Local.com: Turning a –65% ROI Around With Only 1,000 Tracking Codes

A fast-growing business with unsustainable economics

Local.com had quickly become one of the leaders in the emerging local-search category, helping people find nearby businesses by pairing categories with geography. The opportunity was real: consumers were increasingly turning to search to find restaurants, contractors, retailers, and professional services near them, and Local.com had climbed to third place in the category.

The growth was hiding serious structural problems. Paid search was running a negative 65% return. Organic traffic was almost nonexistent. The site was hard to index, link to, bookmark, or share. And the keyword portfolio held millions of combinations while offering almost no visibility into which terms created economic value. Local.com had reached prominence without building a sustainable way to acquire and monetize its audience. I joined to help turn that around.

The problems were all one problem

On the surface, Local.com looked like it had five separate challenges: SEO, paid search, site architecture, attribution, and monetization. In reality they were symptoms of one operating model.

The site was built around forms, cookies, temporary sessions, and dynamically generated results, and that architecture shaped everything downstream. Search engines couldn’t reliably discover the content. Users couldn’t save or share individual pages. Marketing couldn’t attach durable identities to important experiences. The paid-search team couldn’t connect individual keywords to downstream revenue. The company was trying to optimize growth without a stable structure through which customers, search engines, and measurement could interact with the product. This wasn’t a case of marketing needing sharper tactics. The platform was preventing marketing from learning.

An invisible website

The homepage followed the minimalist, Google-style model: a search box and little else. That works fine for someone who already knows the site and wants to run a local search, but it gave crawlers almost no static content to interpret.

The results pages were a bigger obstacle. They depended on form submissions, cookies, session data, and temporary parameters, so individual pages had no stable URLs. Someone could search for a plumber in San Diego and get a useful set of results, but that page wasn’t built to exist on its own. It couldn’t be reliably indexed, linked from another site, shared, bookmarked, revisited at a consistent address, or evaluated as a lasting content asset. Local.com sat on an enormous trove of valuable local information, and from a search engine’s point of view most of it was invisible.

A paid-search model built on volume, not value

Local.com also leaned heavily on traffic arbitrage: it bought visitors from search engines and made money when those visitors engaged with ads on the site. That model lives or dies on a thin margin — the revenue from each visitor has to beat the cost of acquiring them — which makes accurate measurement essential. A small error in bidding, targeting, or attribution can flip a profitable keyword into a loss when it’s repeated across millions of searches.

But the keyword strategy had been built mostly by bolting dictionary words onto geographic locations, which produced enormous scale with very little commercial discipline. A keyword can be perfectly valid and still signal no buying intent whatsoever. Some terms represented categories advertisers paid well for. Others just pulled in curiosity and general information seekers worth almost nothing. Local.com was buying across that entire spectrum with no dependable way to tell the terms apart. It had scaled the number of keywords before understanding which kinds of intent were worth acquiring.

Measurement under severe constraints

Search-engine policy made it harder. To curb arbitrage, the engines capped Local.com at roughly 1,000 tracking codes even though the company bought traffic across millions of keyword combinations, and they blocked us from assigning unique revenue tracking to every destination page.

In an ideal world, each keyword would connect directly to its acquisition cost, customer behavior, page activity, ad engagement, revenue, and profitability. That level of attribution simply wasn’t on the table. Millions of distinct acquisition decisions had to be understood through 1,000 measurement buckets. A conventional tracking approach couldn’t deliver the precision we needed, so we had to find a different way to organize the market entirely.

Reframing the challenge

The obvious questions were “how do we rank more pages?” and “how do we lower PPC costs?” We asked something broader: how can we organize local customer intent so that one structure improves discoverability, acquisition, attribution, experience, and monetization all at once?

That reframing tied together work that had been treated as separate. The site architecture needed to mirror how people actually searched. The keyword strategy needed to reflect the categories the business could monetize. The tracking system needed to group those categories meaningfully. The reporting model needed to link each group to economic performance. Search, paid acquisition, and revenue attribution would all run off the same understanding of the market — and that shared structure became the foundation of the turnaround.

Rebuilding the site around durable intent

Working closely with the technology team, we replaced the site’s dependence on cookies, sessions, and form submissions with an architecture based on URL parameters and rewritten, search-friendly addresses. Instead of existing only as a fleeting response to a form, a local search result got a permanent location. A given business category plus geography could become a durable page that both search engines and customers could recognize — one that could be crawled, indexed, linked, shared, bookmarked, revisited, and measured over time.

It looks like a technical change, but its effect ran across the whole customer journey. A stable URL turns a throwaway search result into a reusable digital asset, and every relevant category-and-location combination became another way for Local.com to be found. The site stopped behaving purely like a search tool and started behaving like a structured library of local commercial information.

