Meit
2023
Hiring home services in Bogotá — designed for both sides of the marketplace
MEIT is a marketplace for hiring domestic and home-maintenance work in Colombia, built as one product with two deliberately different experiences: Aliados, the independent workers who sell their services, and Clientes, the households that hire them.
Problem
Word of mouth is not a gap in this market. It is the market.
In Bogotá, domestic and maintenance work is brokered almost entirely by referral. A household asks a neighbour, a colleague, the porter in the lobby. A cleaner or an electrician gets their next job from the last client who was happy with them.
It is tempting to read that as an absence — no platform, no listings, no reviews. But referral is not nothing. It carries the one thing this market actually runs on: someone you trust has already vouched for the person you are about to let into your home.
What it does not do is admit anyone new. Reputation travels easily between people who already know each other and not one step further. A worker with ten years of experience and no shared contact has no way in, however good she is — and a household’s options stop at the edge of who it happens to know to ask.
" The market is not short of skill. It is short of introductions.
Both sides absorb the cost of that. Providers described income that swings without warning — not because demand does not exist, but because it is invisible to them. Households described choosing between a stranger and doing it themselves, with no way to check who someone is, what they have done before, or what the work should cost.
The exposure is real and it is mutual. One side hands over the keys to their home. The other gives up a day’s work to someone who may not pay.
So the brief was not to build a directory. A directory strips out the only thing referral was providing and leaves both sides worse off than the neighbour they used to ask. MEIT had to manufacture, at scale and between strangers, the trust that a referral gives away for free — and do it before it could ask anyone to make a booking.
Research
How we listened
We ran 20 semi-structured interviews of 30 minutes before designing anything: 10 with people currently doing domestic work, and 10 with the households who hire them. Providers were recruited through my own network of people who employ domestic help, and the sample was varied deliberately across background, income, rates charged and type of service, so it would not describe a single kind of worker.
We asked people to walk us through the last time they found work, or found a worker, rather than what they would want from an app. The second question produces feature requests; the first produces behaviour.
Three findings, and the first was not the one we expected
01
Safety came first, from both sides
Providers described the fear of entering the home of someone they know nothing about. Households described the same fear pointed the other way. It is the same anxiety twice, and it is mutual — which is why verification in this product could never be a one-way check on the person doing the work. That imbalance is intentional. The Aliado's verification is not bureaucracy — it is the entire value proposition being sold to the client side. Friction there produces the trust the marketplace runs on.
02
Word of mouth gates the market, and gates newcomers hardest
Reputation travels between people who already know each other and not beyond them. An experienced worker with no shared contact has no way in.That imbalance is intentional. The Aliado's verification is not bureaucracy — it is the entire value proposition being sold to the client side. Friction there produces the trust the marketplace runs on. Friction on the client side produces nothing but abandonment, so we removed all of it we could.
03
Income fails on planning, not only on demand
Several providers described weeks with no income not because nobody needed the work, but because nothing had been scheduled ahead. Their week is organised, or it is empty.That imbalance is intentional. The Aliado's verification is not bureaucracy — it is the entire value proposition being sold to the client side. Friction there produces the trust the marketplace runs on. Friction on the client side produces nothing but abandonment, so we removed all of it we could.
What the competitive analysis added
We mapped who already served this market and where they were failing — Taskrabbit as the global reference, Jelpit as the incumbent backed by an insurer, Timbrit as the local challenger — and used each as a real user would, booking a service and reaching the payment step.
Coverage
Every platform was concentrated in one city or absent from Colombia entirely
Rigid packages
Competitors sold fixed bundles that fit nobody exactly
Unverifiable experience
None of them let a provider show proof of past work
Transaction fees
A fee appeared as a weakness for both Taskrabbit and Jelpit — raising the price for the client and lowering the take for the provider, in a market where both sides are price-sensitive.
Security
Listed as a threat for Jelpit: users must share personal information and grant access to their homes. Our interviews found the same fear from the other side of the door
Solution
One product, two deliberately different experiences
We built MEIT as two front doors onto the same marketplace. Aliados get a profile that does the selling for them: verified identity, itemised services, photographed past work, client references and their own working hours. Clientes get a search that narrows from category to the specific activities they need, filtered by verified providers near them, with the price agreed before anyone is booked.
