The ticket we stopped filing

Marcus and I ran a five-person web agency in Columbus for six years. We built marketing sites for software companies, clinics, a few restaurants, and one very particular climbing gym. Every site was Laravel, every site had a handful of Blade components we had written a dozen times before, and every site generated the same kind of ticket after launch.

“Can you change the headline on the pricing page?” “We need to swap the photo in the second section.” “Our new head of sales wants the button to say Book a demo.”

In our last full year, 34% of the tickets we closed were copy changes. Each one took a developer about twenty minutes once you counted the context switch, the branch, the review and the deploy. That is roughly 300 hours a year of skilled people changing words.

What we tried first

We tried a headless CMS. The clients could edit text, but every new section needed a schema in one place and a component in another, and the two drifted. We tried a hosted page builder for the marketing pages and Laravel for everything else. The clients liked it until the two halves of their site looked like two companies.

What we wanted was simple to describe: let the client edit the words and images directly on the page they are looking at, and have that edit land in the same Blade files we already wrote. No second system to keep in sync.

The rule that made it work

The idea that turned into Starter was a rule, not a feature. Every visible string, link, image and list in a section is a field, and every field has a default. If a client would want to change it, it is editable. If it is editable, it is in the YAML.

Once that was true, the copy tickets stopped. Not slowed down: stopped. Clients opened the page, clicked the headline, typed, and pressed Publish. We saw the change in the next pull request like everyone else.

What the developers got back

The part I did not expect was how much the developers liked it. Nobody enjoys changing “Get started” to “Start free” for the third time in a month. With copy out of the queue, the tickets that were left were the interesting ones: a new section, a faster page, an integration.

Where the time went instead

In the year after we moved our own clients onto the first version, our copy-change tickets went from 34% of the queue to under 2%, and the ones that remained were usually a request for a section that did not exist yet. We used the hours on two things: faster sites, and a better section library. Both of those became Starter.

That is still the whole pitch, really. Starter is not trying to replace your developers or your stack. It is trying to make sure the people who write the words never have to wait on the people who write the code, and the other way around.