The Land Grab: How Search Became a Three-Company Business

On August 19, 2004, Google’s stock opened on Nasdaq at $85 a share and closed the day around $100. It was the first major tech IPO since the dot-com wreckage, and almost nothing about it went the way Wall Street wanted.

Google used a modified Dutch auction instead of letting banks allocate shares to favored clients. The original price range of $108 to $135 had to be cut to $85 to $95 amid weak demand and open skepticism about the format. Roughly 19.6 million shares changed hands, raising about $1.7 billion and valuing the company near $23 billion.

The skepticism is the part worth sitting with. In 2004, a lot of serious people still weren’t sure that selling small text ads next to search results was a real business. Two years later, that question was settled, the market had consolidated into three companies, and the structure advertisers still work inside today was fully in place.

Here’s how that happened.

Google leaves the search box

In April 2003, Google paid about $102 million for a company called Applied Semantics — previously named Oingo, founded in 1998 by Gil Elbaz, Eytan Elbaz, and Adam Weissman. Applied Semantics had built a semantic text-processing engine called CIRCA. It had also already coined the name for the product Google was about to launch: AdSense.

Sergey Brin’s line in the announcement was that the acquisition would “enable Google to create new technologies that make online advertising more useful to users, publishers, and advertisers alike.”

The mechanic was a genuine expansion of what PPC could be. Search ads match an ad to a query — someone typed “emergency plumber Fort Worth,” so you show a plumber. Contextual ads match an ad to a page. Software reads the meaning of an article about fixing a leaking water heater and serves plumbing ads against it, on a site Google doesn’t own, splitting the revenue with whoever runs that site.

Overnight, the inventory stopped being “Google’s search results” and became “most of the web.” Publishers could monetize with a snippet of code. Advertisers could buy reach beyond the roughly 40% of searches Google was handling.

The scale came fast. By early 2005, AdSense accounted for something like 15% of Google’s total revenue, and the network grew around 73% year over year in 2005 against a total revenue base of roughly $6.1 billion.

This is the origin of the Display Network, and of the single most common way small advertisers waste money. When Google later merged search and content into one campaign creation flow, a lot of accounts ended up running display ads across random websites while their owners thought they were buying search. The default is better now. The trap still exists.

Yahoo buys the competition

Yahoo spent 2002 and 2003 assembling a rival stack.

First Inktomi, for about $235 million, announced in December 2002 and completed March 19, 2003 — that bought algorithmic search. Then Overture, announced July 14, 2003 for roughly $1.63 billion in cash and stock, which bought the paid search auction Bill Gross had invented. Overture itself had just picked up AltaVista for $140 million and AllTheWeb.

Terry Semel’s framing at the time: the combined assets positioned Yahoo as “the largest global player in the rapidly growing Internet advertising sector.”

On paper it should have worked. Yahoo now owned the search engine, the ad platform, the portal traffic, and the patent. What it also owned was Overture’s auction logic — highest bid takes the top slot, with no relevance weighting. Google had abandoned that model in February 2002 for a reason. Yahoo inherited the older, dumber machine and kept running it for years.

There was one more piece of the Overture purchase: the patent lawsuit. Overture had sued Google in 2002 over the pay-per-click ranking patent, and Yahoo inherited the claim. In August 2004, weeks before the IPO, Google settled by issuing 2.7 million shares of Class A stock to Yahoo and taking a non-cash charge in the $260 to $290 million range, in exchange for a perpetual license. Google’s IPO went out clean.

Google gives away the measuring stick

In March 2005, Google announced it was buying Urchin Software, a San Diego analytics company, for a price never officially disclosed but widely reported around $30 million. On November 14, 2005, it relaunched the product as Google Analytics — free.

That word is the whole strategy. Enterprise web analytics at the time ran into the thousands of dollars a month; Urchin’s own hosted product started around $500. Google took a category that had been a paid business tool and turned it into a giveaway. Demand was so heavy that Google suspended new signups about a week after launch and moved to invitation codes.

The logic wasn’t charity. Google’s Jonathan Rosenberg framed it as helping business owners generate a higher return on their advertising spending — which is exactly right, and exactly the point. An advertiser who can see which keyword produced the sale will spend more on that keyword. Free measurement doesn’t just help you; it enlarges the auction.

This is the same move Claude Hopkins made with keyed coupons a century earlier, run at platform scale: whoever controls the measurement controls the spend. The tracking is the moat. — The History of PPC, Part 5

This is the same move Claude Hopkins made with keyed coupons a century earlier, run at platform scale: whoever controls the measurement controls the spend. If you’ve ever wondered why Google Ads, Analytics, Tag Manager, and Merchant Center are all free, that’s the answer. The tracking is the moat.

The Urchin lineage survived a long time, too. Every “UA-” property ID that ran until GA4 replaced it stood for Urchin Account.

