Who watches the store out of hours

The store does not close

A physical shop has a shut time. An ecommerce store does not, and that creates a category of problem most small teams handle by accident: something happens at 9pm, or on Sunday, and whether it gets dealt with depends on whether somebody happened to look at their phone.

That arrangement usually works, which is the trap. It works until it does not, and the failure tends to be expensive because out-of-hours problems compound. A payment issue overnight is a morning of angry emails. A fraudulent order placed on Friday evening ships on Saturday.

Most teams respond either by not thinking about it at all, or by everyone half-watching everything all the time, which is exhausting and does not actually produce coverage because nobody is definitively responsible.

The only question that matters

Not "what might happen overnight" but "what genuinely changes if we act now rather than at 8am?"

Run everything through it honestly and the list gets short fast. Most things that feel urgent at 10pm are not improved by being handled at 10pm, they just feel better to have dealt with. That feeling is worth resisting, because the cost of a team that monitors everything constantly is real and it is paid every night.

Things that genuinely change

• A high-risk order that will ship before anyone looks again. The window closes at dispatch, so if your fulfilment runs early or a 3PL picks up overnight, this is real.
• The store being down or checkout failing. Every hour is lost revenue and there is no catching up.
• A live campaign spending money on something broken. An ad set pointing at a 404, or a promotion with a wrong discount code, keeps costing until stopped.
• A stock error causing oversells during a launch. Each hour adds orders you will have to cancel, and cancellations are more expensive than the sales were worth.

Things that feel urgent and are not

• An unhappy customer email. Replying at 11pm sets an expectation you cannot sustain, and a good reply at 8am is better than a fast one at 11pm.
• A normal order that needs a question answered.
• A supplier problem, when the supplier is closed anyway.
• A low stock alert, unless it is during a launch. Overnight is rarely the difference.

Most stores find only the four in the first list justify contact, and several of those can be designed away entirely.

Designing the problem away first

Before rostering anybody, check whether the event needs a human at all. This is usually cheaper than coverage and is permanent.

• Do not ship overnight. If nothing dispatches between the last check and the first, high-risk orders stop being an overnight problem and become a morning one.
• Pause campaigns you cannot watch. A promotion that starts at midnight with nobody awake is a decision to discover any problem at 8am having paid for it all night.
• Set a buffer on launch stock, so an error causes a stockout rather than an oversell. A stockout is a disappointment; an oversell is a refund, an apology and a review.
• Automate the response where one exists. If an out-of-stock item should be unpublished, have it unpublish rather than alerting somebody to do it.

Covering what is left

After that, most stores have a very small list. Cover it deliberately.

One named person per night, not everyone. Shared responsibility for an unlikely event produces either four people checking constantly or nobody checking, and usually both across a week. One person knows it is them, everybody else genuinely switches off, and the total attention spent drops sharply.

A separate channel for genuinely urgent things. The point is not the notification, it is that everything else can be silenced. If overnight alerts share a channel with routine traffic, the person on call is either reading everything or muting the channel that contains the thing that matters.

Written permission to act. The person on call needs to know what they can do alone: cancel an order, pause a campaign, issue a refund up to a value. Coverage without authority is a person who can only wake you up.

Written permission to not act. Equally important and usually missing. State explicitly what waits until morning, so nobody feels they should have done something about the angry email at midnight.

The weekend, which is different

Weekends are not just longer nights. Order volume often rises while supplier and carrier support is closed, so problems accumulate with no route to resolution, and the Monday queue is the real cost.

The practical adjustment is a short Saturday and Sunday check rather than continuous coverage. Fifteen minutes, at an agreed time, doing one specific thing: look for the four events above, deal with them, close the laptop. A defined check that ends is far more sustainable than a vague sense of being on duty for two days, and it produces better coverage because it actually happens.

The morning handover is the part that gets skipped

Out-of-hours cover is usually designed as a night-time problem and then fails in the morning, because whatever happened overnight has to reach the person who takes over and frequently does not.

The failure looks like this. Something happened at 11pm, the person on call dealt with it, and they are now asleep or off. At 8am somebody else starts, sees an order in an odd state or a customer following up on a conversation they were not part of, and has to reconstruct it. Sometimes they act on the same problem again.

The fix is small and has to be non-negotiable: whoever dealt with something overnight writes two lines before they stop. What happened, and what they did. Not a report, and not a message to an individual who may also be off.

Equally important is the negative case. If nothing happened, say that too. An empty overnight note is information, because silence is otherwise ambiguous: the morning person cannot tell the difference between a quiet night and a night where somebody forgot to write anything.

The first check of the morning

Whoever starts first should have a short fixed list rather than a general intention to catch up. Fixed lists get done and general intentions do not.

• The overnight note. First, before anything else, including email.
• Orders that arrived after the last check, specifically anything flagged.
• Anything that was mid-flight at close. A customer waiting on an answer, a shipment being traced.
• The store loads and checkout works. Thirty seconds, and worth it, because an overnight failure that nobody noticed is still failing.

Five minutes, same order every day. The value is not any single check but that the morning starts from a known state rather than from whatever is loudest in the inbox. The inbox is a queue of other people's priorities, and starting there means the first hour is spent on whatever somebody else thought was urgent.

Rotating without resentment

If more than one person can cover, rotate on a fixed pattern set well in advance rather than arranging it weekly. Ad hoc rotas produce the same conversation every week and quietly land on whoever is least willing to push back, which is how coverage becomes one person's permanent unpaid responsibility.

Two things make a rota sustainable in a small team. Being on call has to mean something specific and bounded, not a vague sense of availability: a defined list of events, and explicit permission to ignore everything else. And whatever compensation applies, whether time back, pay, or simply not being expected in early the next day, should be agreed once rather than negotiated per incident.

Related reading: the first ten minutes of a high-risk order covers the most common overnight decision, and escalation paths without a manager covers who to wake and when.

Common questions