FIN Analysis Process | Documentation

Last updated: Oct-13-2025

Caroline (or Jackie if Caroline's unavailable) exports the previous Sprint's worth of conversation IDs that got a Negative CX score (or neutral if no negatives) plus three positive rated IDs from the same timeframe and creates a JIRA ticket for the person whose turn it is that sprint to do the analysis.
The export looks something like this:

The person doing the analysis does the next steps.
Export the latest Fin in Doc Site Conversations CSV file and rename it fin_conversations.csv.
Put both CSV files in the same folder as fin_csv_to_html2.py (found in cld_docs/fin) and run the script with the conversation-ids option, e.g.:
Take a look at the HTML that's produced to try and understand the conversations and form your own opinions about what could have been improved (docs or Fin guidance).
Use the Fin Conversation Analyst GPT to help you:
1. If a conversation needs translating, either try using Gemini in the spreadsheet, or in the GPT use the following prompt: "Only translate this conversation exactly. Do not perform any analysis yet." Copy the conversation row from fin_conversations.csv (filter on Conversation ID)
  
  and paste it into the GPT after the first prompt (it can be a separate prompt).
2. For another perspective on improvements to make, use the following prompt: "Analyze this conversation and suggest doc improvements and Fin guidance.", then copy and paste the conversation from the CSV file (as in step a).
Create JIRA tickets based on your own analysis and the GPT's analysis for anything that we can improve in the docs, or any guidance that we could provide Fin. Link the JIRA tickets to the JIRA ticket for that Sprint. For guidance tickets, prepend the summary with "Guidance:" and assign them to Caroline. For docs tickets, assign them to the relevant member of the docs team (or just DocTeam). For all tickets, add the label "FIN".