Research
How do you give a business numbers when its real sales channel is an informal one — WhatsApp, a phone call, word of mouth — without taking away the channel that actually works for it?
The problem
Mexican micro-enterprises are usually described as un-digitalised. The description is wrong in an interesting way: they are digital, through messaging. One in three small businesses says it built itself on WhatsApp, and more than 65 % of Mexican online shoppers have bought through a chat. What that channel doesn’t produce is structured data — so the business never learns when its demand arrives, what is asked for most, or how many people looked at its menu and didn’t order.
I ran into this from the inside. Mercadito, a platform I built, started as a delivery marketplace with a 5–8 % commission. On 24 August 2026 it stopped operating deliveries and was redesigned around the channel that had beaten it: orders now leave for the business’s own WhatsApp. The commission code is still in the repository, switched off behind a flag — a model adopted, operated and abandoned, dated and auditable. The same pattern shows up in public reviews of the local competitor, where customers describe couriers phoning the restaurant to place the order by hand.
Four data sources
- Event-level telemetry of the WhatsApp menu channel, one row per menu open and per order sent, prospective since 28 September 2026.
- A prospective delivery logbook, from October 2026: an instrumented field diary a courier fills in himself at the end of each day.
- Timestamped table and point-of-sale records, already running — the only orders the platform sees end to end.
- A public corpus of 1,167 reviews of five delivery apps: the local competitor plus Uber Eats and DiDi Food as a national baseline.
Aggregation by design
Personal data is not anonymised after the fact; it is not captured. Amount bands instead of tickets, neighbourhood instead of street address, time-of-day band instead of the customer’s clock, a random session id that expires in 24 hours and is never linked to an account. Participating businesses are told what is measured and can opt out by email without losing anything — the notice and the exclusion route are public at mercadito.cx/piloto.
Why Japan
Not because Japan solved this. It didn’t: it has its own shuttered shopping streets and its SMEs lag on digital transformation. Japan is the world’s most instructive laboratory precisely because it faces the same structural problem — ageing owners, informal channels, succession gaps — while having produced the most systematic policy response anywhere: shōtengai revitalisation programmes, subsidies for adopting digital tools, and a national SME productivity programme. Both its successes and its failures are documented, and both are what I want to study and adapt.
How the instrument is watched
The platform is both the object and the instrument, so its failures matter to the research. A dated chronology of the corrections — including the three occasions when the series turned out to be blind to an entire channel — is published separately: what we kept finding was missing.
Adrián Ceja Rentería · Sahuayo, Michoacán, Mexico. Applicant to the JICA Knowledge Co-Creation Program (Information Science and Engineering, Ritsumeikan University). Written up in Spanish as a technical note.