As a marketplace grows, moderation quietly becomes a second job. Someone has to notice the fraudulent listing, the miscategorised space, the description that oversells what's actually available. In the early days that someone is a founder, reviewing listings at 11pm because nobody else will.
This works right up until it doesn't. Listing volume grows linearly; a founder's attention doesn't. And unlike most scaling problems, this one fails silently, you don't get an error when a scam listing sits live for three weeks. You get a chargeback, a one-star review, and a trust problem you can't advertise your way out of.
The fix isn't more hours, and it isn't necessarily a hire. It's accepting a simple truth: your users see problems before you do. They're on the listing pages every day, they know the market. What they lack is a structured way to tell you.
On a marketplace I worked on, any signed-in user can report a listing. The flow is deliberately constrained: a fixed set of reasons - inappropriate content, fraudulent listing, inaccurate information, spam, other - plus an optional free-text note. Reports land in an admin queue with the listing, the reporter, and the reason attached.
The constraint matters. Free-text-only reporting produces vague complaints that take longer to triage than to investigate from scratch. A reason taxonomy means the queue can be sorted, prioritised, and measured. Fraud reports get looked at today; "inaccurate information" can wait for the weekly pass.
The effect on operations is a change of kind, not just degree. Manual patrolling, reading everything, mostly finding nothing, becomes triage: reviewing a queue where every item was flagged by a human with a reason. The team stops hunting and starts deciding.
Here's the most important part of building a reporting system, because it's the one most teams skip: reports are rate-limited. Three per user, per hour.
The moment you build a reporting system, you have also built a harassment tool. A competitor can bury a rival's listing under bogus reports. A disgruntled user can flood your review queue until real signals drown. Unthrottled community moderation doesn't solve your abuse problem, it relocates it, one layer up, where you're less likely to be watching.
The rate limit is one line of configuration. What it represents is a habit worth building into every trust feature you ship: before launch, ask what this looks like in the hands of someone acting in bad faith. Reporting, reviews, flagging, referral schemes - every mechanism designed for good-faith participation has a bad-faith use waiting. The good ones are designed with that answer already built in.