Custom WhatsApp Conversation Analytics: 15 Business Questions to Ask Your Chats

What makes a good analysis question

A good analysis question has three parts. Miss one, and the result will be either empty or misleading.

1. PurposeWhat decision will you make once you see the result? “Information about competitors” is not a purpose. “We want to know whether our price list needs updating” is, and it narrows the question.
2. Inclusion ruleWhat in the chat counts as a “yes”? For example, “customer mentioned a competitor” covers only a specific company or brand name, while “it’s cheaper elsewhere” is recorded separately as an unnamed comparison.
3. Missing-data ruleIf the chat has no answer, the result must be “unknown”, not “no”. “The customer did not mention a budget” and “the customer has no budget” are different facts. Without this rule your report will treat silence as a denial.

Compare: “Are customers happy with us?” cannot be answered from a chat, because satisfaction is not directly visible. “Did the customer express dissatisfaction in the chat? Yes/no/unknown; topic; customer’s own words” can be checked. The 15 questions below are written in that form.

Five sales questions

1. 1. Was a quote sent and left unanswered?Result fields: quote sent (yes/no), quote date, the customer’s last reaction (question, objection, thanks, none), days since.
2. 2. Is a follow-up due?Result fields: the date or event the customer mentioned (“after payday”), that date converted to a specific day, whether it has passed. If the date is ambiguous, mark the field “unclear”.
3. 3. Who makes the decision?Result fields: the contact decides alone (yes/no/unknown), other people involved (partner, manager, accountant), the message that shows it.
4. 4. Do we have an open promise?Result fields: what was promised (price, catalogue, call), who promised it, by when, whether a message shows it was done. If there is no evidence, status is “not confirmed”.
5. 5. Is there a buying signal?Result fields: signal type (asked about delivery, payment method, availability or a specific date), the customer’s quote. A price question alone should not count, because people ask it while they are still just browsing.

Five questions on objections and the market

1. 6. Is there a price objection, and what kind?Result fields: objection present (yes/no), type (budget limit, comparison with another offer, unclear value, unknown), the customer’s own words, the rep’s reply.
2. 7. Which competitors do customers name?Result fields: competitor name (as written), comparison topic (price, timing, quality, terms), customer quote. A mention is not a loss to that competitor, so do not merge the two in one field.
3. 8. What do customers ask for that we do not offer?Result fields: requested product, service or term, whether an alternative was offered, the customer’s reaction. This is one of the most useful inputs for product and range decisions.
4. 9. What do customers compare us on?Result fields: comparison criterion (price, delivery, warranty, reviews, a recommendation), the message showing the customer raised it themselves.
5. 10. What is behind “I’ll think about it”?Result fields: vague objection present (yes/no), the open question that came before it, whether the rep asked for the reason. If the chat does not show a reason, the field stays “unknown”.

Five service questions and a result template

1. 11. Was the customer’s issue resolved?Result fields: the customer’s problem, the answer given, resolution status (resolved, still open, not enough evidence), the customer’s confirmation message.
2. 12. Did the customer come back with the same issue?Result fields: repeat contact (yes/no), dates of first and repeat contact, topic. A repeat contact often means the first answer was incomplete.
3. 13. Was context lost when the chat changed hands?Result fields: handover happened (yes/no), the new rep asking a question already answered, customer reactions like “I already told you”.
4. 14. Did an agreed date change?Result fields: original date, new date, who proposed the change, reason. Useful for spotting delays in delivery, appointments and installation work.
5. 15. Was an order or registration left unfinished?Result fields: how far the customer got (asked for details, received a price, asked about payment), what was missing, date of the last message.

When you write your own question, use this template. For each result field, define four things:

• Field name: short and unambiguous, such as “price objection type”.
• Field type: yes/no/unknown, choice from a list, free text, date or amount.
• If it is a list, the options: “budget”, “comparison”, “value”, “unknown”.
• Evidence: a short quote from the message behind the result. Without it, you cannot check the result.

If you want to score sales conversations, the rules in our sales chat audit checklist combine well with the questions above. Counting rules for needs questions are in customer needs analysis from conversations. And WhatsApp statistics vs conversation analytics explains how counts differ from content analysis.

How to run this on your own chats with Chatdue

Chatdue comes with ready-made fields for every chat: summary, what the customer wants, lead temperature, sales stage, sentiment, objections, competitors, amounts mentioned, agent score and a suggested next message. Your company chooses which analytics fields to use and can create new ones for its own questions, like the ones above.

It connects to your company’s WhatsApp number by QR code and loads older conversations too, so you can see first results on your existing history. Voice messages, documents and images are converted to text and included. You can filter results and export them to Excel with every field. One limitation to keep in mind: the analysis only sees what is in the chat. It does not know your till or warehouse data.

Frequently asked questions

1. How specific should a custom analysis question be?Specific enough that two different people reading the same chat would give the same answer. “Is the customer satisfied?” will produce disagreement. “Did the customer write thanks or praise in the chat?” will not. One question should ask about one thing.
2. What should the result look like when the chat has no answer?“Unknown”. It is the most honest form of result and should be counted separately in the report. A high share of “unknown” is useful in itself: it tells you reps are not asking that question, or the information stays on another channel.