Partner with Us|Become An AffiliateNeed a WhatsApp number? Get one in minutes — from $2/mo.WhatsApp number — from $2/mo.Get a number
TimelinesAI
Book a DemoLog inStart Free Trial
3 of 3

Tag every chat

Raw messages aren't analytics yet. Ask the assistant to read each chat once and add a few columns: what it's about, how the customer feels, which stage it's at, any deal value mentioned and any feature requested.

You set the rules, and the assistant applies them to every chat the same way. These columns power every chart in the next two guides.

Spot-check before you trust it. Ten random rows next to the original messages tell you quickly whether the rules need tightening.

Why it matters

Tagging 300 chats by hand takes days, and two people rarely tag them the same way. The assistant does it in minutes and stays consistent. A quick spot check of a few rows tells you whether to trust the result.

  1. 1

    Add the tags.

    Prompt
    Read each chat in messages.jsonl and add these columns to chats.csv: topic (pricing, demo, onboarding, delivery, billing, feature request or other), sentiment from -1 to 1, stage (new lead, demo, proposal, negotiation, won, lost or support), deal value in USD if one was mentioned, and the feature requested, if any.
    chats.csv with topic, sentiment, stage, deal value and feature columns added by the assistant
  2. 2

    Spot-check a sample.

    Prompt
    Show me 10 random rows with the customer's key message next to the tags, so I can check them.
  3. 3

    Tighten a rule and re-tag.

    Prompt
    Count "can you connect to X" as a feature request, not other. Re-tag and tell me how many rows changed.

    Happy with the tags? Keep both files: the next guide turns them into a dashboard.