← All posts

Following up with suppliers on late deliveries (before the customer notices)

Automation · 21 November 2026 · 9 min read

A cafe's supplier follow-up only happens after something has already gone wrong. A six-step routine: one list of open orders with promised dates, a written definition of late, one saved message, a scheduled check that runs in shadow mode for a fortnight, then ten minutes a day with the log.

The scene

A small cafe in Sydney's inner west, run by two people who already work six days a week. Coffee, milk, bread, pastry and a handful of specialty items. Most orders arrive fine. The problem is the ones that do not, and what happens on those mornings.

On a Friday last month the milk delivery did not turn up. Nobody noticed until ten past seven, when the first customer asked for a flat white. The owner rang the supplier, got a voicemail, rang again at eight, was told a replacement would come "this afternoon", and drove to a service station to buy four litres at nearly double the price. A pastry order also arrived with the wrong item, because it had been placed on a supplier's web form in a hurry and never read back.

The owner is not disorganised. She is busy. Supplier follow-up is the classic category of work that only happens when something has already gone wrong, and by then the cost is already paid. This is about turning it into a routine a machine runs, so the owner hears about an order when it is late, not when a customer is standing in front of her.

None of what follows needs new software. It needs order details in one place and something that notices when a date passes. Supplier follow-up suits automation because it is predictable: an order has a promised date, and dates either arrive or they do not. It happens often enough to matter, six or eight orders a week for a cafe, and the cost of missing it is high and one-sided. A late delivery for a retail shop is annoying. A late delivery for a perishable business is lost trade and an expensive emergency substitute.

Step one: one list of open orders

The owner's real problem is that order information lives in five places: texts to the milk supplier, a phone call to the bread supplier, an email confirmation from the coffee roaster, a paper docket in the drawer and a note in her head. You cannot flag a late delivery if the promise is not written anywhere.

Build the smallest list you can that holds every open order. A spreadsheet is fine. A table in whatever tool your accountant or bookkeeper already uses is fine. It needs these columns and no others:

  • Supplier name
  • What was ordered, in plain words
  • Order date
  • Promised delivery date
  • Promised delivery time, if the supplier gave one (leave blank if they did not)
  • The reference or confirmation number, if there is one
  • Contact details the same day someone needs to ring
  • Status, which is one of waiting, arrived, partly arrived, late or cancelled

That is it. No columns for price or invoice matching yet; this list answers one question: what has been promised, and when should it be here?

The rule that makes it work: nothing gets ordered without a line going in the list. Not a text, not a phone call. If the supplier does not give a date, the order does not go ahead until they do. That sounds pedantic, and it will feel awkward for a fortnight, but it changes the supplier relationship in a good way. Suppliers who know you write down promises tend to give you fewer vague ones.

Step two: decide when "late" begins

Late is not a feeling. Pick a number and write it down.

For most perishable businesses, an order is late if it has not arrived by one hour before you need it. A cafe needs milk by six in the morning, so anything not delivered by five is late. For a workshop waiting on parts, late might be the end of the promised day. The rule that matters is that "late" is defined relative to a moment when someone can still do something about it, not relative to the end of the day when it is too late to call.

Work it out in the automation rather than the list, so the rule can change for all suppliers at once.

Then set two check-in points, not one:

  • A soft one, a day before, that does not send anything to the supplier but reminds a person internally.
  • A hard one, at the late moment, that triggers an actual follow-up.

Most small businesses go straight to the hard one, ring the supplier in a bad mood, and get a defensive answer. The soft check is what stops the problem happening: if the order due at five tomorrow was placed on Monday, Tuesday evening is the moment to nudge a supplier without a complaint, because nobody is angry yet.

Step three: write the follow-up message once

The message to a late supplier is short and the same every time. Write it once, save it as a template, and stop composing it in the moment. Something close to this:

Hi Sam, just checking on the milk order from Tuesday, reference 4471. We were told it would be here by five this morning and it has not arrived. Can you tell me if it is on its way and roughly what time? If it is not coming today, let me know now so I can arrange cover. Thanks, Alice.

It states the facts including the reference, asks about timing and the fallback, and does not accuse or use the word "again" even if it is the third time this quarter. The "again" goes into a separate record, covered below.

