When PPC Became a Job: The Optimization Decade

On December 9, 2005, an advertiser logged into a support forum and described what had happened overnight. A keyword that had been running at 18 cents the day before, pulling 70 clicks, was now demanding a $10 minimum bid. He asked why. The Google representative, by his account, couldn’t answer, and eventually offered that the “little google robot” had probably made a mistake.

It hadn’t. The day before, Google had folded landing page quality into Quality Score.

That’s the moment paid search stopped being a media buy and became a discipline. Before it, you set a bid. After it, there was a system evaluating your ad, your keyword, and your website, deciding what you were worth, and adjusting the price without telling you exactly why. Learning to work that system took skill. Skill is what creates a profession.

Quality Score gets teeth

The concept arrived in February 2002 — Ad Rank as bid times click-through rate, covered in Part 4. The product arrived in mid-2005. Google’s own later blog post dates the introduction of Quality Score with quality-based minimum bids to July 2005; other contemporaneous accounts say August. Either way, mid-2005 is when the number got a name and started setting prices.

Then it kept moving, roughly once a year:

December 8, 2005 — landing page quality enters the calculation. Google’s own forum post announced it plainly: they had introduced Quality Score with minimum bids in August, and were now incorporating a new factor, the landing page. This is the update that produced the $10 minimums and the panic.
November 2006 — Quality Score extended to the content network, plus another landing page algorithm update.
February 14, 2007 — keywords get “Poor,” “OK,” and “Great” designations, and minimum bids become visible rather than inferred.
June 2008 — landing page load time starts affecting Quality Score.
August 2008 — Quality Score moves to real-time calculation. Google’s phrasing: calculated at the time of the search, every time. Not a stored grade. A live judgment on every auction.
September 2008 — minimum bids are replaced by first page bid estimates, and keywords stop going “inactive for search.”

That last change is more humane than it sounds. Under the old system, a keyword with a bad score simply stopped running until you raised the bid enough. Under the new one, Google told you what it would take to reach page one and let you decide.

The through-line across all of it: Google kept adding factors it could measure, and kept moving the score closer to real time. Practitioners spent six years chasing a target that was being redefined roughly annually — which is, in a certain light, an excellent description of the job to this day.

The craft that formed around it

If Quality Score sets your price, and Quality Score responds to relevance, then relevance is the lever. That single conclusion generated most of the tactical orthodoxy of the era.

Structure got tighter. The logic was straightforward: a keyword’s expected CTR improves when the ad it triggers contains the searched phrase and the page it lands on matches the promise. The way to guarantee that is to stop putting fifty loosely related keywords in one ad group. Accounts fragmented into narrower and narrower themes, ending at the logical extreme — one keyword per ad group, which the community eventually named SKAGs. That naming happened in practitioner blogs and conference talks rather than any official source, so treat the origin story loosely; the practice was widespread by the early 2010s.

Match types became strategy. Broad, phrase, and exact each triggered a different slice of the query space, and negative keywords were how you kept the broad ones from bleeding. Broad match modifier arrived in 2010 and gave advertisers something in between: control over which words had to appear, without exact match’s rigidity. Managing that layered structure — broad for discovery, exact for control, negatives flowing back from the search terms report — was a genuine skill that separated good accounts from bad ones.

Measurement caught up. Conversion tracking, Analytics goals, and eventually Multi-Channel Funnels in 2011 moved the conversation past clicks. Last-click was the default and stayed the default for a decade, which everyone knew was wrong and almost nobody had a better answer for.

The tools, the conferences, the job title

A profession needs three things: a body of technique, a way to teach it, and a market that pays for it. All three showed up between 2006 and 2009.

The tools came first. Marin Software was founded in April 2006 by Christopher Lien, Wister Walcott, and Joseph Chang, raised over $100 million, and went public in March 2013 at $14 a share. Kenshoo, later renamed Skai, started around the same period. Acquisio was already running. And in 2007, Larry Kim started building scripts to automate the repetitive parts of keyword work — by his own retelling, out of a Panera Bread — and turned it into WordStream, which took a $4 million Series A from Sigma Partners in August 2008 and eventually sold to Gannett in 2018.