Drawing a map of the market

Making the site indexable was only step one. Generating a page for every possible word-and-location combination would just have recreated the paid-search problem — massive volume, inconsistent relevance. We needed to know which local categories people actually searched for and which of those could produce real advertising revenue.

I ran extensive keyword research to find the most popular and profitable business types, then built a comprehensive site map connecting those categories to relevant geographies. It wasn’t merely an SEO inventory; it was a working model of the local-search market. It let us answer which categories drew meaningful demand, which signaled commercial intent, how that intent shifted by location, which combinations deserved dedicated pages, where advertiser demand was strongest, which areas had organic potential, and which terms were worth buying. The site map connected customer demand to the company’s ability to monetize it.

Replacing the keyword portfolio

Working directly with Google, I rebuilt the paid-search portfolio, dropping the dictionary-word approach in favor of profitable business categories paired with local geography. The shift was fundamental. The old portfolio asked whether words could be combined into a search term. The new one asked whether a combination represented a local commercial need Local.com could actually satisfy and monetize.

A broad word might drive traffic while telling us nothing about intent. A recognizable business category tied to a specific location told us what the person needed, where they needed it, which local businesses were relevant, whether advertisers would value the interaction, and which landing experience to show. The keyword itself became a customer-intent signal.

Building a shared commercial taxonomy

To organize the market consistently, I developed a custom taxonomy from a selected subset of Standard Industrial Classification codes plus geography, grouping individual search terms into meaningful commercial categories. Millions of keywords could now be understood through a much smaller set of structured dimensions — business type, industry category, geography, commercial relevance, expected monetization, traffic source, and customer-intent pattern.

That solved more than an organizing problem. It gave SEO, paid search, technology, content, analytics, and monetization one shared language. A business category meant the same thing whether it showed up in a keyword, on a landing page, in the site hierarchy, in a tracking code, in a revenue report, or in an optimization decision. That consistency let information actually move across the system instead of getting stuck at each team’s border.

Designing attribution around the constraint

The 1,000-code limit didn’t go anywhere. We couldn’t measure millions of keywords individually. Rather than treat that as a reason to fly blind, we mapped the taxonomy into the available tracking codes, so each code represented a meaningful slice of the market instead of an arbitrary bag of words. Then we combined several signals — inbound-link tracking, search-engine keyword reports, taxonomy assignments, landing-page categories, ad revenue, and acquisition cost — to infer the economic performance of each keyword group and allocate revenue more accurately.

It wasn’t perfect one-to-one attribution. It was a model built to produce reliable-enough decisions within the constraints we actually had, and that distinction matters. Organizations love to postpone optimization while waiting for complete data, perfect tools, or unrestricted measurement. We had none of those. So we built the strongest possible decision system from the signals available.

Turning imperfect data into intelligence

The taxonomy surfaced patterns individual keyword reports never could. We could see which categories generated the most revenue, which geographies produced profitable traffic, which combinations drew repeat users, which groups justified higher bids, which created volume but little value, which landing structures improved monetization, which organic pages deserved more investment, and where acquisition costs were outrunning revenue.

That moved us from keyword management to portfolio management. A single term can bounce around for a dozen reasons; a category of related expressions gives a far steadier read on demand and value. We could decide on patterns instead of twitching at isolated data points.

Connecting organic and paid search

SEO and paid search usually get run as separate channels. At Local.com they shared the same customer-intent structure. Keyword research fed the site map. The site map decided which category-and-location pages existed. Those pages became destinations for both organic and paid traffic. Paid search revealed which categories produced revenue, which flagged where organic visibility would be especially valuable, and organic behavior added further evidence of durable demand.

The loop ran both directions: research surfaced promising categories, the architecture turned them into indexable pages, paid search generated immediate traffic and performance data, revenue data revealed which categories had value, organic optimization concentrated on the strongest opportunities, growing organic traffic cut dependence on paid acquisition, and customer behavior refined the taxonomy and the investment strategy. The two channels stopped being competing budget lines and became complementary learning systems.

Better architecture, better experience

The new architecture started as a discoverability and measurement fix, but it improved usability too. Customers could return directly to useful local-search pages, send a page to someone else, and find results that external sites had linked to. Search engines could drop people straight onto a relevant category and location instead of funneling everyone through the homepage.

That cut friction. Someone hunting a specific service in a specific place could enter at the point most relevant to their need, and no one had to restart because an internal session had expired. Better architecture lined up business performance with customer convenience rather than trading one against the other.

Scaling what the economics supported

Once we could connect categories, locations, acquisition costs, and ad revenue, paid-search investment got disciplined. We no longer had to treat every keyword combination as equally valuable. The system could make different calls for different segments — raise bids where margins were strong, expand geographically where performance justified it, cut spend on low-revenue combinations, drop terms with weak intent, improve landing pages for promising-but-underperforming segments, and build organic visibility around consistently valuable demand.

