Case Study: A 1999 GoTo.com Pay-Per-Click Win
How a small mail-order retailer used GoTo.com's new pay-for-placement model to outperform its banner ad budget in 1999.
Picture a small mail-order retailer that sells specialty kitchen tools, the kind of business that has run banner ads on cooking sites for two years with mixed results. Click-through rates on those banners have been sliding all year, and the owner is tired of paying a flat rate for impressions whether or not anyone clicks. In early 1999, a friend mentions a search engine called GoTo.com that works differently: instead of buying impressions, advertisers bid on search terms, and they only pay when someone clicks through.
The setup
GoTo.com launched in February 1998 with a model nobody else in search was running: businesses bid for placement against specific keywords, the highest bidder appears at the top of the results for that term and the advertiser pays only on a click, starting as low as a penny. By 1999 the company has gone public on Nasdaq and is signing deals to syndicate its paid listings to smaller portals and search sites, which multiplies the places a winning bid can show up.
Our retailer starts small. Instead of committing a quarter’s banner budget upfront, the owner sets aside a few hundred dollars and bids on a handful of specific terms: “copper cookware,” “espresso machine parts,” “cast iron skillet care.” Nothing broad, nothing expensive. Bids start at five cents and climb only on the terms that convert.
What happened
The first two weeks are unremarkable. Traffic trickles in, some of it converts, some doesn’t. But because every click is tied to a search term and every sale is tied to a click, the owner can see for the first time exactly which keywords make money and which ones don’t. That’s new. The banner ads never gave that kind of clarity, a flat CPM tells you what an impression cost, not what a customer cost.
By week four, the pattern is obvious. “Cast iron skillet care” converts at nearly triple the rate of the broader terms, because the people searching it already own the product and are looking for a replacement part or accessory, not browsing. The owner raises that bid, lets the underperforming terms lapse and reallocates the same monthly budget toward the handful of keywords that pay for themselves.
The result
Over three months, the retailer spends roughly the same total dollars it had been spending on banner placements, but the pay-per-click campaign delivers noticeably more orders, because the spend concentrates on people who are already searching for what the store sells rather than people who happen to be reading a recipe site. The banner ads reached a wider audience; the keyword bids reached a more relevant one, and relevance is what converted.
None of this required a big budget or technical sophistication. It required treating each keyword as its own small experiment, watching what the numbers said and being willing to kill what didn’t work instead of hoping it would improve.
The lesson underneath the numbers
What made pay-per-click advertising different in 1999 wasn’t the technology, bidding systems existed elsewhere, it was the alignment of incentive: you paid for attention that had already expressed intent, not attention that merely existed. A cooking site’s reader might or might not want a new skillet. Someone typing “cast iron skillet care” into a search box almost certainly does.
That distinction is worth remembering as more advertisers discover this model over the next year. The tool matters less than the discipline of testing narrow, measuring honestly and following the data instead of a hunch.
More on Paid Advertising, or see everything published in 1999.
