Six Versions of the Same Button

Tuesday, July 28, 2026 · Jacob Byers

Six mismatched button styles beside a single consistent one

I used to think great design made great websites.

I came up as a designer before I was a developer. So I get it. I have been the person obsessing over a mockup at midnight.

But after a decade of building sites for marketing teams, here is what the work keeps telling me:

Content moves the needle. Design just has to not get in the way.

The mess I kept inheriting

There is a pattern I have seen at almost every client. I call it design creep.

It starts innocently. A new landing page needs a slightly different hero. A webinar page gets a tweaked button. The brand color list grows by one, then three, then ten.

At one client, I counted six variations of the same button.

Three different designers had passed through over the years. Every mockup I got handed had one small adjustment. A new color. A different border radius. A card that almost matched the other cards.

None of it was malicious. But my billable hours kept climbing, not because we were building anything new, but because I was babysitting an incoherent system.

The site was not designed. It was accumulated.

The uncomfortable part

Here is the thing nobody wants to say out loud.

Designers are rewarded for this.

When you are paid by the hour, tinkering is the job. A new mockup every week is income. A component that gets decided once and reused forever is a threat.

I do not think most designers do this consciously. Tinkering does not feel like padding. It feels like craft. You are never fully satisfied, so there is always one more knob to turn.

But the incentive is real. I once pushed a client hard to consolidate everything into one unified design system. It would have cut the design workload dramatically.

The designer did not want that. And I understand why. I was proposing to automate away a big chunk of her job.

I believe pushing for it is part of why that engagement ended.

That one stung. But it clarified something for me.

The test I use now

Every design decision gets one question:

Does this serve the whole, or is it just tinkering?

Do the cards match the border radius of every other card? Does this new icon actually belong in the system? Is this tweak in service of the story the site is telling, or is it busy work?

Most of the time, honestly, it is busy work.

A good design system ends the busy work. It says: the type scale is decided. The colors are decided. The spacing is decided. Nobody has to have an opinion about the button ever again.

Good design is knowing when to stop designing.

What I build instead

After that engagement ended, I started building what I call a design dictionary for every client.

It is the full set of decisions, made once: typography, type scale, colors, spacing, border radius, and the components built on top of them. Every knob you would ever touch on a website, defined and locked in.

The result surprised even me.

When the system is coherent, everything looks good no matter which components you grab. The owner stops thinking about design entirely and focuses on the thing that actually sells: content.

Will it win design awards? No.

Does it move the needle to the sale? Yes. And that is the job.

What deciding once looks like

This is easy to say and harder to prove, so here is this website.

One radius

It is zero. There is no radius scale, so nobody can pick a slightly rounder card.

One font

One web font, loaded once. Serif and mono fall back to the system and ship no bytes.

No blues

There is no palette of blues, only named roles. Nobody can add one more blue because there is nowhere to put it.

Four deleted

Four section types removed by absorbing them into sections that already existed. Zero pages rebuilt.

Adding a new component type costs four files, a menu entry, a preview image, and a permanent obligation to maintain it. Adding an option to something that already exists costs one line. The second is almost always right, and almost nobody picks it, because building the new thing feels more like work.

Then AI changed the math

Here is the part I did not see coming.

Once a design dictionary exists, AI tools can plug into it. My clients now generate full pages using Claude Code, inside their own system, with accessibility and SEO baked into the framework I gave them.

The dictionary is what makes that safe. Without it, an AI writing pages invents a new blue every time, and you get design creep at machine speed. With it, the machine has nowhere to drift to. The colors it can reach are the only colors that exist.

It has grown past web pages. PDFs, invoices, one-pagers, all generated from the same core structure, all looking like they belong to the same brand.

They do not need me for every page anymore. That is the point.

I would rather be the person who built the engine than the person you have to call every time you want to drive.

The real philosophy underneath

If you zoom all the way out, this was never really about design.

It is about incentives.

The hourly model rewards never being finished. Outcomes reward deciding things once and building on top of them. A design system is what outcome-based work looks like when you can actually see it.

The designer paid hourly is financially aligned with touching the knobs forever.

I am aligned with the opposite: decide it once, lock it in, and let your team ship.

That is the whole philosophy. Stop paying for motion. Start paying for the machine.

Want the decisions made once?

See what a design dictionary includes and what it frees your team to do without a developer.

We use cookies

We use cookies to improve your experience and analyze our traffic.