The follow-up should be sent at the moment the hard check fires, and not by the owner if she is in the middle of a rush. The whole point is that the machine sends it, and a human is only involved if there is no answer in a reasonable time.

Forty-five minutes is a reasonable wait during trading hours. If there is no reply by then, the person who can actually solve the problem gets a call or a card in the till. That escalation stays manual, and should.

Step four: build the smallest thing that runs

With a list, a time rule and a template, the automation is a scheduled check, not a project. Something reads the list once or twice a day and does two things: marks anything past its late moment as late, and sends the template to the supplier. Use what the business already pays for, such as a Google Sheet with a small script, or the automation feature inside the booking or bookkeeping tool. Do not add a tool that exists only for this.

Two practical details. Email is often the wrong channel for a milk supplier; many small suppliers prefer a text, and some only read their order system. Where the follow-up has to be a phone call, the automation's job is to put the call in front of a person at the right moment with the right facts, not to make the call. And every message should be logged, one line per event: date, order, supplier, what was sent, what came back. That log later becomes the evidence for a conversation about whether to keep the supplier.

Step five: run it in shadow mode for a fortnight

Before anything sends automatically, run it for two weeks and have it record what it would have done instead. Check that record evening by evening for three things: false alarms (early or cancelled deliveries still marked waiting), missed real problems (a missing row or blank column is nearly always the cause), and messages that read badly to a supplier. Fix the list and the template, not the code.

Shadow mode is dull, and it is the only thing standing between an automated message that saves a morning and one that damages a ten-year supplier relationship.

The failure modes to expect

These are the ones that actually happen, in rough order of frequency.

The list dies.

The owner keeps ordering by text because it is quicker in the moment, and within a month half the orders are not in the list. Fix: make adding a row the first half of the ordering action, not an afterthought. If the supplier has a web form, put the reference straight into the list as part of the order.

The supplier relationship sours.

Automated messages can read as cold, especially when the same supplier gets three in a fortnight. Two guards help. Keep a real conversation about it once a quarter, in person or on the phone, saying plainly that you are tracking deliveries and it is not personal. And add a rule that any supplier with two late deliveries in a month gets a phone call from the owner instead of another automated message. The machine should trigger that call, not replace it.

Messages go out when the supplier is closed.

A check that runs at five in the morning and sends a text is fine. A check that runs at eleven at night and sends an email that sits unread until noon is not. Set the sending window to the supplier's trading hours.

Everyone stops trusting the numbers.

Somebody marks an order as arrived because they saw a delivery van, then the actual item is missing. The list says arrived, the machine is satisfied, the problem reappears. Fix: "partly arrived" is a real status and should be used. Anything partly arrived gets a person's glance the same day.

The automation is invisible when it breaks.

The scheduled check stops running because a password expired or a computer was switched off. Nobody notices for a fortnight because the business is running normally without it. Fix: a weekly count of events in the log. If the number is zero, something is wrong.

What good looks like after a month

The list is the only place orders exist. Every open order has a date. Late is a number, not a mood. The owner rarely composes a supplier message, because the template exists. When a delivery is late, the supplier hears about it before a customer does. The escalation to a phone call happens maybe twice a month, and it is the owner's own decision, made with facts rather than in the middle of a rush.

The measurable change is not a saved cost, it is where the owner's attention goes. She is not scanning a drawer of dockets at six in the morning hoping the milk is there. She reads a short log at the end of the day and moves on. After a while she starts noticing patterns in it, such as which supplier is late most often, and that is a business decision that was previously impossible to make because the information did not exist.

Common questions

How do I automate chasing late supplier deliveries?

Put every open order in one list with the promised delivery date, decide what counts as late, then have a scheduled check flag anything past that time and send a saved message. Run it for a fortnight before letting it send anything so you can fix the false alarms first.

Should a business email or text a supplier when a delivery is late?

Use whichever channel the supplier actually reads. Many small suppliers respond faster to a text or a phone call than to email, so the automation can prepare the message and the facts while a person makes the call.

How do I keep automated follow-ups from damaging a supplier relationship?

Keep the message factual and polite, talk to the supplier in person about it once a quarter, and have any supplier with repeated late deliveries dealt with by a phone call from the owner rather than another automated message.

Want this done for you?

I build one small automation at a time for North Shore businesses, usually starting with whatever is costing you the most hours.

Start a project ↗