Turning a logistics model into an offer people would accept
Envíos Extra was Mercado Libre's first crowdsourced delivery network: independent drivers, their own vehicles, launched in Mexico. Seven documents stood between a person and their first paid route. My job was the words.
Mercado Libre had always depended on third-party couriers. Envíos Extra turned everyday drivers into its own delivery network, from a first idea to a live product in Mexico in under a year. Nothing existed at the start: no product, no offer, no idea whether anyone would say yes.
I was the content designer on it. The whole proposition runs in one sequence: come and deliver packages with us, first we need seven documents from you, then you drive, then you get paid. My job was every word of that: the value proposition, the recruitment landing, the sign-up, the app, and the FAQs. Alongside three product designers, two product managers, and a large engineering team. In parallel I was working on Mercado Envíos Flex, whose team case study is public on Behance, in Spanish.
What made this hard
None of these were design problems on their own. Together they decided what the product could even promise.
A market that wasn't ours
Mexico, and a model Mercado Libre had never run. We could not assume anything about how drivers worked, what they earned, or what would sound trustworthy to them.
Seven documents before a single peso
Licence, vehicle registration, liability insurance, and four tax documents. Some take days to obtain. All were legally required before anyone could drive.
Drivers already had options
Most were already working for other platforms. The offer had to be honest and still better than what they had, or there was no network.
Envíos Extra runs on routes, not packages
Worth understanding before the rest, because the whole content problem sits inside this loop. Purchase volume shapes each route, and drivers choose which ones to take, a day ahead.
Orders shape a route
Based on purchase volume, the system builds an optimized route.
Goes out as an offer
Drivers see it about a day ahead: the zone, the pay, the hours.
Driver accepts or declines
Each driver chooses which routes fit their day. No obligation either way.
Loads up at the centre
The driver collects the packages and scans them in the app.
Delivers the route
Every stop completed through the same app, start to finish.
Gets paid, in-app
Earnings land in Mercado Pago. No separate system, no bank required.
Getting someone to say yes
This was a large team and most of it was not my decision to make. So here is both: what was happening in the project, and what I actually did in it.
Learning the market from the people in it
Mexico was not our home market and this model was new for the company. The research team ran qualitative sessions with drivers already working for other platforms.
I helped write the questions. We studied Uber, Cabify and other existing apps first, then built the guide around what we needed to know: what each tool did well, how drivers organised a day, and how much they earned.
That last one decided everything. If our offer couldn't compete with what they already made, nothing else was worth designing. I sat in on every session and passed follow-ups to the moderator.
Turning regulation into something you can say out loud
Mexican law decided what a driver had to provide before working. Legal and Product shaped the offer around that, before there was a screen to design.
I sat with the legal team in Mexico, not to learn what the documents were called but what each one was for. Localisation was its own step later; I wanted the Spanish right from the start, not translated into shape afterwards. It is hard to explain why you need something when you don't know yourself.
An offer that had to beat what they already had
The recruitment landing page was where a stranger decided whether this was worth their time, their vehicle, and a week of paperwork.
"Earn money delivering to Mercado Libre buyers." Choose your days, know the rate before accepting, get paid without a bank account.
I structured the value proposition and wrote the page. Two decisions I'd still defend.
Earnings as a range, not a promise. Pay is per route and moves with demand, so we showed the ceiling and said so. A confident single number is a lie people discover on day one.
"Get paid without a bank account" as a headline, not a footnote. Straight out of research: the trip to a bank was real friction. Paying into Mercado Pago kept the money inside a system they already used.
Seven documents, without losing the person
Sign-up had to collect everything the law required. Some of those documents take days to obtain.
"Wait while we complete your registration. We are reviewing your details before activating your account." Shown in Spanish, as it shipped.
It came down to a split: everything up front in the FAQ, so nobody hits a blocker halfway through with their data already in, and step by step in the flow, validating as it went rather than failing everything at the end.
And the part I'd call a real decision: we ordered the requests by lead time, not by convenience. The documents you had to go and apply for came first, so the paperwork was already moving while you gathered the rest at home.
Even the waiting screen does both jobs: it says what is happening to your application, and puts the FAQ within reach of the person most likely to have a question.
The route you confirmed and can no longer do
Drivers committed to routes in advance, often for the following day. Life happens between confirming and showing up, and a route nobody turns up for is a hole in the operation.
