← All notes

November 2, 2026 · In Mercadito

A month of WhatsApp errands: the October field log

An independent errand runner in Sahuayo sent the summary of his day over WhatsApp every night. In October that came to 268 trips and 367 drop-offs, and no platform saw a single one. Here's what they showed — and what they can't say yet.

It's one of the sources of Mercadito's pilot: a field diary. Mercadito doesn't deliver — it stopped in August 2026; the collaborator works on his own, with his own customers, over WhatsApp, and at the end of the day he sent what he'd done. From each message, this study uses only the city, the kind of errand and the kind of place. No names, no streets, no neighbourhoods, no which-medicine, no money.

It's one person (n = 1) over one month. What follows are leads for framing questions and deciding whom to interview, not conclusions about delivery work in Sahuayo. And no money is analysed: the pilot doesn't study anyone's fees or income, his included.

days worked

26 of 31

trips

268

drop-offs

367

trips per day worked

10.3

to another town

36 %

for someone who'd ordered before

92 %

The first week was the busiest, at 13 trips a day; after that it settled between 8 and 10. Each trip is one journey; there are more drop-offs because a shop's delivery routes knock on several doors in one go.

It isn't restaurant delivery: it's running errands

Almost six in ten trips are purchases: the customer texts what they want, he buys it — at the supermarket, the pharmacy, the hardware store or the taco stand — and brings it over. One in six combines two or more stops in the same journey. Only a third are deliveries for a business that already had the order ready. The 'restaurant app with a courier' model describes almost none of it.

Food is half of it, but a long tail follows: groceries, parcels, pharmacy, hardware, a phone to fix, stationery, a laundry, flowers. Trips started at 44 different kinds of place, and none accounts for more than one in six.

Almost all of it comes from repeat customers

92 % of trips were for a customer, a business, a neighbourhood or an errand that had come up before: a chicken shop that gave him 44 deliveries over 22 days; a clothing shop that booked 19 delivery routes with over a hundred stops; a parcel that goes from a hospital to a courier office, 13 times in the month; a lunch he carries from a house to an office, another 13. A single neighbourhood in Jiquilpan got 87 trips, on 25 of the 26 days he worked.

That's his strength and his risk: the work hangs on a few customers. It's also a bias in the study: much of what you see here is what those few ask for.

Jiquilpan weighs more than it looks

Jiquilpan, the neighbouring town, got 36 % of the trips, nearly nine in ten of them to a single neighbourhood. Two in three trips to Jiquilpan start in Sahuayo: bought here, delivered there. It's trade between two towns that runs entirely over WhatsApp.

Of the 153 trips that ended in Sahuayo, 111 name the neighbourhood, and half of those went to a single area. Jiquilpan shows the same, more sharply: that's why it counts as one area, since splitting it would point to specific homes. Less is known about origins: only four trips in ten say which neighbourhood the purchase came from, and half of those came from Sahuayo's main boulevard or centre, where the shops are.

When

What to fix in the log

• The time. Only six days came with times, and without the time slot (morning, lunch, afternoon, night) we can't say when the work comes in — the first thing a business wants to know.
• The destination. Fifteen trips don't say which town they went to; a required field keeps them from getting lost.
• Bands, not figures. The log was designed to ask for an amount band, not the amount; over WhatsApp, exact figures came in. The form only captures the band, which is what the study declares.
• Free text brings too much: business names, neighbourhoods, medicines. All of it was stripped on entry, but better not to receive it at all: the log form asks only for what the pilot may keep.
• The gaps. Three days without a report: a reminder at night helps more than a rule.
• One person. With n = 1, every finding may be about him rather than the trade. The log needs more couriers.

What we'd tell someone who runs errands

• Bundle the trips to the other town: post on a WhatsApp broadcast list 'heading to Jiquilpan at such-and-such time' and carry several errands in one journey.
• Build on the trips that already combine stops: one in six goes through two or more shops; posting 'stopping at the supermarket and the pharmacy' draws more errands along the way.
• Agree on fixed times with the biggest customers (the chicken shop, the clothing shop): the work becomes predictable for both sides.
• Use the slow days — Tuesday, Friday and Sunday — to find new customers, for instance on Mercadito's free couriers board.
• Mind the pace: the first week ran at 13 trips a day, and in the second he fell ill.

What Mercadito can do for couriers and shops

Mercadito hasn't run deliveries since August 2026 and won't take a commission: that model was tried and dropped, and its code stays switched off as evidence. What it can offer is access — the documented complaint among couriers is a closed door: 'they already have enough couriers' — and numbers:

• The couriers board: each person posts their neighbourhoods, fee and hours, and the deal is made directly over WhatsApp. It's missing a 'fixed route' ('to Jiquilpan, Tuesdays and Thursdays') for demand between towns.
• The log as their own tool: at month's end, each courier gets their own numbers — trips, repeat customers, good days and slow ones — like the ones in this note. With their consent, the same log feeds the study.
• For shops, on their dashboard: the board's couriers who cover their area, with a WhatsApp button. Mercadito doesn't assign or charge: it only introduces them.
• Fixed routes, like the clothing shop's: a shop with recurring deliveries can agree a day and a route with a courier from the board.
• And say what's true: today the board gives visibility, not work. Promising income with today's demand would be lying to whoever signs up.

Nothing gets deleted: the October log stays in the pilot's append-only database, and the trip-by-trip dataset is kept for next year's analysis.

All notes

• Mercadito goes free: a research pilot
• The client never sends a number