Replaced the platform from underneath 7 live sites without losing revenue
Cole-Parmer’s CMS and search platform had aged beyond support, and the sites still had to sell daily. I was responsible for all front-end tools to guide replacing half the site’s CMS, build a new search and merchandising approach with AI in it, then redesign the item detail page (IDP) on top of the new foundation.
AI widget revenue grew 400%+, click-through rate rose 40%, and the migration cost no revenue. This is how the work went, including the choices I made to keep the sites steady while they changed.
What changed
Each line jumps to the part of the story behind the number.
What had to be true when we finished
The platform could not be patched any further, and customers would keep ordering the whole time it was replaced. So the list was long and had one hard edge.
- A modern CMS
- AI-driven search and merchandising
- Smarter faceted navigation
- Redesigned item detail pages
- Scalable content architecture
- No downtime and no revenue loss during the transition
That meant replacing half of the site’s CMS, building a new search and merchandising strategy, and adding and adjusting AI as we went.
Owning UX, CMS, and AI widgets end‑to‑end while replatforming
Being the sole lead meant the business rules for each group lived with me, and so did the job of keeping everyone informed as the platform changed under them.
I worked with product managers, IT and engineering, marketing, customer service, SEO, analytics, and the Bloomreach AI platform specialists. I kept each group’s business rules intact, kept them updated, and sold them on the new widgets and what we discovered along the way.
At least one person from each group tested after launch, so I could see the work from their side and improve it. The clearest example came from customer service.
Helping customer service use the AI widgets to improve their manual process
- They did it by handCustomer service was used to finding cross-sells and upsells manually when helping customers.
- I showed them the new wayI demonstrated that the item page widgets would surface cross-sells and upsells automatically, and kept the team in tune with how those widgets worked.
- They helped keep the AI honestI asked them to keep watching the results. Their vigilance turned up ways to improve what the widgets showed.
Phase 1: replace the platform and keep the site steady
My decisions for UX, CMS structure, search logic, merchandising rules, and AI widget configuration kept the site stable while replatforming as well as fitting in upgrades.
Flexible on strategy and on colleagues. Never flexible on the goals.
No downtime and no revenue loss were on the list from the first day, so every choice below was made to protect that while still moving the site forward.
- Same look, new foundationsThe replatform was not meant to change how the site looked or worked. When we built new widgets, I designed them in the modern style and slipped them into the older look, which started a piece-by-piece update.
- Templates built from the bottom upLearning a new CMS while building page and merchandising templates could have gotten cluttered. Building from the bottom up, always from the customer’s point of view, meant few or no follow-up fixes.
- Search kept simple on purposeSimple logic let the AI learn, and let us collect inputs and pain points to boost, bury, and merchandise at scale later. Holding enhancements back let me do them thoughtfully, and they became the visual typeahead.
- Merchandising rules kept simple and scalableProprietary items got priority. I tailored global rules and built AI widgets that could run anywhere on the site, on the item page redesign, and in a future AI-powered quoting system.
- Tuned once the AI had learnedAfter the AI picked up customer tendencies, I quickly adjusted search and filters so customers could choose items faster and better, using input from customer service and user testing.
Phase 2: redesign the item page on the new platform
With the replatform done, the item detail page could finally use the new tools.
I guided my UX designer through the redesign. We applied new designs, microinteractions, and a modern look, removed old code and tools, placed new AI widgets from the new CMS in strategic spots, and built frameworks for future dropzones for other widgets and ads.
What the new page prioritizes
- Purchase-critical information above the fold
- Specs, documentation, compatibility, and certifications, clearly structured
- A guided journey through all the supporting content
- AI-powered cross-sell and upsell modules
- Less cognitive load for technical buyers
Opinions came, so I made the design easy to follow
Plenty of people had views on how the page should look and work. I kept stakeholders and the curious calm with a design template I shared early and explanations of how their work would fit in. After launch, once everyone had adjusted to the new page I hosted a session for their input.
3 AI widgets made for how customers needed them
These three widgets scaled across thousands of item pages and could be placed anywhere else on the site. Each runs on logic I set: product attributes, compatibility, category-level rules, global and individual business rules and priorities, and AI relevance scoring. I ordered them to allow customers build bigger and better carts with the accessories and cross sells they needed.
- Buy It Together First to show because it brings cross-sells, upsells, and accessories together so customers can see and buy every piece they need for working equipment. It grew add-to-cart numbers and drove the 400%+ increase in widget revenue.
- Customers Also BoughtIn the middle to scale with the same attribute and rule logic as the others yet could be tailored towards specific product families.
- Compare With Similar ItemsLast as it lazy-loads a lot of comparative content to show customers their upgrade options.
Search that keeps learning
Customers were moving from search terms to search phrases. Search had to keep up.
I knew the next challenge coming out of the change was that customers would ask more of the search and use whole phrases and certain specifications on products instead of terms and category names, and expect accurate results to match.
So I planned for it. I set up a partnership with buyers and customer service who pushed search and demanded more tailored answers, using the specific phrases people type into Google and LLMs. I trained the AI toward better results through specific scientific term synonyms and prioritizing often-used filters to show search.
It was a wildly rewarding long-term project and partnership, and the visual typeahead grew out of it.
The overall result provided a scalable CMS architecture across 7 sites and 4 languages, without any revenue disruption during migration, and building a foundation for future AI-driven UX.
Need your digital strategy and design pain points solved? Let’s talk.
I’m looking for a team to grow with a team long term in a remote or hybrid in the northwest Chicago suburbs.