Cross-border transfers
Who receives it, how much lands, when it landed.
Fintech · Cross-border transfers
Bancolombia: who receives it, how much lands, when it landed
A four-day design challenge for Bancolombia's Experience Design team, judged by that team. I audited the live app before drawing a single screen, and designed one product for both directions when the brief let me pick just one.
The brief
Four days, a real panel, and a choice I did not take
Assumptions
What I bet, before showing anything
Starting point
I went into the product before writing a problem
On the Saturday my own audit said the opposite: that receiving from abroad was an empty table with no action on it. I went back into the app on Sunday, found the screen, and corrected the audit in the three places where it was wrong. The framing did not get weaker for it. The missing screen is not the one nobody designs. It is the one that exists and is incomplete.
The problem
The bank already knows how to accompany a transfer. It does it inside Colombia.
| With Bre-B, inside Colombia | Across the border, today | |
|---|---|---|
| Who receives it | With Bre-B, inside ColombiaIt shows you the account holder's name before you confirm. | Across the border, todayI check the details myself, digit by digit. |
| How much lands | With Bre-B, inside ColombiaThe exact amount arrives, at no cost. | Across the border, todayHow much actually landed, I find out once it has landed. |
| When it landed | With Bre-B, inside ColombiaTwenty seconds, and it tells you. | Across the border, todayThe confirmation comes from the other person, over WhatsApp. |
What I take from Bre-B is the experience, not the rail. The transfer still goes over SWIFT, with its days and its costs, so I cannot promise free and I cannot promise instant. What I can do is make the app answer. They are the same three questions whether you send or receive. What changes is who starts.
How I got there
Three statements I threw out first
Research
Five interviews, ten patterns, five that changed the design
Three things the evidence does not support, and I said so in the file. Paperwork is not the pain when the purpose is asked in human language, although the raw regulatory code in a dropdown is. Nobody lost money, so I do not claim people lose money. I claim they behave as if they could. And one of my own questions failed, because it used my vocabulary instead of theirs.
Two archetypes
Daniel sends. Mariana collects.
Daniel, 42, sends to a third party
Mariana, 38, collects from abroad
Mariana was badly built at first: her problem came from a single interview and was presented as the problem of all three people who receive. I rebuilt her on the pattern that is three out of three, that the exchange rate governs the income and there is no visibility forwards or backwards. What one person said did not make it into the archetype.
Benchmark
The advantage nobody is communicating
- Bancolombia
- No commission for receivingCorrespondent banks can still charge, and the bank's own site says so.
- BBVA
- About USD 20 to the recipientPlus an administrative charge in pesos.
- Davivienda
- USD 12 digital, up to USD 20 in branchSources differ slightly, so it goes in as a range and not as an exact figure.
- Banco de Bogotá
- Not available to individualsSWIFT into corporate accounts only.
The proposal does not fix a broken product. It makes visible an advantage that is currently buried.
The decision
Two entrances, one task, three points of divergence
The risk is written on the board itself: unifying is harder to design and easier to criticise than picking one lane. What lands is the lead number and what you send is secondary. The rate shows the official reference beside it so the margin is visible. Deadlines are stated with a date and a time, never in business days.
The screens
The missing screen, and the money made legible
Closing the loop
The decision I care about most, and the only one that comes straight out of research
The move from A8 to A9 is not triggered by the sender. It is triggered by the other person, from the link. No competitor in the benchmark designs that close.
The only transition in the file executed by someone who is not the user
System and handoff
Contrast measured by script, not estimated by eye
The handoff has two routes. The one I deliver here is Dev Mode, with per-screen annotations and data contracts. The other is a branch on the repo with the tokens as CSS variables, which I did not do because I have neither the repo nor the bank's real design system. The difference between the two is not the tool. It is where the source of truth lives.
How I used AI
AI proposes, I verify and decide
The tool proposed
I verified
On Sunday it ran the other way, and that is the part worth telling.
I proposed
The reviewer pushed back
The rule holds in both directions: framing and judgement are not delegated, and verification runs both ways. I also designed for AI inside the product, which is a different thing: autofill the beneficiary details from a photo, check that the ABA and the bank agree, suggest the purpose from history. Always with the person reviewing before anything continues.
What I did not verify
Declaring it adds. Hiding it takes away.
- I did not open a dollar account for a design exercise, so the send flow beyond the requirement page is still an assumption.
- I did not see what the app does when a transfer actually arrives. I saw the settings screen, not a crediting.
- I did not audit the desktop version of the web branch, or the BBVA and Davivienda flows.
- n=5, all from Medellín, all highly digitally literate, and one of them is the designer.
- No interviewee lost money. The fear is documented. The failure did not occur in the sample.
- That the currency paperwork can be resolved inside the app is a declared assumption, and the first question I would ask the team.
- Two opportunities are declared and deliberately out of scope: keeping the money in the currency it arrived in, and the rate on recurring collections.
What I take from it
