The Mint Content Library

A style guide built channel by channel

Mint didn't have one place that said how the product should sound. Terms, emails, push notifications, error states, even emoji use—each channel had its own quiet folklore, and every writer filled the gaps with their own judgment. Two other content designers and I set out to write it down.

Each entry followed the same shape: a principle stated plainly, then a real do and don't pulled from actual copy we'd written or rejected—words we'd use to sound optimistic without promising something we couldn't keep, and the ones we'd learned made customers feel patronized instead of reassured.

We put tone to the test

The system grew out of a standing weekly sync. Every few weeks we ran a working session pulling in whatever content questions product and marketing partners were actually asking us most often, then used that list to decide what the guide needed next and what old guidance to retire. Nothing got written because one of us thought it sounded right—it got written because the teams we serve asked the same question twice.

The clearest example was tone itself. We had a real disagreement about how direct a message should be when the news wasn't good—apologetic and soft, or plain and specific. Instead of deciding by seniority, we built 2 versions of a real in-product moment, one framed reassuringly and one framed directly, and tested both with actual users, tracking their reaction rather than asking them to rate it.

Both halves of the test taught us something. On the positive side, people needed to see the actual number in the headline before the message felt real—value that stayed abstract didn't land. On the negative side, the direct, less-cushioned version won by a wide margin; softening bad news read as vague rather than kind. Neither is what any of us would have guessed sitting in a room.

A guideline is just an opinion until it's tested

Our standard was simple: we based our style guide on what real users did when they interacted with the content.

Nobody's seniority outranks what a real user does when they see the copy.

Since 3 of us built it in parity, every rule had to be agreed upon by 2 people who could just as easily have written it themselves. The system that came out the other side held up better for it.

Mid-build when Mint went away

We had the terms, email, push, and data guidance drafted and in real use inside the team, plus the error-states and emoji frameworks close behind. The tone study had run its first round and pointed us toward a second, sharper question we never got to test.

Then Mint was discontinued, and we were told to stop the content library work so we could start the transition to Credit Karma. The library never got a formal publish moment—not because it stalled, but because the company had somewhere more urgent for the three of us to be.