry costs you nothing in ranking terms. An accepted job you cancel costs you a completion-rate hit that takes many jobs to average out. This is the entire argument for declining politely and quickly, and it is a ranking argument as much as a reputation one.
Stay active. Steady small volume outranks sporadic large volume, because recency decays and completion count accumulates. Five jobs a month for six months puts you well ahead of thirty jobs in one month and nothing after.
The two things you cannot control, and should stop trying to
Tenure. Accounts with years of history rank above you and will continue to, and no amount of optimisation closes that gap directly. The workable response is positional rather than competitive, which is what positioning against the people above you is about.
The algorithm changing. It does, without notice, and the visible effect is a week where bookings drop for no reason you can identify. Treat a single bad week as noise. Treat a bad month against a stable pipeline as a signal, and check whether anything about your own behaviour changed before concluding the platform did.
The cold-start exemption, and its expiry
Most marketplaces give new listings a temporary boost, on the reasoning that they need data to rank you and the only way to get it is to send you traffic.
This is the single most misread feature of any marketplace. Early bookings arrive more easily than they will in month three, and the drop that follows is not a decline in your performance - it is the exemption ending.
Two consequences. Use the window: it is the cheapest traffic you will ever get and the point of it is to convert into completed jobs and reviews. The cold start problem covers how to spend it. And do not benchmark month three against month one, because they are measuring different things.
Volume compounds, which is the whole strategy
Completed jobs improve ranking, ranking produces enquiries, enquiries let you decline the bad ones, declining the bad ones protects completion rate and review score, which improves ranking.
The loop is slow to start and difficult to stop, and it is the reason volume sometimes beats price early on even when the per-hour arithmetic says otherwise. The corollary is that anything which breaks the loop - a cancelled job, a fortnight offline, a dispute - costs more than the single event suggests, because it costs the compounding too.
What ranking cannot see
It cannot see that your work is better than the person above you. It cannot see that a dispute was the buyer's fault. It cannot see a repeat buyer who came straight to your profile, which is why the most valuable demand you have contributes least to your position.
That last point is the useful one: a mature earner's income increasingly comes from routes ranking does not govern, and the anxiety about position should fade in proportion.
The buyer's own account of how much they rely on the ordering, versus reading three profiles and deciding on other grounds, is written up from the commissioning side and is less algorithmic than earners assume. Where ranking does sit on measured quality rather than behaviour, it uses whatever scoring the platform has, and the technical account of what automated scoring can and cannot establish is the honest boundary on that. The free tools buyers use before they arrive shape what a top-ranked human listing has to be worth, which the tooling overview makes concrete.
For how listing position is determined on the platform specifically, including what new judges are shown and for how long, Rate Cock's judges page is the current statement.