What a twelve-hour ad slot costs, in every unit that matters

Ad slot cost is a single number when you pay it and four different numbers when you try to compare it with anything else. A twelve-hour slot on a board is bought as one flat amount. The rest of the advertising world prices attention per thousand impressions, per click, per visit, or per hour of airtime, and each of those units makes the same purchase look cheap or expensive depending on which one you hold it up to. This post takes one hypothetical slot and prices it every way, so the conversion is done once and you can reuse it. The examples use round numbers and are labelled as hypotheticals throughout; the real figures for our board are on the front page and on each brand's page, and we do not type them into prose because they change.

The slot, as bought

On BidSurvivor — ours, so weigh it accordingly — the day is two slots of twelve hours, 00:00–12:00 and 12:00–24:00 UTC. The first brand into an empty slot pays $0. Taking a held slot means beating the standing bid, from a floor of $1.00 up to a ceiling of $5,000 that no bid can exceed. A winning bid is charged at once and the logo goes up at once. The rules in full are in how the board works; the general shape of the format is in what a time-boxed advertising auction is.

So the input to every calculation below is one number: what you paid. Take a hypothetical slot bought for $12 and held for the full twelve hours.

Unit one: per hour

$12 over twelve hours is $1.00 an hour. This is the unit broadcast and out-of-home use, and it is the unit almost nobody uses online, which is a pity, because it is the one that survives being outbid.

Now the wrinkle that makes the flat price less flat than it looks. Being outbid on this board does not refund you. If a higher bid displaces you four hours in, you paid $12 for four hours: $3.00 an hour, three times the rate you thought you were paying. Hold for the full twelve and the rate is what you bid. Get displaced in the first hour and it is $12 an hour.

Per-hour cost is therefore not fixed at purchase. It is settled when the slot closes, and the only way to pin it down is to be the brand still holding at the end. That is the auction doing what auctions do, and it is the reason a bid is a decision about how much you want the whole twelve hours, not a price per hour.

Unit two: per open

An open is a card opened on the board — somebody looked at your entry, not just the page. Suppose the hypothetical $12 slot produced 600 opens over its twelve hours. That is 2 cents an open.

Opens are closer to what display advertising calls a viewable impression than to an impression: the card had to be opened, not merely scrolled past. There is no industry price for "an open" because no other format counts one, which is inconvenient for comparison and honest as a measure. Treat it as the top of a small funnel.

Two cautions. First, an open count on a young board is a small number and a rate on a small number is noisy; we only show the open-to-click rate once a brand has twenty opens for that reason. Second, opens are counted, not sampled, and bought or synthetic traffic is flagged at source and never counted in them — which is a fact about our counting, not a guarantee about anybody else's.

Unit three: per click

A click-through is somebody leaving the board for your site. Suppose 60 of those 600 opens clicked through. The $12 slot cost 20 cents a click.

That is the unit search advertising is priced in, and it is the one that makes a slot look either very cheap or irrelevant depending on your business. Twenty cents a click is far below what most search terms cost; but a search click comes from a person who typed your category into a box, and a board click comes from a person who was looking at a board. CPM vs CPC vs flat fee goes through why those clicks are not the same click, and the short version is that a cheap click from the wrong person is not cheap.

Change the hypothetical and the number moves fast. Ten click-throughs instead of sixty and the slot cost $1.20 a click. Zero and the unit is undefined, which is the honest answer for a slot that sent nobody.

Unit four: per thousand visitors

The display world's unit is cost per mille — CPM, the price per thousand impressions. To put a slot in that unit you need the number of people who saw the board while you held it. Suppose 2,000 visitors came to the board during the hypothetical twelve hours. $12 for 2,000 is $6 CPM.

This is the unit that flatters the slot least and the one to be most careful with. A visitor to the board is not an impression of your ad in the display sense; your entry is one of two on the page, and it is a card rather than a banner in the margin. It is also the unit where a typed-in number would be most misleading, because visitor totals on a young board move a lot from day to day. The live total is on the front page, refreshed every thirty seconds, and the honest thing is to read it there on the day.

The same slot at $0 and at $5,000

Two edge cases, because the ceiling and the floor matter.

The free claim. An empty slot's first bidder pays nothing. Every unit above is zero: $0 an hour, $0 an open, $0 a click. The slot still produces the counts, which is the point — a free slot is a measurement, and the measurement is what tells you whether a paid one is worth bidding for. Why the first bid is free explains what we get out of that.

The ceiling. A bid at the $5,000 cap works out at about $417 an hour. For that to be a $1 click it would need 5,000 click-throughs in twelve hours, and for it to be a $6 CPM it would need roughly 830,000 visitors. Neither is a number this board has any business promising. The cap is not a price target; it exists so that a slipped decimal or a bidding war cannot produce a figure nobody meant, and why a bid ceiling makes an auction fairer makes the case. But the arithmetic is useful in the other direction: before any bid, divide it by the visitors you can see on the front page and ask whether the CPM you get is one you would pay elsewhere.

Doing the sum before you bid

The procedure, in four lines:

1. Read the visitors, opens and click-throughs from the front page and from the brand pages of recent winners. Those are the real inputs. Do not use the ones in this post.
2. Divide your intended bid by each. That gives you per-visitor, per-open and per-click at the board's current size.
3. Divide it by twelve, then by the number of hours you think you will actually hold. If a higher bid is likely, the second number is the honest one.
4. Compare each unit with what the same money buys on the channel you would otherwise use, remembering that the people are different.

And remember that a bid inside the $100 of credit every visitor starts with costs nothing out of pocket, which changes the sum in a way house credit and what you test goes into. Credit is spent only by a bid that wins, and a bid that fails to take a slot costs nothing at all, so the downside of running the numbers on a real slot rather than a hypothetical one is small.

The point of pricing the slot four ways is not that one unit is right. It is that a flat price hides which unit you are actually buying in, and the board publishes enough for you to work it out before you pay rather than after.