How I would fix a bank's online card application

I run online sales for an insurer in Tashkent. A bank's online card application looks like a different product, but it is the same problem: a regulated product, an identity check, a payment instrument at the end, and a customer holding a phone. Three visitors in four come to gross.uz from a phone, and I would expect at least as many for a bank.

This is how I would approach a card application if it were my product. It is not a review of any bank's app: the steps and hypotheses below are what I would check first, and the numbers would come from the bank's own funnel.

Make the funnel readable first

Before changing anything I would set one goal per step:

1. Opened the card page
2. Started the application
3. Passed identification with MyID
4. Filled in the remaining details
5. Application approved
6. Card issued, in the app or for delivery
7. Card activated
8. First payment within seven days

The last two steps matter most and are measured least. A card that is issued and never used costs the bank money and brings nothing. At Gross the step-by-step funnel showed that a third of the June growth came from the path to the calculator, not from the calculator itself. Without a goal per step we would have credited the wrong change.

Five hypotheses I would test first

1. Show the outcome before asking for data

On the OSAGO page the price used to come after the form. When it moved to the first step, more people finished. For a card the outcome is the conditions: the fee, the cashback, the limit, how soon the card is ready. If a customer has to type a passport number before seeing them, many will leave at that field. Hypothesis: the conditions on the first screen, the data after.

2. Identification first, so the form fills itself

MyID already does the hard part: it confirms the person and returns their details. When we moved sign-in to MyID at Gross, there was no password to forget and nothing to type twice. For a card, MyID at the start can fill in the name, the passport and the address, so the customer only confirms them. Hypothesis: fewer fields after MyID, more people through step 4.

3. Meet the customer in Telegram

Gross sells OSAGO and travel insurance through a Telegram Mini App, and it already brings sales. Many people would rather not install another app just to try a product. Hypothesis: a Mini App that starts the card application, with MyID and the same conditions screen, brings applications from people who never open the bank's app.

4. Activation is part of the product

The application ends when the card is used, not when it is issued. I would design the first week as a product: a message the day the card is ready, a nudge to link it to the bills people pay every month, a reason to make the first payment. Hypothesis: a first-week path raises the share of cards with a payment in the first seven days.

5. Designed for the phone, not squeezed onto it

The mobile versions were part of the June redesign at Gross, not an afterthought. A form built for a desktop and squeezed onto a phone loses people at every dropdown and date picker. Hypothesis: a phone-first form, one question per screen where it helps, and the keyboard that matches each field.

How I would test them in a bank

The users of the US logistics software I managed noticed every change, so every release went the same way: customer interviews, a pilot with a few customers, then everyone. A bank needs that order even more. I would talk to people who started an application and stopped, release each change to a share of the traffic first, run an A/B test where the volume allows it, and read the result in the funnel above.

Regulation belongs in the plan from the first day, not the last week. At Gross I work inside the rules for a joint-stock insurer, from the disclosure section to the payment and identity steps, and the US product worked inside FMCSA rules. In a bank I would bring risk, legal and security into the problem before the design, so that a good idea does not die in approval.

What I would measure in the first 90 days

• The completion of each step, and the step that loses the most people
• The time from the first screen to an approved application
• The share of issued cards with a payment in the first seven days
• Applications by channel: the site, the app, Telegram

The first month is for making the funnel readable and talking to customers, the second for the first two hypotheses, the third for activation. Every release is linked to the step it should move, so the team and management look at the same numbers.