We went looking for something else entirely. We were trying to work out how long a shopper waits before giving up on a reply, and the answer turned out to be uninteresting. What we found on the way there was not: an enormous share of the questions people ask online stores are asked into an empty room.
The night shift is real, and it is enormous
Here is every message we have, arranged by the hour it arrived.
View as table
| Hour | Messages | Share |
|---|---|---|
| 00:00 | 358 | 1.7% |
| 01:00 | 258 | 1.3% |
| 02:00 | 93 | 0.5% |
| 03:00 | 81 | 0.4% |
| 04:00 | 81 | 0.4% |
| 05:00 | 74 | 0.4% |
| 06:00 | 150 | 0.7% |
| 07:00 | 453 | 2.2% |
| 08:00 | 797 | 3.9% |
| 09:00 | 1,212 | 5.8% |
| 10:00 | 1,456 | 7.0% |
| 11:00 | 1,493 | 7.2% |
| 12:00 | 1,541 | 7.4% |
| 13:00 | 1,404 | 6.8% |
| 14:00 | 1,366 | 6.6% |
| 15:00 | 1,344 | 6.5% |
| 16:00 | 1,373 | 6.6% |
| 17:00 | 1,144 | 5.5% |
| 18:00 | 1,189 | 5.7% |
| 19:00 | 1,118 | 5.4% |
| 20:00 | 1,040 | 5.0% |
| 21:00 | 1,084 | 5.2% |
| 22:00 | 960 | 4.6% |
| 23:00 | 646 | 3.1% |
The daytime shape is the one you'd expect. Volume climbs through the morning, peaks at 12:00 (7.4% of all messages), and holds through the afternoon. If you staffed a support inbox from nine to six, you'd feel like you had it covered.
You wouldn't. Look at what happens after the shops shut. Volume doesn't fall off a cliff at six — it steps down to a plateau and then stays there for five more hours. 29.1% of all messages arrive after 18:00. A further 12.3% land between ten at night and six in the morning, in what you'd be forgiven for assuming was dead air.
The detail that stopped us: more people message a store at midnight than at six in the morning — 1.7% versus 0.7%, close to two and a half times as many. Midnight isn't the tail of the distribution. Six a.m. is.
This isn't an artefact of where we drew the line
The obvious objection to a statistic like 40.5% is that we picked a flattering definition of "closing time." So here it is under every definition we could think of, including some very generous ones.
| If the working day is… | Messages outside it | Share |
|---|---|---|
| 09:00-17:00 | 9,526 | 46.0% |
| 09:00-18:00← headline | 8,382 | 40.5% |
| 10:00-19:00 | 8,405 | 40.6% |
| 09:00-20:00 | 6,075 | 29.3% |
| 08:00-20:00 | 5,278 | 25.5% |
Even if you define the working day as 08:00–20:00, seven days a week — twelve hours, no weekends off, a schedule almost nobody actually works — 25.5% of messages still arrive outside it. Apply the local weekend and the figure goes the other way, to 48.9%.
There is no cutoff that makes this problem small.
Most shoppers ask exactly once
The second finding reframes the first, and it's the one we keep coming back to.
View as table
| Messages | Conversations | Share |
|---|---|---|
| 1 | 2,751 | 37.8% |
| 2 | 1,371 | 18.8% |
| 3 | 1,081 | 14.9% |
| 4-5 | 1,316 | 18.1% |
| 6-10 | 695 | 9.6% |
| 11+ | 63 | 0.9% |
37.8% of conversations are a single message. One question, no follow-up, gone. Add the two-message conversations and 56.6% of all shopper conversations are over in two exchanges or fewer. By three, you've accounted for 71.5%.
This is not what a support conversation looks like. It's what a checkout hesitation looks like — one unresolved question standing between someone and the purchase. Does this ship to me. Is this the right size. Can I return it if it's wrong.
Which means the reply window is far narrower than the volume alone suggests. These people are not filing a ticket and waiting for an answer; they are resolving a doubt or abandoning the cart. A response that arrives the next morning isn't a slow answer to their question. It's an answer to a question that no longer exists, sent to someone who already bought elsewhere.
Put the two findings together and the shape of the problem is clear: the largest single block of these micro-moments happens when there is nobody there, and by the time anybody reads them, the moment has closed.
About one message in eleven is about your own policies
9.0% of messages — 1,862 of them — required looking something up in the store's own stated policies: shipping times, delivery costs, returns, exchanges, warranty terms.
Every one of those answers was already published somewhere on those sites. People asked anyway. Not because the information was missing, but because finding it meant breaking off what they were doing to go hunting for a policy document written for a lawyer — all while holding a question that takes eight words to ask.
What we'd take from this if we ran a store
Nothing here requires buying anything, so here is the honest version.
Look at your own after-hours share before you do anything else. Your analytics already know when your contact form is submitted, when your live chat gets missed, when your inbox fills up. If your split looks anything like ours, the staffing question answers itself, and it isn't "hire someone for evenings."
Judge response time in minutes, not hours. A first-reply SLA measured in business hours is measuring the wrong thing when 56.6% of conversations end within two messages. The comparison that matters isn't your competitor's response time — it's how long the shopper is willing to stand still.
Put the answers where the question gets asked. Shipping cost, delivery window, and return terms belong on the product page, in a sentence, near the price. The 9.0% of messages that go looking for them are the ones you can remove entirely without answering a single one.
Read your own unanswered messages. Not the tickets — those are the ones that survived long enough to become tickets. The single-message conversations that went nowhere are a list, in your customers' own words, of the reasons people didn't buy. It is the cheapest research available to you and almost nobody reads it.
None of that touches the after-hours window, which is the one part you can't fix from your own side of the counter. 40.5% of these questions arrive when no one's staffed to answer them. On the stores running our agent, 71% of them get answered anyway — and 12% end in a sale.
If there's a cut of this you'd find useful — a different definition of the working day, a breakdown we didn't think to run — say so and we'll run it.
oomates builds AI agents that work inside WooCommerce stores, which is why we had these logs to look at in the first place.