I measured a leaderboard for three days. Then it fixed itself.
A week ago a leaderboard appeared with one rule: your rank is whatever you paid. No ads, no algorithm, no accounts. By day four it had taken $175,438 from 1,097 listings. By day eight, $241,173 from 2,134.
It is a genuinely clever piece of work and I want to say so plainly, because most of what follows sounds like criticism and is not meant as one. But outbid.lol publishes its whole database through a public API, including what every listing paid and how many clicks it received. That makes something possible which normally is not: you can check, precisely, whether people got what they paid for.
So I checked. Three times. The third reading is the one worth reading this far for.
01Seven buyers paid for half of it
The distribution is the story. The mean bid is many times the median, which is not a market: it is two populations sharing a page. A handful of whales, and a very long tail of five-dollar lottery tickets.
On 23 August the top seven listings were 52% of all revenue. A day later, 43%. The concentration eases as the board grows, exactly as you would expect, but the shape does not change. The median bid over those two days moved from $5 to $6, and roughly two thirds of every listing on the board paid $10 or less.
02I measured twice, and the answer changed
Every listing reports its click count, so cost per click is simple arithmetic. I ran it on 23 August. Then, because something looked wrong, I ran it again exactly 24 hours later.
The listing at the top did not get better at marketing overnight. It simply sat there another day. Its clicks went up 6.3x, its price went up 13%, and its cost per click fell by 82%. Across the whole board the median fell from $1.91 to $0.67 in those same 24 hours.
Cost per click here is a function of how long a listing has been up, not of what it paid or where it ranks.
You can watch it work in any single snapshot. On 24 August the two worst deals on the board were its two newest arrivals: #2 paid $16,000 for $2.36 a click, and #8 paid $4,000 for $11.63 a click, while #1 above them was getting $0.51. Not because they ranked lower. Because they bought in later.
03Why that is a problem, not a curiosity
Every ad product in the world sells you a quantity of exposure: impressions, clicks, a slot for a defined period. You know roughly what you are buying before you pay.
This board sells you a position, and exposure accrues to it afterward, for as long as nobody displaces you and the site keeps drawing a crowd. Neither is promised or knowable at the moment you pay. So a $16,000 bid might turn out to be $0.50 a click or $2.50 a click, and which one you get is decided by events after the transaction. That is not a price. It is a place in a queue whose length nobody has told you.
04The rule that should have ended it
Bids are permanent. Nothing expires. Nothing is refunded.The rules, as they read on 24 August
So the top only ratchets upward. Day one it cost a few dollars. By day five, taking #1 costs $17,005. Nobody arriving on day thirty outbids that, so the top freezes and the people who would have generated the next $17,000 quietly leave.
And there is no second sale. The whales have paid, and there is nothing to buy twice. outbid.lol has no accounts, so it has no email addresses, so it cannot even tell a buyer they have been outranked. A dollar auction gives you a spectacular first week. It does not give you a week eight.
05Then, on day seven, it fixed itself
outbid.lol now runs three boards instead of one.
And in their own words: “The same payment also adds to your all-time rank.” One payment, two effects. The frozen top is gone, because the top resets every day. The missing second sale is gone, because the daily board is the second sale.
I want to be careful about what this does and does not mean. I never contacted anyone, published anything, or influenced this in any way. I spent three days measuring a seven-day-old product, and it moved underneath me. Which makes the lesson bigger than this board:
A structural criticism of a young product has a shelf life measured in days. The operator is reading the same numbers you are.
06What is still true
Exposure still cannot be priced. A daily board changes how long you hold a position, not what a position is. You still buy rank and receive whatever exposure the following hours happen to deliver.
And there are still no accounts. No accounts means no email addresses, which means the site still cannot tell a buyer they have been outranked. That gap was awkward when the board never reset. On a board that resets every 24 hours, the only way to know you have lost the position you paid for is to remember to go and look.
And the fix did not stop the decline. On 24 August the board took $45,620 in a day. On 28 August, $5,128, with 33 new listings. The daily board that was meant to keep it moving held 319 listings on the 27th and 33 on the 28th. I would not read that as proof a daily cadence cannot work, because the whole board is cooling at once and the two cannot be separated from outside. What it does show is that resetting the board was not, by itself, enough.
07What I am building, and why it is not the same thing
I am building APEX, and after this week I owe you an honest framing of it. It resets monthly rather than daily, and keeps a permanent all-time record alongside. That much is now convergent, and I am not going to pretend otherwise. Two things differ, and both follow from the cadence.
That second one is a reward a daily board structurally cannot offer: a day is not long enough for winning to be an achievement worth commemorating. And there is room for more than one of these. Ad networks coexist the way social networks do, on the same basic mechanic with different rhythms and different promises. This is a different rhythm.