Paid-search traffic eventually grew tenfold. But the number that mattered wasn’t the volume. It was that the economics moved with it: ROI went from negative 65% to positive 35%. The company didn’t just acquire more traffic. It got better at deciding which traffic was worth acquiring.

The results

The transformation reshaped the whole growth model. Organic traffic doubled as the newly indexable architecture exposed the category-and-location content to search engines. PPC traffic grew tenfold as the rebuilt portfolio captured a wider range of commercially relevant local searches. Repeat visitors rose 50%, a sign customers were finding enough value to come back. Revenue per 1,000 pages climbed from $8 to $29, more than tripling what the audience earned. And most important, paid-search ROI swung from –65% to +35%.

Local.com had traded an unsustainable acquisition model for a system that could balance scale against economic value. By the end of 2006 it had become a Top 100 website, with CEO Heath Clarke reporting that organic traffic had nearly doubled and visits per user had risen.

The adaptive marketing lessons

The Local.com transformation demonstrates several principles at the center of The Adaptive CMO.

Channel problems are often system problems. Local.com looked like it had an SEO problem and a paid-search problem. Both traced back to the same architecture, taxonomy, and measurement limits. Treating the channels separately would have produced fragmented gains; we redesigned the system underneath them. Adaptive leaders look past the underperforming metric to the structure producing it.

Architecture determines how fast marketing can learn. The original platform created temporary experiences that couldn’t be indexed, linked, or reliably measured, which limited far more than search visibility — it limited the company’s ability to understand how customers found and used the site. Durable URLs created stable units of customer intent that could accumulate traffic, links, behavior, and revenue over time. Flexible marketing needs an architecture that preserves learning.

More data isn’t more insight. Local.com had millions of keywords and no understanding to show for it. The taxonomy reduced apparent complexity by organizing terms around meaningful categories and geographies — fewer decision units, more clarity. Adaptive marketers don’t react to every data point independently. They find the patterns that make the data actionable.

Let customer intent organize the system. The same category-and-geography model drove paid keywords, organic pages, navigation, tracking codes, landing experiences, revenue reporting, and optimization. It worked because it mirrored how customers actually expressed local needs, so the operating model got organized around observable behavior rather than internal channel boundaries.

Constraints can force better models. We had 1,000 tracking codes for millions of keywords. That killed perfect attribution — and forced us to build a taxonomy that revealed broader patterns of intent. The limitation became a design condition. Adaptive teams don’t ignore constraints, but they don’t let constraints end the learning either. They ask what reliable decisions are still possible.

Build decision-grade measurement, not theoretical perfection. Our attribution leaned on mapping, correlation, and inference. It was not a perfect record of every interaction. It was strong enough to separate profitable categories from unprofitable ones and guide better bids and investment. Measurement creates value when it improves a choice. Credible evidence you can act on while the opportunity is still open beats perfect precision that shows up too late.

Scale profitable patterns, not raw activity. The old strategy manufactured millions of keyword combinations without checking whether they represented monetizable demand. We rebuilt around categories backed by research and economic evidence, and PPC traffic grew tenfold because the system had found a stronger basis for expansion. Adaptive growth doesn’t start by doing more. It starts by discovering what deserves more.

Organic and paid should learn from each other. Paid search gave fast feedback on demand and monetization; organic built a compounding traffic base around the strongest categories. Both ran off the same taxonomy and architecture, so lessons from one improved the other. Adaptive marketing lowers the walls that trap useful information inside individual teams.

Improve customer value and business value together. The new URLs helped search engines and analytics, and they also let customers bookmark, share, and return to useful pages. The redesign worked because it didn’t treat optimization as something done to the user. The same changes that improved business performance made the experience more convenient.

A shared model creates organizational leverage. The taxonomy gave technology, marketing, analytics, and monetization a common picture of the market. Without it, each team might have optimized its own corner while creating problems elsewhere. The shared model let the organization coordinate around one understanding of customer intent and economic value.

The lasting lesson

Local.com had impressive market visibility when I arrived and a growth model that couldn’t last. The site hid its content from search engines. Its architecture blocked durable customer journeys. Its paid-search portfolio chased volume over intent. Its measurement limits made profitability hard to even see.

The fix wasn’t a single SEO tactic, bidding tweak, or redesign. We built a shared system. Stable URLs made local content discoverable and reusable. A commercial taxonomy organized customer intent. The site map connected that intent to relevant experiences. The paid-search strategy concentrated on categories with real monetization potential. And the attribution model turned constrained signals into useful economic insight. Together, those changes grew organic and paid traffic while substantially improving revenue and profit.

That’s what adaptive marketing looks like when resources and measurement are constrained. It doesn’t wait for perfect information or unlimited technology. It builds a useful model of the market, connects the signals it can get, acts on the strongest evidence, and keeps improving the system as it learns.