Semrush was founded in 2008 by Oleg Shchegolev and Dmitri Melnikov, growing out of an earlier browser extension called SeoQuake. Criteo had been founded in Paris in 2005 and pivoted to pay-per-click retargeting in 2008. AdRoll was operating by 2007.

The teaching infrastructure formed alongside it — Search Engine Land, SMX, PPC Hero, Hero Conf. Practitioners started writing publicly about technique, testing each other’s claims, and arguing about account structure in a way that looked a lot like a trade forming.

And the market followed. “PPC manager” and “paid search specialist” became things you could be hired as, rather than things a marketing generalist did on Tuesdays.

This matters for a specific reason. For roughly six years, the person operating the account was a meaningful input to performance. Two advertisers with identical budgets and identical products could produce wildly different results based purely on who was managing. That gap was the entire value proposition of an agency. Understanding when that window opened — and, as Part 9 will cover, when it started closing — explains a lot about how the industry sells itself today.

The alliance that didn’t work

Meanwhile, the two companies that weren’t Google decided to merge their weaknesses.

Microsoft and Yahoo announced a search alliance on July 29, 2009 — a ten-year deal in which Microsoft would power Yahoo’s search while Yahoo’s sales force handled premium search advertisers for both. Yahoo kept 88% of search ad revenue from its own sites for the first five years. Steve Ballmer’s line: the agreement gave them “the scale and resources to create the future of search.”

Regulators approved it in February 2010, and the migration of Yahoo Search Marketing advertisers onto adCenter followed.

It’s remembered as a disappointment, and Yahoo’s own commentary through the following years said as much — revenue per search on the Microsoft platform kept falling short of what Yahoo had hoped for. Combining two subscale search businesses produced one subscale search business.

What this means for your account

Relevance is still the cheapest lever you have. Everything in this era was downstream of one fact: Google prices your clicks partly on how well your ad and page match the query. That’s still true. If you’re a roofer bidding on “metal roof installation” and the ad says “Roofing Services” and the click lands on your homepage, you’re paying a relevance tax on every click. Matching the ad and the landing page to the query costs nothing and lowers your CPC.

You don’t need SKAGs, but you do need themes. Modern match types and Smart Bidding have made single-keyword ad groups mostly counterproductive — they starve the algorithm of data. The underlying principle survives the tactic: keywords in an ad group should share an intent, so one ad can honestly speak to all of them. “Emergency AC repair” and “AC installation cost” do not belong together.

Negative keywords are not optional maintenance. The search terms report existed in full then and is restricted now, which makes the negatives you can build more valuable, not less. Jobs, DIY, free, and the name of every competitor you don’t want to pay for are a starting list.

Judge an agency on what it does when there’s no lever to pull. The optimization decade produced a lot of people whose skill was tactical account manipulation. Most of that work is now automated. What isn’t automated is offer, creative, conversion tracking integrity, and knowing what a booked job is actually worth to you.

Carryover: the vocabulary that survived

Terms established in this era and the instruments they became.

Quality Score · 2005: Google’s relevance grade, driving minimum bids. The 1–10 keyword diagnostic, plus the real-time score that actually prices each auction.
Minimum bid · 2005: The price floor a keyword had to clear or go inactive. Replaced in 2008 by first page bid estimates; keywords no longer go inactive.
Landing page experience · Dec 2005: Page quality as a Quality Score input. A named Quality Score component; still one of three.
Broad match modifier · 2010: The middle ground between broad and phrase match. Retired in 2021; its behaviour absorbed into phrase match.
Negative keywords · 2005–11: Query exclusions used to control broad match. Unchanged in function, and more important now that search term data is restricted.
SKAG · early 2010s: Single keyword ad group — the extreme of tight structure. Largely obsolete; the intent-theming principle behind it survives.
Search terms report · 2005–11: Full visibility into the queries that triggered your ads. Restricted since 2020; still the source for negatives and new keywords.
Bid management platform · 2006: Third-party software that adjusted bids on rules. Mostly displaced by Smart Bidding; survives as scripts and Optmyzr-style tooling.
Last-click attribution · 2005–11: The default rule crediting the final click. Replaced by data-driven attribution as the Google Ads default.
Multi-Channel Funnels · 2011: The report showing assist paths, not just the last click. Path exploration and attribution reports in GA4.