The Global Product Hierarchy Project

The situation

It’s hard to be strategic and to make intelligent decisions if you don’t know what it is you’re working with. At a Fortune 100 industrial manufacturer, no one had a full picture of how big the digital footprint of the company’s websites was. It had grown for years across many teams, platforms, and strategic initiatives. My team had to crawl each global domain one at a time to try and get an idea of what we were working with. We stopped counting at roughly 92 million URLs.

Once we scoped it, the problems multiplied. Some catalogue category pages returned no products. Page types had never been formally defined, but had to be. Additionally, the company’s analytics could not yet be filtered by division or business group, which limited how much the data could support strategic decisions.

One business group alone, in one of the priority countries, carried about 160,000 URLs: roughly 117,000 product pages, 43,000 catalog category pages, a few hundred content pages, and a few hundred more on a hand-coded blog platform. That count left out the resource library of data sheets, instructions, and brochures, which was a cleanup project of its own.

Then the company decided to move everything to a new global product hierarchy. Every URL would change, and the business could not afford any interruption while it happened.

The task

My team’s job was the search side of the migration. We needed to figure out what was worth saving, make sure the search value moved with it, and keep the business running the whole time. We also had to work with the IT and analytics teams to make sure the infrastructure and tags were in place for us to be able to: 1) scope the problem, 2) create benchmarks and prioritizations, 3) work with a taxonomist to create the new catalogue structure, 4) create base-variant canonical structures, 5) roll out the changes successfully.

Our scoping summary opened with a picture of a mouse looking through a block of Swiss cheese. That was the honest status. We knew where the holes were; we did not know how big they were, and we needed decisions from the business before we could estimate the rest.

The action

I segmented the countries before anything launched. I sorted every country into a 3-tiered structure (A, B, and C levels) by business priority, so the rollout could start where a mistake was cheap and finish where it was expensive.

We named the keystone. Every downstream activity depended on one thing: an approved information architecture. URL recommendations, keyword and call-to-action mapping, backlink analysis, redirects, content recommendations, analytics, and tool stack updates all waited on it. Saying so in writing put the decision where it belonged, with the business, instead of letting the delay land on the search team.

We built a decision tree for what deserved to survive. We pulled crawl data, connected a crawler to Search Console and two backlink databases, and ran every URL through the same test. A live page with traffic in the last 30 days qualified. A live page with no recent traffic qualified if it still showed up in Search Console or carried backlinks. Anything with none of those was deprioritized: no redirect, a noindex tag, and a removal sitemap to clear it out of the search engines. Broken URLs got one recrawl, then went through the same test to see if it survived or not. In one business group, fewer than 4 percent of non-product URLs earned a place in the redirect plan.

“Lift-and-shift” vs Mosaic pages: Pages moving in their entirety to a new location in the IA got their backlink analysis before content mapping. Pages splitting across several new pages got it after, once we knew where each piece landed. We salvaged content instead of starting over. After screenshotting every priority page, we kept the sections still doing their job, built new wireframes from that existing content, and filled the gaps with new copy. Not everything had to be net new.

We defined every page type and mapped it to the customer journey. For the first time, each page type had a stated purpose and intent. Mapping those to the customer journey showed marketers exactly where the content gaps were, so they could fill them with their own work. Content-rich category pages with no products underneath them moved to the content platform, where they could actually be indexed and ranked.

We made the data filterable by the business structure. We worked out how to cut the data down to the division level, so each division could see its own pages and buy in, and how to roll it back up to the business group. Then we got a metadata tag implemented in the page code so analytics tools could segment everything by the business structure.

We wrote the to-do list nobody else would. A changed URL breaks things far from the website. We documented every downstream change marketers would face: campaign tracking links, analytics filters and vanity URLs, advertising landing pages, marketing automation landing pages, channel partner and press backlinks, packaging, trade show materials, brochures and PDFs, video descriptions, and hard-coded links.

The rollout launched where a mistake wasn’t costly. My teammates ran the project through 2019 on the plan we scoped for approval, and the new hierarchy went live in a smaller country first. Then a handful of other non-priority countries, where a problem would not hurt the business much. Then bigger and more important markets, one wave at a time. By the time they reached the USA, the team was confident enough to just switch it over, and the whole hierarchy was live by the end of 2019.

The result

The footprint went from roughly 92 million URLs to 20 million URLs across 83 global domains in 47 unique languages with no business interruption.

The country segmentation outlasted the project. The A, B, and C levels evolved into Tiers 1, 2, and 3, and later into a tiered country model that set which experience each market got, from export-only and catalog-only countries to markets with the full experience of content pages, catalog, and tool and product selectors.

The metadata tag became the basis for segmenting analytics by division and business group in the years that followed. And the page type definitions gave marketers a map of their own content gaps instead of a list of broken pages.

It took a lot of people: product marketers, business group marketers and marketing centers, the corporate teams, IT, and regional marketers around the world. The coordination was the project.

The lesson

Not everything deserves to survive a migration. Decide what earns a redirect before you decide what the new site looks like. Then launch where a mistake is cheap before you launch where it is expensive.