Local shops that deliver: warn customers without calling

Half past seven in the evening. The shop is shut, the van is half loaded, and you're on the phone telling twenty-two people exactly the same thing: "I'll be round tomorrow between ten and twelve." Four don't pick up. Two want a chat. One asks for the afternoon instead, which means you're redoing the route at nine at night on the back of an envelope.

This one is for shops that do their own deliveries — greengrocers, health-food shops, corner shops, bakeries, florists, pet-feed suppliers. One van, a route that repeats most weeks, and a customer list where you know everyone by name. By the end you'll know which three messages are worth automating, what you need in place first, and when you're better off leaving things exactly as they are.

The phone calls aren't the expensive part. The unanswered doorbell is.

The calls are tiring, but they're forty minutes. The costly event is the delivery that fails.

When you arrive and nobody's in, three things happen at once. You lose the slot in your route, with the van sitting there on hazards. You carry the goods back, and if it's fresh, some of it gets binned. Then you have to renegotiate a second attempt by phone, and that attempt doesn't fit neatly into tomorrow's route either. A failed delivery doesn't cost you one delivery. It costs you two, plus the stock.

Most of those misses aren't the customer's fault. They're an information problem. She knew it was coming "this week"; she didn't know it was Thursday at eleven. That information existed — it was in your head — and it never reached her phone in time.

The three messages that do most of the work

You don't need a logistics platform. Three well-placed messages cover the bulk of it.

1. The night before: confirm the window. Goes out around seven in the evening with the order, the day, the time window and one closed question: does that work for you? The customer replies "yes" or "afternoon's better." That's where the real gain is, because you re-sequence the route with a whole evening in hand, not at eight in the morning with the van already loaded.

2. In the morning: the actual running order. When the van pulls out, each customer gets an approximate time based on where they sit in the route. Not "between 10 and 2", which nobody can plan around, but "you're fourth today, should be with you around 11:15."

3. Ten minutes out: I'm nearly there. This is the one that genuinely kills the unanswered doorbell. People are in — they're in the shower, hanging washing, halfway through unpacking something. One short message and they come down.

An agent connected to WhatsApp can send all three off the back of your delivery list. And all three go past you before they go out the first time: you approve the wording, you approve the route, and after that you only look at the replies that go off-script.

What you need in place before automating anything

This is where half of these projects fall over, and almost never for technical reasons.

• A customer list with usable mobile numbers. Country code included, no entries like "Mari (flat 3B)". If your numbers live in a notebook, step one is moving the notebook into a spreadsheet. An afternoon's work.
• A running order, even a rough one. You don't need route optimisation or anything close to it. You need somebody to write down the order deliveries actually happen in, which is usually already in the driver's head.
• Permission to message on WhatsApp. If a customer gave you their number for deliveries, messaging them about a delivery is within reason; sending them offers is not. Have a clear line about it when they sign up, and honour "stop messaging me" the first time it's said. That's baseline GDPR, and it's also just how you behave on your own street.
• Three statuses, and no more. Pending, delivered, nobody in. That's enough to sit down on Monday and see what happened last week. A fourth status nobody fills in is worse than having none at all.

With those four things, wiring up the messages takes days. Without them, automating first just automates the mess.

The actual messages

Short, name first, no flourishes, and nothing that reads like a parcel carrier. A local shop competes on being the opposite of that.

Hi Carmen, it's Javi from the greengrocer's. I'm bringing your order tomorrow, Thursday, between 10:00 and 12:00. Does that work, or shall we make it the afternoon?

The following morning:

Morning Carmen. You're fourth on the route today, I reckon around 11:15. If something comes up, tell me and we'll move it.

And the arrival one is shorter still:

I'm on your street, five minutes away.

Two details worth copying. First: the message is signed by a person and a shop, not by "Customer Service". Second: there's always an easy way out for the customer — that "tell me and we'll move it". A message that doesn't invite a reply isn't a heads-up, it's a notification, and people tune notifications out.

When to leave this alone

Honestly, there are cases where it isn't worth the effort.

If you deliver to eight regulars you've known for years, a two-minute phone call does the same job and costs nothing to set up. This starts paying for itself somewhere around fifteen to twenty drops a day, or at the point where someone other than you is driving and you need to coordinate with them every morning.

Don't build it as described if your route is rewritten daily by orders that land up until eleven in the morning. The night-before message would be lying, and a message with the wrong time does more damage than no message at all. In that case start with the third one only — "five minutes away" — because that one can never be wrong.

And don't build it if the goal is to stop talking to your customers. A shop with its own delivery run survives on Carmen feeling looked after. The point is to remove twenty-two identical evening phone calls, not to remove Carmen.

One week is enough of a trial

Pick a single delivery day. Take that day's route and send only the first message, the night-before one. Write down two numbers: how many replied, and how many deliveries failed. Compare it with a normal week. If it improves, add the second message the week after, and the third once the second runs itself.

At Yaqbot this is the operations agent's job: it takes your delivery list, drafts the messages, you approve them and it sends them, and anything off-script comes straight to you. You approve, it executes.