MoMo PSB

A bank that lives on a phone, and a door for businesses.

I redesigned the app in sprints. I started with a heuristic review of the existing app, then studied how other finance apps handle the same flows, using Mobbin, and moved flow by flow: sign-up, sign-in, home, add money, transfers, airtime, data, bills and statements in the first release. Every flow went to engineering with its error and empty states already drawn.

After the first release the requests kept coming: 64 of them, each on its own Figma page, from tap to pay to a text size setting. Alongside the app I designed MoMo Corporate, the web platform where a business opens a corporate account and MoMo's staff check it. Both had the same job: tell people what is happening to their money.

Say it failed, the moment we know.

The clearest signal in the whole project was a pair of screenshots. A customer bought ₦50 of data. The app said the request had been submitted and was processing. At the same moment, an SMS from the same bank said the payment had failed.

Then I went flow by flow and drew what failure looks like in each one: transfers, bank transfers, airtime, data, bills, betting, beneficiaries, sign-in, sign-up and forgot PIN. Each error says what happened in plain words and gives a way forward. When the connection drops, the error offers the USSD code, so a customer can still buy data without the app.

Warn people before they send.

A transfer to another bank can fail on the other bank's side, after the customer has done everything right. From their phone, a failure at the other bank looks the same as a failure at MoMo. So the next question was: can we tell people before they press send?

The rest of the transfer follows one pattern, the same for every service in the app: confirm who you're paying (the name comes back from the bank), review the amount, the fee and the total, enter your PIN, and get a result with a receipt. A customer learns it once, sending money, and meets it again buying airtime, paying for electricity or funding a betting wallet.

A sign-in that doesn't wait for a text.

The first question in MoMo's own FAQ was "Why didn't I receive the SMS OTP code?". A code that arrives late, or never, stops someone from reaching their own money.

Biometrics needed the same care. Signing in with your face is optional, and using it to approve payments is a separate switch, behind a plain statement: if you don't want to transact with your biometrics on this phone, don't turn this on. A phone that someone else uses can see "Not you? Switch account" at sign-in.

Regulation you can read.

A payment service bank works inside rules set by the Central Bank of Nigeria. A customer's tier decides how much they can send in a day and hold in their account, and moving up a tier means proving more about who they are. I couldn't change the rules. I could decide whether a customer meets them as an error or as a clear path.

So the tier is on the balance card from day one, the limits for all three tiers sit on one screen, and the upgrade is one tap from home. When someone types an amount above their limit, the field says so before they reach the PIN. When the CBN asked banks to update customers' KYC details, the request came in the app's own words, with the reason: to comply with the CBN's requirements.

Two sign-up paths came from the same rules. An MTN customer starts with their phone number and NIN. A customer on another network adds their BVN and a selfie check. And when an account is blocked because a NIN didn't validate, the customer can unblock it themselves, from the blocked screen, without a branch or a call.

Small things, for people who pay every day.

Most of the 64 requests after the first release were small, and they're the ones people notice every day. Frequent amounts as chips, so ₦500 of airtime is two taps. A Self or Others switch, so buying for your mum doesn't mean typing your own number away. Repeat a payment from its receipt. A text size setting, from small to extra large, for people who hold the phone at arm's length.

And a few that made the app feel local. Receipts with themes for Valentine's, anniversaries and the new year, and one for the most Nigerian request of all: "Urgent 2K". Payments by NFC tap and QR scan at the counter, with the fee shown before the PIN. Transfers to other African countries, with the exchange rate on the confirmation. The rating prompt only appears after a successful transfer, when people have a reason to be kind.

The same promise, for a business.

A business account carries more risk than a personal one, so opening it takes more: company registration, the directors, the owners, a board resolution, proof of address, and checks on every person behind the company. Before Corporate, that ran on forms and follow-ups, and nobody could see where an application was.

I designed it as one path with five stages, and a person who owns each one. The business only ever sees its own stages. MoMo's staff see all of them.

