Priya

One caveat on the counts: this is a keyword pass over review text, so accuracy is almost certainly undercounted. People do not type the word accuracy, they paste the sentence that came out wrong. The ordering of the top three is safe, the exact 12 is not.

One more from the accessibility pile worth carrying over: a visually impaired reviewer complaining there is no dark mode. In a category whose peak usage is a dark bedroom at three in the morning, that is not a preference, that is the use case.

Add the sold listings, not just the open ones. An asking price is a wish. The sold set is the only place a number somebody actually paid shows up, and it is what lets you turn a monthly profit into a rough valuation instead of guessing at one.

Four low star reviews out of 98 is also a sample size caveat. That is a rate you could get from luck. It is enough to say the category is not obviously broken, not enough to say it is loved.

The spread is the interesting part of that list, not the top. Two million at the top and eighty thousand at the bottom is a twenty five to one gap on identical input. Whatever explains that gap is the entire skill in this business, and it is not the data.

Both, surely. Being surprised by four pounds and being surprised by forty are not the same review.

Before. It is the raw sale price against declared annual profit, so treat it as a ceiling. Fees, escrow and whatever transition support you agree to all come out of that number afterwards.

Agreed on the motive. My complaint is what it does to everyone else's arithmetic. Those eight rows are the ones people screenshot, and they are the only ones nobody can check.

The margin figure is doing something subtle here as well. Zero cost means zero cost to the seller, who was not paying themselves. Add one person answering support and one platform fee and the profit line looks very different for the buyer.

Yes, and that is the right way to use it. It is a filter for whether you can get in, not for whether you should want to. You need both readings, and I was only doing one.

That is the right test and it moves me. The content of the complaints is about monetisation choices rather than about the product being free, which is a different thing.

The repost subreddit thing is worth generalising. Before you count anything from a community, count the authors. Three thousand posts by four hundred people and three thousand posts by nine people are different data sets with the same row count.

The self selection caveat is doing more work than you are giving it credit for. A paywall is a filter on who gets to be disappointed. Put the same PDF reader behind three pounds and its rating goes up without one line of code changing, because the people who would have rated it 1 never arrive.

Two of 195 on price is the finding here. Whole category, and the thing everyone assumes is the friction is barely mentioned. That means a paid authenticator is viable in a way the received wisdom says it is not, provided it earns the money on the recovery story.