Patrick Dwyer

In May 2025, Charter Communications and Cox Communications announced a $34.5 billion merger — one of the largest deals in cable industry history. When it closed in August 2026, the agreement meant every customer-facing Cox product, across 45 states, needed to start presenting itself under Charter's consumer brand, Spectrum, on a hard external deadline: mid-September 2026.

I work on the shared component library that every Cox digital property is built from — the design system behind both our marketing sites and our account-management tools. The merger didn't give us a new coat of paint to apply; it gave us a deadline to hit and a brand we didn't design, and it fell to me to figure out whether our foundation could take it.

Before proposing anything, I needed to know exactly what we were working with. I audited how our design tokens were actually structured and found two compounding problems.

First, "brand" wasn't a clean, swappable concept in our system. Cox Residential and Cox Business styling was mixed directly into components rather than abstracted behind it, so there was no lever to pull for a third brand — just components that would each need to be found and edited by hand.

Second, the tokens we did have were a mess to work with on their own terms. Naming had grown organically over years, with no shared convention behind it — similarly-purposed tokens were named differently depending on who wrote them, some encoded their value directly in the name, others didn't, and there was no reliable way to guess a token's name from what it did. That made even routine maintenance slower than it should have been, and it meant any new system we built risked inheriting the same confusion at a larger scale.

I also had to get ahead of a wrinkle Charter's own brand introduced: Spectrum wasn't one design system, it was two, split by use case —

• Spectrum Nova, used on marketing and acquisition pages
• Spectrum Kite, used in customer-management tools like My Account

Each came with its own component library and its own token set — related, but not interchangeable. Any fix I proposed had to account for two target systems, not one.

With the shape of the problem clear, I worked out what a real fix would take. The core idea: separate what a design decision means (a button's primary color, an error state's text color) from what that decision actually resolves to for a given brand. If components referenced meanings instead of hardcoded values, swapping brands would become a matter of repointing meanings — not rewriting components one by one under deadline pressure.

But a swappable architecture is only as good as the names it's built on, so before I mapped a single token to Spectrum, I tackled naming head-on. I researched established approaches to token taxonomy and standardized on the naming framework popularized by Nathan Curtis of EightShapes — a structured, layered approach ("Naming Tokens in Design Systems") that's become something of an industry reference point for exactly this problem. Adopting it gave us a predictable, self-documenting pattern: given a token's purpose, anyone on the team could now guess its name correctly, and given a name, they could guess its purpose — something that had been genuinely impossible under the old system.

I also settled a question that had never been formally answered on our team: where does a token actually live? I established Figma's Token Studio plugin as the single source of truth for the entire system — every token, in every brand, authored and governed there first, then synced downstream into code. That gave designers and engineers one shared, authoritative place to define a value, instead of tokens drifting out of sync between what a designer intended and what shipped.

That naming standard and source-of-truth decision became the scaffolding for everything after it. From there, I defined the actual token mapping — the translation layer from our legacy Residential/Business values into the new Spectrum identity — and because Nova and Kite each had their own slightly different token sets, this meant working out two parallel mappings side by side, both authored in Token Studio and built on the same naming convention, checking each against its respective component library rather than assuming one mapping would serve both.

I also made the case that this shouldn't be scoped to just our application. If we were going to solve this mapping and naming problem properly, other teams across the organization adopting Spectrum would run into the exact same problem — so I pushed for the tokens to be extracted into their own standalone package rather than living inside our codebase.

What shipped, over roughly seven months:

• Figma's Token Studio plugin established as the single, governed source of truth for every token, in every brand — the anchor point everything downstream synced from
• A standardized, industry-informed naming convention applied consistently across every token — the structural fix that made everything built on top of it legible instead of guesswork
• A restructured token architecture that centralizes brand identity in one swappable layer instead of scattering it through component code
• Two parallel, fully-mapped token sets translating our legacy values into Spectrum Nova and Spectrum Kite, each validated against its own component library
• The tokens published as a standalone npm package, versioned for org-wide consumption rather than owned by a single application
• A phased rollout behind feature flags tied to real merger milestones — early cosmetic "day-one" branding, then a unified account sign-out experience — so we could de-risk the cutover instead of flipping one irreversible switch
• Internal documentation tooling so designers and engineers could see exactly how Nova and Kite diverged, once the mapping was too large to track by reading raw config

By the time Spectrum branding needed to appear in former Cox markets, our platform was ready to present the new identity without a component rewrite — correctly split across marketing and account-management contexts, and reusable by any team in the organization that needed the same tokens rather than solving this independently. Having a single source of truth in Token Studio meant designers and engineers were never working from conflicting versions of a value, the naming convention alone paid dividends beyond the rebrand, and the phased, flag-gated rollout meant we could catch problems in production without the risk of a full-brand outage.