Uplift: Fixing the Org Before Fixing the Design System

The moment it clicked

There’s a pattern I’ve learned to recognize. You ask a developer why they built a custom component instead of using the design system, and the answer is vague: “it was faster to just build it.” What they almost never say is: I stopped trusting it.

That was TVH when I arrived. Uplift was a real, thoughtfully built system on MUI with a Figma component library. But when I looked at what ran in production, I kept finding components that didn’t exist in the library. Only one development team used the coded system at all, and even they used it largely without UX guidance.

TVH is a multi-brand, international operation: the main TVH brand serving western Europe, TVH America covering North America, and Bepco, primarily agricultural and based in France. The company’s direction was a cautious consolidation toward a single platform for buying industrial parts. That made the design system the foundation for a major strategic move, and right now one team used it, inconsistently.

The diagnosis

The cause was structural. Not design quality, not documentation, not communication. Capacity.

Design system engineers sat inside the Development organization. When a product sprint got tight, and product sprints always get tight, the design system work lost. The engineers meant to maintain Uplift’s repository, fix bugs, and keep the Figma library in sync with production code were reassigned, consistently. Nothing protected that capacity.

The drift ran in both directions. Without reliable design tokens, developers used eyedroppers to sample colors and estimated spacing values. And decisions made during delivery rarely traveled back to design, so the Figma library collected manual variations that approximated what the codebase had become instead of representing it. Teams found the drift when it hurt delivery, concluded, correctly, that using Uplift added risk, and stopped.

That protection couldn’t be informal. It had to be organizational.

The argument

My case to engineering leadership was that the design system isn’t a development project that occasionally needs UX input. It is a UX project that happens to produce code, so its health is a design concern before it is a delivery concern. I made it in the terms tech leadership measured, quality and performance. A system teams had opted out of produced subtle inconsistency and translation drift, and every custom component built outside it was engineering time that didn’t need to be spent.

The deeper argument was strategic. The ecommerce sites were already being templatized across brands, with distinct identities kept through token-driven theming, and that only worked if the design system was mature enough to carry it. I pushed to mature Uplift quickly because the window wasn’t unlimited.

Building credibility

I hired Conor Sixta as DS Design Lead in August 2023. He carried the token and governance work well before the engineers moved under UX around mid-2024.

Getting the DS engineers to report through me took about six months, and most of that wasn’t spent arguing. It was spent finding the right person to argue with. TVH has a large, mature tech organization, I led a peer org, and UX didn’t traditionally have front-end engineers on its teams. The risk wasn’t a no. It was that the idea would sit politely in a queue.

What broke it open was Engineering Director Jennifer Dennon, who understood what I was trying to do and championed it internally. Then something more valuable happened: she pushed back on how I was planning to run it. Her feedback on how contribution workflows would meet engineering review, and on what “stable” meant to an engineer, made the model significantly stronger. It wasn’t a governance process designed by UX and handed to engineers. It was stress-tested by someone who understood both sides, and when other development teams saw that engineering fingerprints were on it, the conversation changed.

Once the engineers were under UX, the repository stabilized almost immediately. That was the straightforward result of protected capacity, insulated from product delivery reprioritization. Stability wasn’t enough to rebuild trust, though. Teams needed a way to participate in the system rather than just consume it, so governance had three parts: a contribution model that let product teams propose components, engineering review and approval before anything shipped to the shared library, and documentation standards that made the system navigable for designers and engineers. Jennifer and I shaped that model; Conor took it over, ran it, and completed it.

The DS Dev team also ran two-week sprints that it scoped and timelined itself in JIRA. Priority went to bugs found by design or QA, revisions to existing patterns, and new patterns for active redesigns. The team that owned the system also owned what it worked on next.

