Rethinking the chatbot: less friction at cancellation

The problem

The chatbot that was supposed to make support easier ended up adding to the frustration. The main friction points:

• Inconsistent language throughout the flow
• Translation errors and no real localization
• Difficulty completing simple actions
• No clear guidance at decision points

Analysis

I mapped the cancellation flow end to end and pinpointed the critical issues using Nielsen’s heuristics and content analysis:

• Inconsistent terms for the same concept — the flow mixed filiação, associação, and assinatura: three different Portuguese words for “subscription”
• Literal translations that broke natural understanding in Portuguese
• No user control — no clear way to go back or exit
• Messages that described the problem without helping resolve it

The proposal

I worked on three fronts:

Clarity and consistency — I rewrote the microcopy to remove ambiguity and standardize the language across the whole flow, so the person always knew where they were and what would happen next.

Localization — I adapted the content into real Brazilian Portuguese: not just the translation, but the cultural context, the rhythm of the sentence, and how natural the text felt.

Tone of voice — I defined a closer, more guiding style of communication, consistent with the brand’s identity.

Yuzu — a bot persona

The centerpiece of the proposal was creating a persona to give the chatbot an identity: Yuzu, the pet cat of Hime, Crunchyroll’s official mascot.

Yuzu is warm, attentive, and curious — with a light, friendly way of communicating that never gives up clarity. The tone balances the brand’s playfulness with the directness support demands, creating a more human experience without losing focus on solving the problem.

The persona worked as a guide for every linguistic decision, keeping things consistent from the start of the conversation to the very end — right down to the sign-off:

A product decision

One of the most interesting moments of the project was thinking through the balance between easy cancellation and business goals. The decision was clear: prioritize transparency and user control, offering alternative paths — like a discount — without blocking or complicating the main action.

Expected impact

With the proposed changes, the impact hypotheses are:

• Higher CSAT in the support flow
• Fewer cancellations driven by frustration with the experience
• More trust in the chatbot as a channel that actually resolves things
• More qualified retention — people who stay because they want to, not because they couldn’t get out