Inbox Audit Checklist: Unassigned, Untagged, and After-Hours Unanswered Conversations

A WhatsApp Business inbox doesn't break all at once: it degrades. First there are three unassigned conversations that nobody sees because they're mixed in with the rest. Then there are ten without a label, so they can't be filtered or measured. In the end there are twenty that went unanswered on a Sunday and nobody found out until the customer wrote again, angry. This audit is meant to find those three failures before they pile up. It's not a review of the quality of what was written: it's a review of whether the work system left conversations hanging.

The audit is done on the real inbox, not on what the team thinks it does. That's why the checklist starts with a count, continues with a classification and ends with a sample of conversations read one by one. If you don't feel like doing the full work, at least do block 1: in twenty minutes you'll know whether you have a volume problem or a process problem.

Before you start: what you need to have on hand

• Access to the inbox with permissions to see all conversations, not just your own. If each agent only sees their own queue, the audit can't be done.
• The list of active agents with their real working hours (not the hours published on the website: the ones they actually keep).
• The business hours the business communicates to the outside, on the website, in the WhatsApp Business profile and in the away message.
• The labels the team says it uses. If there's no written list, the first finding of the audit is already there: nobody can consistently use a labeling system that only exists in one person's head.
• A specific date range. Audit a full week, Monday through Sunday, not a single day: out-of-hours problems don't show up if you only look at Tuesday morning.

Block 1: unassigned conversations

An unassigned conversation is one that doesn't have a named person responsible for it. It can be open, it can have a reply, but if nobody has it assigned, nobody is going to follow up. The most common mistake is confusing 'unassigned' with 'unanswered': there are conversations answered three days ago that still have no owner, and those are the ones that fall apart when the customer writes again.

• Count the total number of open conversations in the audited range.
• Count how many of those don't have an assigned agent. Note the number and the percentage of the total.
• From the unassigned ones, separate those that never got a reply from those that were answered but were left orphaned.
• For each unassigned conversation, look at the customer's last message: is it a question waiting for an answer or is it a 'thanks' that closes the topic? Only the former count as a failure.
• Check whether there's a time pattern: are the unassigned ones concentrated in a shift, on a day or at the moment when a high volume comes in?
• Verify who can assign. If only one person on the team has permission to hand out conversations, that's the bottleneck and there's no need to keep looking.

Block 2: conversations without a tag

The tag is not decoration: it is what allows you to filter, measure and find. A conversation without a tag is a conversation you will not be able to locate in two weeks. The problem is not the tag itself, it is that without tags you cannot answer basic questions: how many price inquiries came in, how many converted, how many were forwarded to another area.

• List the tags that exist today in the system. If there are more than fifteen, the problem is not the lack of tags but the excess: nobody uses fifteen tags consistently.
• Count how many conversations in the audited range have no tag at all.
• Of those that do have tags, count how many have two or more tags that contradict each other (for example, 'resolved' and 'pending' together).
• Check whether tags are applied in the moment or at the end. If they are only applied on closing, conversations that remain open are never tagged.
• Review whether there are tags created by a single person and used by no one else. Those are personal tags disguised as system tags.
• Verify that tags can be filtered in the inbox. A tag that cannot be filtered is dead data.

Block 3: out-of-hours reply

This isn't about answering at three in the morning. It's about knowing what happens when someone writes outside business hours: whether they get an auto-reply, whether they end up in the queue for the next day, or whether they get nothing at all. All three options can be fine; what can't happen is that nobody knows which of the three is occurring.

• Define the real business hours and compare them with the published ones. If they don't match, that's the first finding, before looking at a single conversation.
• Count how many messages came in outside business hours during the audited week and what percentage they represent of the total.
• For each out-of-hours message, check whether there was an auto-reply. If there is one, read it: does it say when they will respond, or does it just say 'thanks for writing'? The second one is useless.
• Measure the time until the first human response in those cases. It doesn't need to be exact to the minute: it's enough to know whether it was first thing the next day or whether two days went by.
• Check whether out-of-hours conversations get assigned to someone or fall into the same pit as unassigned ones.
• Verify that the auto-reply doesn't also trigger during business hours. It's a common mistake and it makes the customer receive an automatic message when people are working.

Block 4: the intersection of the three problems

The three previous blocks can be audited separately, but the real damage appears when they combine. A conversation with no assignment, no tag, and no out-of-hours response is an inquiry that came in on a Sunday, nobody took it, nobody classified it, and nobody answered it. That's the case you need to report and bring to the meeting.

• Filter the conversations that meet all three conditions at once.
• Count how many there are. If it's more than zero, that number is the main indicator of the audit.
• For each one, note the category or the reason for the inquiry, even if it has no tag: it can be inferred by reading the first message.
• Check whether the customer wrote again. If they did, the failure already had a visible cost.
• Note whether that conversation ended in a sale, a referral, or nothing. If you can't tell, note it as 'no data': it's useful information for the next audit.
• Bring the list to the team meeting without names. The audit looks for process failures, not people to blame.

How to measure whether the audit worked

An audit without a second measurement is useless. Two or three weeks later, repeat the same three counts over a comparable week and compare. The numbers do not need to drop to zero: they need to move in the right direction and the team needs to know why.

• Percentage of unassigned conversations out of total open conversations.
• Percentage of untagged conversations out of the total for the range.
• Number of conversations that meet all three conditions at once.
• Time to first human response on messages outside business hours.
• Number of active tags: if it grew a lot, the system is getting messy again.

What to do when something fails

Each failure has a different cause and therefore a different solution. The most common mistake is treating everything as a problem of 'lack of attention' and responding with a motivational meeting. That does not change any number.

• If the problem is unassigned conversations: the cause is almost always that assignment depends on a single person or that nobody has the task explicitly. Define an owner per shift and an assignment rule when opening the conversation, not when closing it.
• If the problem is untagged conversations: the cause is that there are too many tags or that they are only applied at the end. Reduce the list to the ones you actually use and apply them on first contact.
• If the problem is the reply outside business hours: review the auto-reply first. If it does not say when they will reply, the customer does not know whether their message exists. Then check whether those conversations stay assigned to the first shift of the next day.
• If the problem is that the team does not see the conversations: review the permissions and the inbox filters. A conversation that exists but that nobody sees is the same as one that does not exist.
• If the problem is that you cannot measure anything: the first fix is not about process but about tooling. Without being able to filter by agent, by tag and by hour, the audit cannot be repeated.
• If the problem is that the numbers improve but the customer keeps complaining: then what fails is not the assignment or the tag, it is the quality of the response. That is another audit.

How this relates to the tool

This audit can be done with any WhatsApp Business inbox, including the app alone. If the team uses a tool with a shared inbox, labels, assignment and time-based auto-replies, the three blocks are audited faster because the filters already exist. Wando, for example, has an inbox with AI suggestions, labels, multi-agent and time-based auto-replies, and the AI learns from the business's real replies instead of requiring programmed flows. But the audit does not depend on the tool: it depends on someone sitting down once a week to count stalled conversations.