Getting more from Fin analytics with conversation evidence

Your Fin resolution rate has moved since last week. Before you change an article or adjust a handoff rule, you need to know which conversations explain the change.

That is where a useful weekly review starts. Choose one question, find the relevant threads, and use them to decide what to investigate next.

Start with the definition behind the number

Intercom distinguishes between confirmed and assumed resolutions. A confirmed resolution includes a customer's positive response. An assumed resolution can follow when the customer leaves after an answer without asking for more help. See Intercom's outcome definitions for the details.

Keep that distinction visible in your review. A customer saying the answer helped gives you different evidence from a customer going quiet. Silence alone does not establish whether the problem was solved.

Use the definition and filters in the report you are reading. Save the date range, channel, and conversation group alongside the result so you can repeat the comparison.

Ask a question narrow enough to answer

"Why did resolution improve?" can send you through unrelated conversations for an hour.

"Are customers getting usable password-reset instructions after Tuesday's article update?" gives you a place to start.

Consider a hypothetical example. You added instructions for customers whose reset links expire. The overall resolution rate rises that week. That is encouraging, but the week may also have brought more simple account questions.

Review password-reset conversations from before and after the edit. Check whether customers with expired links received the new instructions and whether they could complete the next step. Keep ordinary password resets separate from account-access cases that require identity checks.

The result may be more specific than the headline metric. Perhaps the instructions now work for expired links, while customers who no longer have access to their email still need a clearer handoff.

Keep a short evidence record

For each relevant conversation, record:

• The customer's request and the reported outcome.
• The answer or action that matters to your question.
• The source guidance, where available.
• What the conversation confirms and what remains unknown.
• A link to the thread.

Write "customer says the new link worked" rather than "good interaction." If there is no reply, record that. If completion requires an account event, check that event before calling the task complete.

A small sample can reveal a problem worth fixing. It cannot establish how often that problem occurs across all Fin conversations.

Turn the finding into a follow-up

Assign the next step to the person who can change it. Missing instructions go to the content owner. A supported answer that uses the wrong eligibility rule needs a guidance review. An incomplete transfer needs someone to check the handoff workflow.

After the change, review the same kind of conversation again. Compare similar customer requests and note other changes that could affect the result. A better rate after an article edit is evidence to investigate, not proof that the edit caused it.

CraftCX connects quality findings to their source conversations so your team can inspect the evidence and choose a follow-up. Explore Support Quality to see how that review fits alongside your Fin reporting.