What to do with WhatsApp chats that arrive overnight

An overnight WhatsApp message carries a deadline: the 24-hour customer service window it opens closes 24 hours later, and after that you can only reply with an approved template. Most teams do not realise the clock exists, and discover it when a Monday reply fails.

Meta's sending messages documentation sets out the mechanic: the window opens when a customer messages or calls you, and resets each time they do. Nothing you send extends it.

The Friday-evening problem

Work through the timings and the trap is obvious:

Customer messages	Window closes	Team back at desk	Can you reply freely?
Tuesday 11pm	Wednesday 11pm	Wednesday 9am	Yes
Friday 6pm	Saturday 6pm	Monday 9am	No
Saturday 10am	Sunday 10am	Monday 9am	No
Sunday 8pm	Monday 8pm	Monday 9am	Yes

For a Monday-to-Friday team, everything arriving between Friday evening and Sunday morning is unanswerable by free text on Monday. The attempt fails with error 131047, and most stores have no template written for "sorry for the delay".

Four ways to handle it

1. Clear Friday before you leave. The cheapest fix. A reply that prompts the customer to respond keeps the thread alive into the weekend.
2. Check once on Saturday morning. Ten minutes clears most of what arrived overnight and resets every window it touches.
3. Keep one re-opening template approved. A plain utility template — "we're back and picking up your message" — charged once, and free again the moment they reply. Cheap insurance for a team that genuinely does not work weekends.
4. Automate the overnight self-service so a share of those conversations resolve without anyone.

Most teams need the first and the third. The second is a nice-to-have and the fourth is the one with the best return.

Make the overnight queue smaller

The conversations arriving at 11pm are disproportionately answerable without a person: order status, delivery timing, returns policy. Three flows cover most of them, and all three run at 2am exactly as they do at 2pm.

• An after-hours reply that says when you reopen — not just that you are closed. See after-hours auto-replies.
• Order-status answering, which handles the single most common overnight question — the lookup flow.
• A knowledge base for policy questions — AI answers from your own website.

All of these are free to run, because the customer messaged first and everything inside an open window is free under Meta's pricing rules — service conversations have been free and unlimited since 1 November 2024.

Use the pending status properly

Set overnight conversations to pending in the out-of-hours flow, so your team opens a defined queue in the morning rather than scrolling a general list. Then work it oldest first — the oldest conversations are the ones closest to their window closing, which inverts the usual instinct to answer the newest.

That single ordering change is worth more than it sounds. Answering newest-first means the messages you eventually reach are exactly the ones you can no longer reply to freely.

A morning routine that takes ten minutes

1. Open the pending queue, sorted oldest first.
2. Reply to anything from more than 12 hours ago before anything else.
3. Assign whatever needs a specialist rather than answering it badly.
4. Note anything answered more than twice — that is a candidate for a flow or a knowledge-base entry.
5. Clear the queue before starting on new arrivals.

Step four is where the overnight queue pays for itself. The questions people ask at midnight are the same ones they ask at midday, and they are unusually easy to spot in a small batch.

What about hiring for it?

Usually unnecessary, and worth checking before you do. Count the overnight conversations for a fortnight and split them into three: answerable by an existing flow, answerable by a flow you could build, and genuinely needing a person at that hour.

Most stores find the third bucket is small — and that the real problem was never coverage but the fact that nobody told the customer when to expect a reply. That is a copy fix, not a staffing one.

The one thing to change today

If you do nothing else, write an after-hours reply that names a real reopening time, and set those conversations to pending. It takes ten minutes, costs nothing to run, and removes both the "is anyone there?" follow-up and the Monday-morning scramble to find what arrived. Team mechanics are in running one WhatsApp number across a team.

What actually arrives overnight

Before deciding how to staff it, it is worth knowing what the queue is made of. In most stores it splits roughly three ways, and only one part needs a person at that hour.

Type	Typical share	Needs a person?
Order status and delivery timing	Largest	No — a lookup answers it
Pre-purchase questions: sizing, stock, policy	Substantial	Mostly no — a knowledge base answers it
Complaints, damaged items, payment problems	Small	Yes

The implication is that overnight coverage is usually a content problem rather than a rota problem. Automating the first two shrinks the queue to something a morning routine can absorb, and the third is small enough that nobody needs to be awake for it.

Do not automate the third row

An angry customer at midnight does not want a menu. Route anything that looks like a complaint straight to the pending queue with an honest message about when someone will reply, rather than putting them through self-service first.

Two signals worth branching on: a message containing complaint keywords, and a message that is a photograph rather than text. A keyword condition cannot read an image, so use the message-type condition to detect attachments and send those to a person — someone photographing a damaged item at 11pm is exactly the conversation you do not want a bot handling.

Timezones make this harder than it looks

If you sell across borders, "overnight" is not one period. A customer in a different timezone is messaging during their working day, and telling them you are closed without saying where you are reads as unhelpful.

Naming the timezone fixes most of it: "we're back at 9am UK time" is actionable in a way that "we're closed" is not. If a meaningful share of your orders come from one distant market, consider saying plainly that replies there take a day — customers plan around accurate information and complain about optimistic information.

A weekly check on the queue

1. How many conversations arrived outside hours?
2. What share were answered by a flow rather than a person?
3. How many were still unanswered 24 hours later — that is, past their reply window?
4. What was asked more than twice?

The third number is the one to watch. Anything above zero means customers are being left in a state where you cannot reply freely, and the fix is either an earlier morning routine or a re-opening template kept ready.

What good looks like

A store handling this well has: an after-hours reply naming a real reopening time, order-status and policy questions answered automatically at any hour, complaints routed to a pending queue rather than a menu, a morning routine that works oldest-first, and one approved template held in reserve for conversations that went past their window.

None of that requires anyone to work nights, and the whole thing costs nothing to run — every message involved is a reply inside an open conversation window.

The cost of getting it wrong

Nothing about an unanswered overnight message shows up as an error, which is why it persists. The costs are all indirect:

• A customer who messages again. Two unanswered messages feel worse than one, and the second often carries more frustration than the first.
• A conversation you can no longer answer freely, which means either an approved template or an apology sent through some other channel.
• A block. Customers who feel ignored mute or block the number, and blocks are the strongest negative signal feeding your quality rating.
• A sale lost to whoever replied first. For pre-purchase questions this is the common outcome and the one nobody measures.

Start with the message, not the rota

If you take one thing from this page: most overnight dissatisfaction comes from silence rather than from delay. A customer told clearly that you reopen at 9am and answered at 9:15 is satisfied. The same customer left with no reply until 9:15 is not — identical response time, completely different experience.

That makes the after-hours reply the highest-return change available here, and it costs nothing to run. Everything else on this page is an optimisation on top of it.

Write it once, publish it, and the problem largely stops being one. Ten minutes of work removes both the midnight follow-up asking whether anyone is there and the Monday-morning scramble to find what arrived over the weekend.