One source, several audiences

The tokens had lived inside Figma Variables, tool-locked and invisible to version control. I started the Tokens Studio setup and scoped Bitbucket as the JSON source of truth. Conor migrated the tokens into a git-managed JSON file that Tokens Studio consumed and manipulated, so a token change was versioned and reviewable like any other code change, and Figma and the codebase fed from the same source. Conor refined and completed the token work. I selected Storybook as the developer reference, living documentation of components in their coded state. Zeroheight gave product, marketing, and other teams a readable reference without opening a repository, and marketing used it to design non-ecommerce sites directly, which extended Uplift’s reach without creating a parallel system. Figma Libraries served UX design. I also selected Zeroheight and the Figma Dev Mode workflow. Original Figma, Bitbucket, and token-governance artifacts have not been independently inspected.

The multi-brand challenge was handled through scoping: the primary brand governed and stable first, secondary brands documented with a clear path into the system incrementally. Solving all three at once would have diluted the effort enough to stall the trust recovery on the primary product.

What teams were using

By my own count when I left TVH in October 2024, Uplift had gone from one development team using the coded system, mostly without UX guidance, to approximately four brand tech teams: three primary and one secondary. That covered approximately 15 front-end developers. These are my historical internal adoption counts, not an independently audited census or evidence of a completed cross-brand customer platform.

The token architecture was probably the most concrete day-to-day improvement for developers. Before stable tokens, they sampled colors from design files with eyedroppers and estimated spacing, and that compounded across hundreds of components into inconsistency no single person caused. With mature tokens and Figma Dev Mode, developers could extract consistent CSS directly from the design file. The guesswork left the process.

View larger (opens the full image)

The clearest proof of adoption wasn’t new UI. Most of what changed looked no different. The Product Details Card, a monolithic, hard-to-extend component used across nearly every product page, was rebuilt as a reusable, configurable pattern, for two reasons at once: current performance metrics and the expansion the roadmap already called for. It became the largest single acceptance the governance model produced, a full rebuild that Product signed off on rather than routed around. Historical team/JIRA reporting put documented pattern updates at approximately one sprint instead of three, after the pattern answered questions that used to go back and forth in tickets. Original reports have not been independently inspected, and the figure does not describe whole-system delivery or time to release Innovatis.

I also contributed directly to the PDP design while leading it, working on low- and mid-fidelity product specs/details tabs alongside Geoffrey Van Meirhaeghe and Suzan. Suzan finalized both displayed eShop Neo and Innovatis screens. I made the comparison composite from design-file exports; these are not authenticated production captures. The Innovatis redesign and cross-brand templating I led remained in progress when I left in October 2024.

How much time the template saved per brand is hard to state cleanly. TVH, TVH America, and Bepco ran on different data backends with different page layouts, so extending the template wasn’t a like-for-like copy job. What’s true instead is qualitative: each brand’s build started from a common token and component foundation rather than from nothing. TVH’s direction was consolidation, but the process was cautious, and three brand experiences were not going to merge overnight.

Across the broader UX organization, my team reached a maximum of 11: seven designers, a researcher, and three design system engineers. That was its largest composition, not a constant staffing level. Conor’s August 2023 hire gave Uplift dedicated design leadership early; the hire itself was not evidence that the system was stable or fully adopted, and system work remained ongoing during my tenure.

The larger argument

The lesson I’d take from Uplift to any organization with a struggling design system: before you invest in better documentation, tooling, or design quality, ask where the engineers are. Not which engineers. Where they sit, who sets their priorities, and what happens to their backlog when a product sprint gets tight.

A design system is infrastructure, and infrastructure needs protected maintenance capacity. When that capacity is embedded in a delivery organization, it loses to delivery pressure, not because anyone made a bad decision but because delivery pressure is immediate and visible in ways infrastructure debt isn’t. The system I inherited wasn’t bad work. It was good work that had been structurally set up to fail, and fixing it meant fixing the structure first.

The strategic part is what I find most interesting in retrospect. Consolidating three international brands into one platform isn’t primarily a technology challenge. It’s a design language challenge, and the work was building the foundation for a move the company hadn’t fully committed to making yet.