The design work concentrated on trust and on payment, because those were the two points where the market was actually failing. Identity verification, work history and a two-axis rating system carry the trust. Card, bank transfer, in-app wallet and cash — with withdrawal to Nequi — carry the payment, so that being unbanked is not a reason to be excluded from either side.
The structural decision the whole product rests on
The two onboardings are deliberately asymmetric. An Aliado passes through roughly fourteen steps: identity document, a live selfie, personal data, service categories and activities, experience with photographs, client references and weekly availability. A Cliente reaches a usable home screen in about five.
That imbalance is intentional. The Aliado’s verification is not bureaucracy — it is the entire value proposition being sold to the client side. Friction there produces the trust the marketplace runs on.
Flows
App Architecture
With the research settled, we mapped both journeys end to end before drawing a single screen — two flows sharing a backend, a navigation shell and a component library, but almost no screens. Working at flow level let us remove redundant steps while they were still cheap. This is the stage the project rests on.
Design
Built as a system, not a set of screens
The interface had to read as competent and safe to two audiences with very different relationships to technology, and survive being used one-handed, outdoors, on a mid-range Android phone. One type scale, a restricted palette with a fixed semantic meaning for every colour, and a service iconography set that lets a category be recognised without reading.
Each service category carries its own illustrated icon — cleaning, beauty, plumbing, maintenance, installations — drawn in a single style so a category is recognisable at a glance, before it is read. In a product used across a wide range of literacy and familiarity with apps, that is an accessibility decision rather than a decorative one.
Colour that survives an accessibility audit
The first palette was built for brand recognition and failed on contrast. Measured against white, the primary blue reached 2.5:1 and the warning amber 2.0:1 — well below the 4.5:1 that WCAG AA requires for text. That mattered more here than in most products, because two of those colours carry the states this marketplace runs on: verificado and pendiente.
We kept the hues and split each role in two. The original colour fills large surfaces and buttons; a darkened variant carries text, icons and labels. Same brand, same semantics, legible in sunlight on a mid-range phone. No state is ever carried by colour alone — every status pairs its colour with a text label and an icon.
Where it landed
Prototyped to handover
MEIT was designed and prototyped to developer handover in 2021 and did not reach production. What the project produced was a complete two-sided service model, a design system covering around 110 screens across both flows, and an architecture for a marketplace built on verification rather than volume.
The clearest signal came from the research rather than from launch metrics. Across 20 interviews the first thing named, by both sides, was the same thing: the fear of a stranger in your home, and of entering one. Everything in this product that looks like friction exists to answer that.
How it would have launched
Depth before breadth: seed three localidades to real density rather than covering Bogotá thinly. Forty verified cleaners in one part of the city is a working marketplace; five spread across twenty is a directory with nobody in it. The search flow was built to hold that constraint — coverage is declared by localidad, and an empty result captures the demand instead of apologising for it.
Reflection
What I’d do differently
I would have acted on my own research sooner. Safety was the first thing both sides named, and the product I designed verifies only one of them. The client side onboards in five steps and proves nothing to the person walking into their home. I have since designed the other half — elective client verification, reciprocal ratings, and the ability to share a job with a trusted contact — but it should have been in the first version, because the interviews were pointing at it from the beginning.
The Aliado registration is too long. It is long for defensible reasons, but fourteen consecutive steps before any value is returned is a lot to ask of someone who has not yet earned a peso through the platform. I would split it: verify identity, let them into the app, and prompt for services, experience and availability contextually — when the first matching job appears.
The plan tiers were priced before there was data to price them against. Tiers should have shipped after we understood what an Aliado earns in a month, not before.
And the dispute path should have existed from the start. The flow handled a service that goes well and a payment that fails, but not a client who is unhappy after the work is done. I designed it next, and doing so changed the product rather than adding to it — payments became held rather than taken, and the activity list turned out to be the contract all along.