Onboarding checks. Compliance decides.

Two teams review every application, in order. The support analyst checks that each document is there, readable and right. Then they send it to compliance, who look at the people behind the company: are any of them politically exposed, are any on a sanctions list, how risky is this business? Each team has its own queue and its own dashboard, and the hand-over between them is one button: Send to compliance.

Compliance changed the most between versions. In the first version, compliance uploaded two documents: their compliance check and a risk assessment. In the update, a third step arrived for businesses rated medium or high risk: screening for politically exposed people and sanctions. Low-risk businesses skip it. The business's own form gained the matching question: are you, or a family member, politically exposed?

Ask about each person once.

The application form is where a business either finishes or gives up. In the first version it asked about directors, then shareholders, then signatories, each in its own section, each asking for a BVN, an ID and the same personal details. In a small company one person is often all three, and they were typing themselves in three times.

Smaller changes in the same spirit: expected income became a band to pick instead of a number to guess, and the forms the business has to sign are offered for download before it starts, so nobody discovers a board resolution on step four.

Show where each application is stuck.

The first dashboard counted activity: invites sent, invites accepted, registrations completed, compliance checks done. Useful for a report, but it couldn't answer the question a manager asks every morning: where is everything stuck?

The updated dashboard counts by stage instead, from started KYC to account opened, and leads with what needs a person: "20 supplier forms waiting for compliance check". Behind it, every action is logged with who did it, their role, the account and the time, and the reports the business asked for are a click away, including a weekly summary sent on a schedule to compliance, marketing and management.

Built from one set of rules.

Both products sit on the same colour system, which the file calls Sunshine: white for most of a screen, MoMo blue for structure, and yellow, MTN's yellow, kept for the things that need a tap. The rule is written beside the palette, 60, 30 and 10, so anyone picking a colour knows how much of it a screen can take. The colours are set up as tokens with a light and a dark value, which is how the app got its dark mode without a second set of screens.

Handing over mattered as much as the screens. Each flow went to engineering with its error and empty states, changes from the business analysts and product owners were noted on the page where they happened, and the open questions sat beside the screens they belonged to, like "Retries limit screen, what happens?" on sign-in. I'd rather an engineer find the answer in the file than in my messages.

Shipped, release after release.

The app shipped on iOS and Android and kept shipping: the first release in three sprints, then 64 feature requests on top, the bank network monitor and the passcode sign-in among them. MoMo Corporate shipped with its staff side and is live today; a later update added role management, the audit log and the reports. After launch, onboarding drop-off fell 30%, transaction success rose 20% and corporate invitations rose 20%. Those numbers cover whole products, so each decision above still names the narrower signal that would show its own effect.

How the work moved: business analysts and product owners brought requests, often from customer complaints and from compliance; I turned each into a Figma page with every state drawn; engineers built from that page. When a request changed, the change was noted on the page where it happened, so nobody had to reconstruct a decision from chat.

Funmi never sees a success rate being worked out, a compliance queue or an audit log. She sees a name come back before she pays, a plain answer when something fails, a receipt she can show, and one day an account number for her company.

What I'd do differently:

1. Agree the numbers on day one. Each decision above has a "how we'd know". I'd put those to the product owner at the start of each release and ask for them back after it, so the next release starts from evidence.
2. Give the business the tracker staff have. In the designs, after a business submits, it sees that review takes 48 hours, and then waits. I'd show it the same stages MoMo's staff see, with the one it's in.
3. Test the monitor's words with customers. "Success rate" is a banker's phrase. I'd test whether people read 40% as "don't send now" before trusting the screen to change behaviour.
4. Put the colours in code from the first sprint. In 2023 we weren't building with AI tools yet, so the Sunshine tokens lived in the Figma file and engineers built from it by hand. Today I'd use AI with the engineers to turn the tokens into code at the start, so the app and the file can't drift apart. It's how I build my own products now.