Chapter 4 of 11 · 8 min read
The Price of Intent
From The Market for Your Mind by Wrotebook
Most of the attention in the history of advertising has been borrowed. A family settles down for a Western and is interrupted by margarine. A reader opens a page for the news and finds a strip of colour asking about his mouse. In each case the advertiser is trying to divert attention that was pointed at something else, and much of the industry's ingenuity has gone into making that diversion louder, cleverer or harder to refuse.
Search is different, and the difference is worth a great deal of money. At seven o'clock on a January morning in Leeds, a person standing in a cold kitchen in a dressing gown types four words into a search box: boiler repair Leeds today. Nobody has had to interrupt them. They have told a machine, unprompted and in their own words, what they want, where they want it and roughly how urgently, which is more than most shop assistants are told by most customers. They have, in effect, walked into the shop and asked. For an advertiser there is no more valuable attention than this, because it arrives with intent attached, and the companies that worked out how to sell it built some of the most profitable businesses in the history of commerce.
The idea did not begin with Google, and it is only fair to say so plainly, because the story is usually told as though it did. In 1998 a company called GoTo.com, one of the businesses hatched at the Idealab incubator in Pasadena, launched a search engine with a frank premise: advertisers would bid for the right to appear when people searched for particular words, the highest bidder would be listed first, and the advertiser would pay only when somebody clicked. The results, that is, were partly for sale, and the company did not pretend otherwise. Respectable opinion was appalled. A search engine was supposed to be a kind of librarian, and here was one taking tips at the desk; critics said that nobody would trust a list of answers that had been auctioned, and that GoTo had mistaken a reference service for a market stall. The critics were right about the principle and wrong about the users, who turned out to be perfectly happy to click on a paid listing for a florist if the florist could deliver on Saturday. It worked. GoTo, renamed Overture in October 2001, supplied its paid listings to other search engines and portals and became for a few years a large and profitable business, so profitable that Yahoo agreed in July 2003 to buy it for about $1.63 billion.
Google arrived at the same place by a longer road. Its founders had written, as graduate students, that a search engine paid for by advertising would be under constant temptation to favour its advertisers over its users, which was a sound observation and turned out to be an excellent description of the business they went on to build. But a search engine that is used by millions of people and earns nothing is a charity, and in October 2000 Google launched an advertising system called AdWords, which placed short text advertisements beside the ordinary results. At first they were sold by the thousand impressions, like banners. In 2002 Google relaunched the system as an auction in which advertisers bid for each click, and it added a twist that mattered more than anything else in the arrangement. The position of an advertisement was decided not by the bid alone but by the bid combined with Google's estimate of how likely people were to click on it, the seed of what Google came to call Quality Score.
Overture took the view that this was rather close to its own system and sued Google over a patent in April 2002. The case was settled in August 2004, just before Google's flotation, when Google issued Yahoo, by then Overture's owner, 2.7 million shares, which settled the patent suit and a separate dispute about share warrants at the same time. As licence fees go, it was paid in a currency whose value the payer was in an unusually good position to predict.
The twist explains both why search advertising was so useful and why it was so lucrative, and it is easiest to see in an illustrative auction for that cold morning in Leeds. The firms and figures that follow are invented. Three heating companies want to appear when someone searches for boiler repair in the city. The first, a national chain, bids £6 a click, but its advertisement is generic and people rarely click on it. The second, a family firm in Headingley, bids £4, and its advertisement promises a same-day visit in the searcher's part of town, so people click on it often. The third bids £2 with an advertisement so dull that almost nobody clicks at all. A pure price auction would put the chain at the top. Google's auction multiplies each bid by its estimate of how clickable the advertisement is, and on that basis the family firm wins despite bidding less, because an advertisement people actually click on earns more for a platform that is paid per click. The winner then pays not its full bid but roughly what it needed to beat the advertiser below it. Everyone pays a price set by competition, relevance is rewarded, and the person in the dressing gown sees the firm that can come today.
