Brand Voice
Guardian.

Brand voice is hard to transfer

At most companies, brand voice lives in one or two people's heads. You can write guidelines, run workshops, add a style guide to the shared drive. But when someone new joins, a freelancer needs to turn around three emails by Thursday, or the team scales faster than onboarding can keep up -- the voice drifts. Gradually at first, then noticeably.

The review process that catches this drift is expensive. It requires the one person who truly understands the brand to read everything before it goes out. That person becomes a bottleneck. The feedback they give is often subjective, hard to act on quickly, and inconsistent across channels.

Why Mailchimp's guidelines were the right foundation

Mailchimp publishes one of the most detailed and thoughtful content style guides available publicly. It covers tone and word choice, but also specific rules by channel -- email, social, blog, paid ads, product writing -- with concrete examples of what to do and what to avoid.

That depth made it the ideal foundation for this version of the tool. A brand voice evaluator is only as good as the guidelines it runs on. Most companies don't have documentation this specific, which is itself a finding worth sitting with. The Mailchimp guide also includes a words-to-avoid list, plain English principles, and active voice rules that translate cleanly into evaluable AI criteria.

Mailchimp's guidelines are public, which makes them useful for a portfolio demonstration. But the architecture here is the point. A team with their own internal style guide -- voice pillars, channel rules, banned phrases -- could swap in that documentation and have the same tool running against their actual brand in an afternoon. The hard work is having the guidelines. The tool just makes them accessible at scale.

Encoding the brand, not just describing it

The core challenge was editorial, not technical. How do you translate nuanced, qualitative brand guidance into something an AI can evaluate consistently? Carefully, and with a lot of rewriting.

Mailchimp's guidelines are comprehensive but written for humans. The first phase of this project was translating that documentation into precise, evaluable criteria for each channel. Email guidelines became a specific checklist: sentence case subject lines, single clear CTA, lead with what the reader needs to know. Twitter rules became measurable: 280 characters, no shortened spellings, one thought, no engagement bait. Each channel has its own evaluation logic built in.

What the tool actually does

A marketer selects their input channel, pastes copy, and chooses an output format for the rewrite. The tool evaluates against Mailchimp's voice guidelines and returns a score across four criteria -- tone and personality, word choice, clarity and readability, and channel fit -- with specific feedback quoting the submitted copy directly.

It also generates a rewrite in whichever format was requested: email subject line, social post, blog intro, customer story, paid ad, and more. The house rules feature lets users add brand-specific overrides on top of the Mailchimp defaults -- so a team that never uses the word "seamless" can encode that once and have it applied to every evaluation. The tool is designed as a starting point for revision, not a verdict. Human judgment still leads.