Falling conversion at a steady price is the first sign

By the time the visible prices in a category have fallen, the crowding happened months ago and you have been absorbing it. The early signal is in your own numbers, and it is specific: the same listing at the same price converting a smaller share of enquiries, or attracting fewer enquiries per week, with nothing on your side changed.

That is the whole detection method. Everything else in this piece is about not misreading it.

The four indicators, in the order they appear

Enquiry volume falls at a steady listing. You changed nothing and fewer people are asking. This is the earliest signal and the noisiest, which is why it needs several weeks before it means anything.

Conversion falls while enquiry volume holds. More people are asking and fewer are booking, which means they are comparing you against something and losing interest. This is the strongest early indicator, because it is a direct measurement of your listing's relative position rather than of the market's size.

Price questions arrive earlier in the conversation. When "can you do it cheaper" moves from the third message to the first, buyers have a comparison set and are working through it.

The visible band narrows. New listings cluster tightly around a figure. This is the last thing to happen and the one everyone notices first, because it is the only one visible without records.

You cannot see any of the first three without having written them down. Enquiries that did not convert are the most commonly discarded data in this trade and the most useful - tracking the ones that got away is a two-column habit that makes the difference between noticing this in month two and noticing it in month eight.

What it is not

Three things look identical to crowding in a small sample.

Seasonality. Demand has an annual shape, and a quiet stretch that recurs at the same point each year is a season, not a market change. Compare against the same weeks last year if you have them, and if you do not, this is the reason to start keeping them.

A ranking change. If a platform changed how listings are sorted, your enquiry volume moves without anything happening to the category at all. Check whether anything landed in your inbox before concluding anything about competitors.

Your own drift. Response times slipping, a stale listing, an unrepaired gap after a break. These produce the same curve and are entirely within your control. Treating a quiet month as data rather than a verdict covers how to tell these apart before making a decision you cannot walk back.

What crowding actually does to prices

Not what people expect. Categories under pressure rarely see a general collapse; they see the bottom fall out and the top hold, so the band widens downward and the median drops while the upper quartile does not move.

This matters because it determines your options. If the top of the band is intact, the problem is that you are in the part of the category that crowded, not that the category is finished.

The generic low end of assessment work is the clearest example: it did not crowd, it was replaced, and the replacement is instant and free. The same shift has been measured in freelance writing: after ChatGPT's release, Hui, Reshef and Zhou's study of a large online freelance marketplace found writing freelancers' monthly jobs fell 2% and monthly earnings 5.2%, with higher-quality workers hit harder rather than spared. The technical account of what automated scoring produces is the description of what happened to that segment, and it is a permanent change rather than a cycle.

The three options

Differentiate. Change the format so the comparison stops being like for like: a different length, a scored structure, a recorded component, a stated turnaround guarantee. This is the cheapest option and the one to try first, because it costs a rewrite rather than a repositioning.

Reposition upward. Move out of the crowded segment entirely by raising your floor and removing your entry tier. This works when the upper part of the band is intact and you have enough evidence to be credible there. It reduces volume deliberately and improves take-home per hour, and it is a real strategy rather than a consolation.

Move. Add a format or a platform where the same skills apply and the crowding has not happened. Slowest and most expensive, and correct when the category has genuinely closed rather than merely thinned.

Most people should do the first, consider the second, and reach for the third far later than instinct suggests. Adding a second format is the structured version of the first two, and it is worth doing before you are forced to rather than during.

Decide on a horizon, then stop looking

The failure mode after noticing crowding is checking competitors' listings weekly, which produces anxiety and no information, because week-to-week variation in a small category is entirely noise.

Set a horizon - a quarter is about right - make one change, and leave it alone until the horizon is up. Resample the band properly at the end of it rather than continuously, and keep the sheets so you have a series rather than a snapshot.

Precision-based differentiation is one of the few positions that does not crowd easily, because it can be checked and most listings will not survive being checked; the conventions to hold yourself to are documented on the measurement property. And the tools buyers reach for before they consider a person are catalogued here, which is the fastest way to see what your entry tier is actually competing against.

For what the current field of listings looks like on the platform side, Rate Cock's judges directory is one useful sample - though a category can look crowded on one marketplace and empty on the next, which is itself the finding worth having.