"This cancellation may affect your future requests… to avoid that, cancel before 20:00." Shown in Spanish, as it shipped.
The easy version of this screen is a threat. Cancelling affects your future offers, are you sure. That produces two things: anxiety, and drivers who quietly do not show up rather than press a button that sounds like a punishment.
So the modal is honest about the consequence and then tells you exactly how to avoid it: cancel before 20:00. Same rule, same policy, written as a door instead of a wall.
A warning with a way out is worth more than a warning. It turns a no-show into a cancellation with notice, which is a slot the team can still fill, and it keeps a driver who had one bad week rather than losing them to a screen that made them feel like a problem.
The questions that arrive after launch
FAQs and support, plus operations dashboards giving the team visibility into demand and performance as the network grew.
I wrote the FAQs, some anticipated from other platforms, others from what people actually asked once it was live.
They lived in a blog-style CMS where real formatting needed HTML, so the team normally waited on a developer to build a table. I knew HTML, so I did it myself, and taught the rest of the content team to do it too.
From choosing a route to getting paid
Every screen here needed words, including the ones nobody thinks about: the empty list, the scan that fails, the screen that confirms you are done.
None of it was built from nothing. We reused what already existed in Mercado Libre's other delivery apps and adapted only what Envíos Extra needed, which is part of why something this size shipped in under a year.
Scroll sideways →
The push knows where to go
One offer opens that route, several open the list. One less tap on a notification someone reads at a traffic light.
The deadline is a branch, not a clause
The 20:00 rule lives in the flow, so the interface can tell you which side of it you are on instead of burying it in terms.
Every state has words, including the empty one
"You have no confirmed routes yet, check what's available" is a screen where nothing happened that still tells you what to do next.
Looking at these screens again, we centred far too much text. Centred paragraphs are harder to read in any language that runs left to right, because every line starts in a different place and your eye has to hunt for it. At the time it was the house style across the company and I did not question it. I would now.
Four numbers and a decision
A driver reads this once and decides whether to give up a day. Every field on it was a content decision. It shipped in Spanish, so here is what each line says.
Hover or tap a number →
-
Earnings first, and largest
Ganancia · $ 642
Not "route", not "job". The word is the money, because that is the question being asked. Everything else on the screen exists to qualify this number.
-
Duration and the window, not one or the other
Duración 6 hs · Horario 08:00 – 14:00
Six hours tells you the cost. Eight to two tells you which part of your life it takes. A driver deciding between this and another platform needs both to compare properly.
-
Where it starts, named the way locals name it
Centro de distribución · Iztacalco
The neighbourhood, not a warehouse code. This is the line that tells someone in Mexico City whether the route is on their side of town before they read anything else.
-
The distance you don't get paid for
Av. Pdte. Plutarco Elías Calles 269 · Benito Juárez (3,8 km)
Getting to the pickup point is unpaid time, so we put the distance next to the address instead of leaving it to be discovered. A number that makes the offer look slightly worse, and makes it honest.
-
A commitment, not an interest
Confirmar recorrido
"Confirm", not "accept" or "take". You are agreeing to be somewhere at eight in the morning, and the button should sound like it.
Screens weren't the hard part. Context was.
- Built for a glance, not a read. Drivers navigated cities and managed packages under time pressure.
- Clarity over polish. Plain language and forgiving error recovery mattered more than visual refinement.
- Adapted, not translated. Written for Mexico from the start, not localised into shape afterwards.
What launching Envíos Extra changed
third-party deliveries replaced by the internal driver network
from the first ideation session to launch in Mexico
to semi-senior, mid-project, on the strength of the work on Envíos Extra
Where I learned to think in products, by writing them
I was in a room where everybody had been at the company longer than I had, on a product none of us had built before. What I took from it lasted longer than the promotion did.
I learned what good research looks like by shaping the questions, not by reading the findings afterwards. I still write them the same way: work out what decision the answer will change, then ask about that.
And I learned that in a product like this, the words are not a layer on top of the design, they are the product. Whether a stranger becomes a driver depends on what you ask for, in what order, and how you explain why. It is still how I work, and I have written about why content belongs at the start rather than at the end.
The HTML was not a one-off either. Learning enough to do the thing myself, rather than wait for someone else to have time, is the same move I made years later when I built an app end to end.
Hola, I'm Lu. Senior Product Designer, 8+ years across different industries. I start with the real problem, make complexity understandable, and stay close to the outcome.
More about me →