The auction does not stop there, because it is run again for every search, all day, and the bidders learn. The chain, having noticed that it is losing to a smaller firm, can raise its bid, or rewrite its advertisement to mention same-day visits, or both. The family firm, having noticed that its clicks have become more expensive, can raise its bid in turn, narrow its advertising to the postcodes it can actually reach before lunch, or decide that January mornings are worth paying for and July afternoons are not. Each of these decisions is made against a screen of figures supplied by the auctioneer, showing how often each advertisement appeared, how often it was clicked and what each click cost. Nobody in Leeds sees the other bidders' figures. Everybody sees the price going up.
The arrangement did two things at once. For a while, it aligned the interests of everybody involved more neatly than anything the attention market had produced before. The searcher saw advertisements that were often genuinely helpful. The advertiser paid only when someone interested arrived, and could work out, nearly to the penny, what each new customer had cost. And the platform did best when it showed advertisements people wanted to click on. At the same time, the formula deciding who won was Google's own. Advertisers could see their bids. They could not see the calculation that turned a bid into a position, and they could not audit the estimate of their own advertisement's worth, which was produced, maintained and adjusted by the company running the auction. The auctioneer was also the judge of which lots deserved to sell, and nobody else was in a position to check its arithmetic.
The receipt for search at its best runs to a paragraph. In exchange for four typed words and one click, the person in Leeds is given the name of a firm that can come this morning, which is precisely what they asked for. The family firm pays Google a few pounds and gains a customer who already wanted it, at a cost it can set against the job. Google takes those few pounds for a fraction of a second of computing and an auction the customer never saw. What was measured was a click and, increasingly, whether the click became a phone call. What was left behind was a warm kitchen, a small transfer of money from a heating engineer in West Yorkshire to a company in California, and an engineer's remark, on leaving, that the boiler was sixteen years old and would not see many more winters. It is hard to find a cleaner transaction anywhere in the market.
It did not stay that clean, for reasons that had nothing to do with anyone's character and everything to do with arithmetic. Google's advertising became the engine of a very large company, and shareholders expected the engine to grow every quarter. Over the years the number of advertisements above the ordinary results increased, especially on searches that suggested someone was about to spend money, until on many queries, and particularly on a small phone screen, the first thing a person saw was a block of paid listings. The ordinary results, the ones that had made Google trusted in the first place, moved down the page by degrees, like a guest at a dinner gradually being reseated nearer the kitchen.
The line between advertisement and result blurred at the same time. In the early years paid listings sat in a shaded box or down the side of the page, plainly apart. The shading paled, then went. The marker that identified an advertisement shrank to a small label in a colour chosen with some care. Each change could be defended on the grounds of tidiness, and taken together they made it slightly harder to tell what had been paid for, which is exactly what an advertiser paying by the click would want and exactly what the person searching might not.
The unpaid results bent the web in their own way. If appearing near the top of the ordinary list was worth a fortune, it was worth paying people to get you there, and search engine optimisation became a profession. Much of it was sensible: making pages that loaded quickly and answered the question clearly. Some of it produced a literature written for an audience of one ranking algorithm. Recipes acquired long personal essays before the list of ingredients. Answers to simple questions were buried beneath paragraphs that existed chiefly to keep a reader on the page long enough for the advertisements to load. The search engine was trying to rank the most useful pages, a great many people were trying to look useful to the search engine, and the reader was the ground over which the contest was fought.
For the small advertiser the arithmetic changed too. An auction is designed to discover what a click is worth to the most determined bidder, and in a busy trade that tends to approach the whole of the profit the click might bring. In the early years a family firm could buy clicks cheaply because few competitors had found the auction. As more firms joined, and as large companies with large budgets bid on the same words, the price of a click in competitive trades rose steadily, and the platform, which ran the auction and kept the formula, collected the difference. A heating engineer in Leeds could end up handing over a meaningful share of each job's margin for the privilege of being found by people who were already looking for a heating engineer.
The most instructive turn came when the logic of intent was applied to a business's own name. Someone who types the name of a particular firm into a search box almost certainly wants that firm. But its competitors are free to bid on its name and appear above it, and so the firm must bid on its own name too, to make sure that the first thing its own customers see is its own website rather than a rival's. A large number of businesses now pay, every month, for clicks from people who typed their name, knew what they wanted and would have found them anyway, which means that the most valuable attention in the market has become, in part, a toll on the road between a customer and the front door they were already walking towards, collected by the company that owns the road and paid by the company that owns the door.