Conversation Closing Audit Checklist: What Must Be Recorded Before Closing

Closing a conversation is not the same as stopping replying. It is the moment when the conversation moves from open to resolved, and when it is decided what stays on record so that someone else can understand what happened without asking anyone. Most customer service audits look at response time or message volume; almost none look at the close. And the close is where information gets lost: when the customer comes back twenty days later, nobody knows what was promised, who handled it, or why the conversation was considered finished.

This page is a checklist. You go through one closed conversation at a time, in order, and each item says what to check and how to know it is done right. You do not need a special tool: you can do it on the WhatsApp Business inbox, by exporting the conversation or by reviewing the team's history. What you do need is a single standard, because if everyone audits differently, the results cannot be compared with each other.

Before you start: what is a close and what is not

A conversation is closed when the team decided it requires no further action. That includes three different endings, and the audit has to distinguish them because each one requires a different record: close by resolution (what the customer needed was answered), close by referral (the case went to another area or another person and that area replied) and close by abandonment (the customer stopped replying and the team considered the matter finished). The fourth case, the one that causes the most problems, is the false close: the conversation was marked as resolved but in fact a question was left unanswered.

Stage 1 — Verify that the close is real

• The customer's last question has an answer. A partial answer or an 'I'll look into it and let you know' with no follow-up does not count: if the team promised to look into something, the later message with the answer must be there.
• The customer confirmed or gave a sign of satisfaction. A 'thanks', a 'perfect' or a 'sure' count. If there is no sign at all, the conversation is not closed by resolution: it is closed by abandonment, and it is recorded as such.
• No pending request was left inside the thread. Review the customer's last five messages and verify that none asks for something that was not answered.
• If the close was by referral, there is evidence that the destination area replied. Referring is not closing: the close is done by whoever receives, not by whoever passes the case on.
• If the close was by abandonment, there is a timestamp of the customer's last message and at least one later reactivation attempt. Without a reactivation attempt, it is not abandonment: it is a forgotten conversation.

Stage 2 — Review what was recorded

Here is the heart of the audit. The log is not to fill out a form: it is so that the next person who opens that conversation understands the case in thirty seconds. The practical test is simple: if someone new joins the team tomorrow and reads only the log, can they pick the case back up without asking anything?

Stage 3 — Review the form of the closing

• There was an explicit closing. The agent said something that marks the end ('let's leave it there', 'write to me if anything comes up', 'I'll send you the info in a bit'). A thread that simply stops having messages is not a closing.
• The closing does not promise more than the team can deliver. Check that no closing phrase includes a deadline or an action that is not upheld afterward.
• If the case was referred, the customer knows where it went and what to expect. A plain 'I'll pass it to the area' leaves the customer without a reference.
• The tone matches the rest of the conversation. A dry closing after twenty warm messages reads as a cut-off, not as a closing.
• No question that the next step depends on was left open. If the customer has to decide something in order to move forward, the conversation does not close: it stays on hold, which is another state.

Stage 4 — Sampling: how many conversations to review

Auditing every closed conversation does not scale. What works is a fixed weekly sample, with the same size every week, so that the numbers are comparable. The formula is simple and does not depend on the tool:

Conversations to review per week = (conversations closed per week × 10%) rounded up, with a floor of 10 and a ceiling of 40. If the team closes 60 conversations per week, 10 are reviewed. If it closes 300, 30 are reviewed. The floor of 10 prevents conclusions from small samples; the ceiling of 40 prevents the audit from eating half a workday.

Stage 5 — Measure the audit result

An audit without numbers doesn't hold up over time. These are the indicators that come straight from the previous list, all calculable with what has already been recorded. No new tool is needed: counting over the sample is enough.

What to do when something fails

Each failure on the list has a different correction, and confusing them means one gets fixed and another breaks. These are the responses according to the type of problem:

• Missing reason for contact: it isn't solved by asking people to 'write better.' It's solved with a mandatory field at closing time and a short list of reasons to choose from, instead of free text.
• Missing commitment made: the problem is that the promise is made in the chat and not recorded. The correction is that every promise of a deadline or action is copied to the record at the same moment the message is written.
• The closure is false: you have to review the definition of 'resolved' the team uses. If the criterion is 'the agent stopped responding,' any conversation closes itself. The criterion has to be 'the customer's question has an answer.'
• Nobody knows who resolved it: this happens when the case passed through several hands without a formal handoff. It's corrected with a handoff that includes a new owner before the previous one lets go of the case.
• The record is completed late: it's usually a consequence of the team closing conversations in bulk at the end of the day. The correction is to record at closing, not at archiving.
• The indicator worsens suddenly in one week: before changing the process, check whether the person auditing changed or whether the sample size changed. A change of auditor is more noticeable than a real change in performance.

What a well-audited closure looks like

A closure that passes the audit has five visible things: a reason written in one sentence, a concrete response recorded, a commitment with a date if there was one, an identified owner, and a result tag. In addition, the conversation ends with a message from the team that marks the end, and if the case was referred, the customer knows where it went. None of that requires a special tool: it can be done on any WhatsApp Business inbox that allows tagging conversations and leaving an internal note.

If the team uses a tool with AI suggestions on the official WhatsApp API, such as Wando, the record is completed the same way: the AI suggests the response and the team approves it, but the reason, the commitment, and the closure tag are set by a person. The audit does not change because there is AI in the middle; only the point where it is reviewed changes. If you want to see what plans are available today, check the current plans at wando.online/pricing.

Common mistakes when auditing closures

• Auditing response time instead of the closure. They are two different things: a conversation can be closed quickly and badly.
• Taking the archive date as the closure date. It distorts all reports by time slot and by duration.
• Reviewing only the conversations the team chooses to show. The sample must include those with risk signals, not only the tidy ones.
• Turning the audit into an individual evaluation. The goal is to find process failures; if it is used to grade people, the team starts hiding conversations.
• Closing due to abandonment without an attempt to reactivate. That is not abandonment: it is a conversation without an owner.
• Defining 'resolved' so broadly that almost everything falls into the category. If 95% of closures are 'resolved', the indicator has stopped measuring anything.

Template to check off, conversation by conversation

Copy this list and use it once per audited conversation. It is enough to mark yes or no on each line; the indicator comes from counting the yeses over the total.

• The customer's last question has a concrete answer.
• There is a sign of agreement from the customer or it is recorded as abandoned.
• No order was left pending in the last five messages.
• The reason for the contact is written down.
• What was resolved is recorded, with the concrete answer.
• If there was a promise of a deadline, price or action, it is recorded with a date.
• It is identified who resolved the case.
• The conversation has a result tag (resolved, referred or abandoned).
• The date used is that of the last message, not that of archiving.
• There was an explicit closing message from the team.
• If it was referred, there is a record that the destination area responded.
• If it was abandoned, there was at least one reactivation attempt.