How to Analyse Old WhatsApp Conversations: What to Check in Your First Audit

Recording the audit’s coverage

The most important, and most often skipped, step in a first audit is recording coverage. Coverage tells you what data the findings rest on. Write it down in a few lines:

1. Date rangeWhen is the oldest conversation from, and the newest? The history stored on a phone can differ for each number: the phone was replaced, chats were deleted, a backup was never restored. Check the actual range. Do not guess.
2. Conversations includedHow many chats are in the analysis? Have group chats, personal chats and staff-to-staff chats been separated out? For business findings, you only want customer conversations.
3. What is not includedWhat cannot be seen? Phone calls, in-person meetings, conversations on other numbers or channels, deleted chats. Record these openly so anyone reading the findings knows: “This is what was on WhatsApp, not the full picture of sales.”

A coverage note looks simple, but it decides how trustworthy the whole audit is. If the history covers the last three months, then “customers complain most about delivery” is true for those three months. When the season changes, the picture may change too.

The first three analyses

Do not ask too many questions in your first audit. Three are enough: they produce results quickly and let you check that the method works.

1. Unanswered questionsThe customer asked something and got no reply. Old history usually has many of these, but most are no longer relevant. So split the result into two groups: the last month (you can still go back) and older ones (for statistics only).
2. Recurring objectionsWhat makes customers hesitate most: price, delivery, quality, terms? This is the most valuable analysis on old history, because several months of data filter out one-offs and show real patterns.
3. Open commitmentsPromises staff made to customers with no sign they were kept. Old promises may be long overdue, but their number and type reveal gaps in the process.

Checking results against samples

Any analysis, manual or AI-based, can get things wrong. In your first audit, check the result before trusting it: pick 10–20 chats per question and compare what was extracted with the original messages. The table below is an illustrative example of typical error types.

1. Error: a thank-you counted as an unanswered questionExtracted: “unanswered question”. Original: the customer’s last message was “thanks, that’s clear”. Fix: add a condition to the criterion — “must contain a question or expectation”.
2. Error: a phone call was missedExtracted: “promise not kept”. Original: the agent wrote “I’ll call you”, then the customer wrote “we spoke, thanks”. Fix: accept the customer’s later confirmation as evidence.
3. Error: wrong objection typeExtracted: “price objection”. Original: “delivery is free elsewhere”. That is about delivery terms, not price. Fix: refine the objection categories.
4. Correct resultExtracted: “open promise: send catalogue”. Original: “I’ll send the catalogue tomorrow”, then no file. Result confirmed.
5. UnknownExtracted: “sale lost”. Original: the customer simply went quiet. Fix: record silence as “unknown”, not as a loss.

After the check, write a short note for each question: how many of the samples were correct, and what kinds of error appeared. That note is the basis for your second audit.

An audit card: one-page result

Collect the first audit’s result on a one-page card. The card below is an illustrative example: the first audit at an appliance store.

• Coverage: 1 July – 30 September, customer chats; group chats and staff conversations removed.
• Not included: phone calls, in-store meetings, conversations on the second number.
• Question 1, unanswered questions: a handful in the last month, mostly evening messages; older ones for statistics only.
• Question 2, recurring objections: delivery time and return policy; price comes third.
• Question 3, open commitments: “I’ll send the catalogue” and “I’ll check with the warehouse” are most often late.
• Sample check: 15 chats per question; the main error was counting thank-you messages as unanswered questions.
• Next step: fix the criteria, run a second audit on October’s chats, add a new question: “return requests”.

The card is both a short report for managers and the starting point for the next audit. The coverage line sits at the very top, so nobody mistakes the result for the whole history.

Moving on to the second audit

The goal of a first audit is not a perfect result but a method that works. Do two things before the second one.

1. Fix the criteria errorsAdd what the sample check found to your criteria. Add the thank-you exception to “unanswered question”, and customer confirmation to “open promise”. Small fixes make the result noticeably more accurate.
2. Choose the next questionAfter the first three, add one question specific to your business: “buyers silent after a viewing” in property, “no enrolment after a trial lesson” at a training centre. One new question per audit is enough.

Choose the second audit’s period in advance too. Rather than re-analysing old history, look at the month of conversations after the first audit. With corrected criteria, the new result also shows the effect of changes you made after the first one.

We show which old customers are worth returning to after the audit in how to reactivate old WhatsApp leads. The difference between statistics and content analysis is covered in WhatsApp statistics vs conversation analytics. Questions for later audits can be picked from 15 business questions to ask your chats.

How to check your own chats with Chatdue

Chatdue connects to your company’s WhatsApp number by QR code through Linked devices and loads older conversations as well. Check which period is included as part of your first audit: it depends on the history stored on the phone. Voice messages, documents and images are converted to text and included.

Each chat gets a summary, what the customer wanted, objections, sales stage and other fields. By creating your own analytics fields, you can add the three questions above and the ones that follow. For the sample check, compare each selected chat’s summary and fields with the original messages in WhatsApp. You can also export the full result to Excel.

Frequently asked questions

1. Is an audit useful if part of the history is missing?Yes, as long as the coverage is written down. Three months of history is enough to show recurring objections and process gaps. Just do not present the result as “all history”, and keep seasonal differences in mind.
2. What should I note if voice messages are not included?State it plainly in the coverage note: “voice messages not included”. If your customers send a lot of voice notes, that gap can change the result significantly, because objections and questions may sit exactly there. In that case, a tool that transcribes voice messages makes the audit far more complete.