Inbox triage is one of the automations I build for clients: incoming mail gets read, classified and filed, and the messages that need a human rise to the top. It seemed only fair to find out what that looks like on my own inbox before telling anyone else it would help theirs.
So I sorted my business inbox properly, once, and wrote down what I found. The numbers are small. The finding is not.
What was there
About fifty threads in the inbox. Two labels already existed, one for DMARC reports and one for GDPR correspondence, set up by hand at some point and then, as it turned out, only half kept up.
Everything else had no classification at all. It was just the inbox, in date order, which is to say in no useful order.
The labels
I read every thread and sorted it into seven categories. These are the actual counts:
| Label | Threads | What it catches |
|---|---|---|
| Marketing & Ads | 18 | Google Ads and Business Profile notifications, performance reports |
| Account & Security | 14 | Login codes, two-step verification notices, sign-in alerts |
| DMARC Reports | 31 | Daily aggregate reports from Microsoft, Yahoo and others |
| GDPR | 28 | My own data-subject request threads with third parties |
| Billing | 5 | Workspace invoices and subscription notices |
| Admin & Registration | 5 | One-off business registration mail |
| Leads & Prospects | 1 | A real reply from a real prospect |
The DMARC and GDPR counts include threads that had already been filed out of the inbox, which is why the table adds up to more than fifty.
The finding: one in fifty
Of roughly fifty threads in the inbox, one was a reply from a prospect. The rest was infrastructure noise and account admin: ad platforms telling me things, security systems confirming things, invoices for tools I already knew I paid for.
That ratio is the honest result of this exercise, and I would rather use it than a made-up efficiency percentage. The value of triage is not that it saves some number of minutes. It is that the one message that matters is not buried under forty-nine that do not. On a busy day, in a flat inbox, that one reply is exactly the one that gets missed.
The other finding: hand-kept systems drift
The DMARC label was the interesting part. Twenty-seven older reports had been filed correctly. Four recent ones, from the last two or three days, were sitting in the inbox unlabelled.
Nobody decided to stop filing them. The filing just stopped, the way it always does when a system depends on someone remembering to do it. It works until the week you are busy, and then it silently drifts, and nothing tells you it has drifted because the whole point was that you would not have to look.
That is the real argument for automating this rather than asking someone to be more disciplined about it. Discipline is not the scarce resource. Attention is. A rule that runs on every incoming message does not have busy weeks.
What this was, and what it was not
To be straight about it: what I did was a one-time classification pass using Gmail’s labels. It was not a running automation, and so on its own it will drift exactly like the DMARC label did, just more slowly.
There are two ways to make it last. Gmail’s own filters handle the easy cases, the senders that are always noise. Anything that needs reading rather than matching, like telling a prospect’s reply from a supplier’s, needs a workflow that polls the inbox on a schedule and classifies each new message. The second is what I build for clients, and it is next on my own list.
If you want to try this on your own inbox
You do not need any software to learn the useful thing here. Take your inbox as it is, and for one afternoon, sort every thread into a handful of labels. Count them. Then look at the smallest pile.
For most businesses the smallest pile is the one with the customers in it, and seeing it at the bottom of that table is usually the moment the case for automating the rest makes itself.
If your smallest pile is getting lost under the others, that is a conversation I am happy to have.