Microsoft finally builds its own

Microsoft had been renting. MSN Search ran ads first from LookSmart, then Overture, then Yahoo — meaning Microsoft was sending its own search traffic to a competitor and taking a cut.

Microsoft adCenter came out of US beta and launched officially on May 4, 2006, announced by Steve Ballmer at the MSN Strategic Account Summit. The Yahoo syndication deal expired the following month, and MSN went fully in-house.

Microsoft was last to the party by four years. That platform is the one you know today as Microsoft Advertising, and the four-year head start Google got is a decent part of why the market share gap never closed.

Yahoo’s response came late. It announced a rebuild of its ad platform in May 2006; the new relevance-weighted ranking system — Project Panama — didn’t go live in the US until February 5, 2007. It worked, somewhat: revenue per search rose about 20% in Q3 2007, Yahoo’s first uptick in six quarters. It was five years after Google made the same change, and it wasn’t enough.

The first fraud reckoning

Advertisers had been complaining for years that they were being billed for clicks that weren’t real customers — competitors clicking your ads to drain your budget, and low-quality publishers clicking their own AdSense ads to generate revenue.

Lane’s Gifts & Collectibles, a small Texarkana retailer, filed a class action in Miller County, Arkansas. In March 2006, Google settled for up to $90 million: up to $60 million in advertising credits covering clicks back to 2002, and up to $30 million to plaintiffs’ attorneys. No cash to advertisers, no admission of liability. A court-appointed expert, NYU’s Alexander Tuzhilin, reviewed Google’s invalid-click detection.

Wired’s January 2006 cover had asked how click fraud could swallow the internet. It didn’t. But the settlement established something permanent: the platform decides what counts as a billable click, and you find out afterward. The “invalid activity” credit line on your billing page is a direct descendant of this case.

Where things stood by the end of 2006

US internet ad revenue ran $7.3 billion in 2003, $9.6 billion in 2004, $12.5 billion in 2005, and $16.9 billion in 2006, with search consistently the largest single format — around 40% of the total.

Query share told one story. comScore had Google at 36.5%, Yahoo at 30.5%, and MSN at 15.5% in July 2005; by June 2006 it was Google 44.7%, Yahoo 28.5%, MSN 12.8%.

Ad dollars told a much lopsided-er one. Google monetized each query far better than Yahoo did, so its share of search advertising revenue ran well ahead of its share of searches. A 2005 industry survey found 95% of advertisers using Google AdWords, 59% Yahoo, and 29% MSN.

Don’t confuse those two numbers when you read search-share stats. They still diverge today.

What this means for your account

Check what network you’re actually buying. The Search/Display split created in 2003 is still the most common source of silent waste in a small account. Search Partners and Display are separate inventory with separate economics. Look at the network segment in your reports before you conclude a campaign isn’t working.

The free tools are not neutral. Analytics, Tag Manager, and conversion tracking are excellent and you should use all of them. Just hold onto the fact that the company measuring your results also sells you the media. Server-side tracking and your own CRM data exist partly to give you a second opinion.

Invalid clicks are handled, not eliminated. Google filters and credits automatically. If you’re in a competitive local trade and see a spike in clicks with no calls, pull the IP exclusion tools and your call recordings before you assume the keyword went bad.

Diversification has been hard advice for twenty years for structural reasons. Microsoft was four years late and never recovered the ground. That’s why “just move budget to Bing” isn’t a strategy on its own — but it’s also why Microsoft Advertising CPCs are usually cheaper. Both facts come from 2006.

Carryover: the vocabulary that survived

Terms established in this era and the instruments they became.

AdSense · 2003: Google’s contextual ad product for publishers. Still AdSense on the publisher side; the advertiser side became the Display Network.
Content Network · 2003: Non-search inventory across third-party sites. Renamed Google Display Network in 2010; now also feeds Performance Max placements.
Search Partners · 2003: Google’s search ads syndicated to third-party search boxes. Still a checkbox in every Search campaign, with separate performance worth segmenting.
Contextual targeting · 2003: Matching ads to the meaning of a page, not a query. Topics, placements, and content keywords in Display and PMax.
Urchin / “UA-” · 2005: The analytics company Google bought and gave away. Google Analytics; UA property IDs ran until GA4 replaced them.
Conversion tracking · 2005: Free measurement Google gave away to enlarge the auction. Google Ads conversion actions, GA4 imports, Enhanced Conversions.
Click fraud · 2006: Clicks billed to advertisers that weren’t genuine customers. “Invalid activity” credits and Google’s invalid traffic filtering.
adCenter · 2006: Microsoft’s first in-house PPC platform. Microsoft Advertising.
Bid-to-position ranking · Overture legacy: The model where the highest bid simply won. Abandoned everywhere; replaced by quality-weighted Ad Rank.
Revenue per search (RPS) · 2003–2006: How much money a single query earned the platform. The platform-side metric behind auction pressure and rising CPCs.