Fail early, not at the worst moment.
The proposal doesn't stop banks saying no. It moves the no to a moment with time to fix it, instead of a checkout with food in the cart.
Now I ask where a failure should land before I design the happy path.
foodpandaExploration 02Self-initiated, 2026
I wanted to save a new card. foodpanda wanted me to buy dinner first.
My debit card expired. I opened foodpanda to add the new one and there was nowhere to do it. The only way in is to spend: top up the wallet or place an order, and the card is kept only if that payment goes through. I didn't want dinner. I wanted the card ready for whenever I order next. So I mapped every route in, tested where people look for it, and designed card setup that lives in the account instead.
Role Research + design, soloApp foodpanda Pakistan 26.40.2Evidence 2 accounts, iOS + Android
Not affiliated with or commissioned by foodpanda or Delivery Hero.
Try it
Your card expired, or you have a new one. Where would you add it?
This is the real pandapay wallet. Tap where you'd go. Every answer tells you what that tap actually does.
Pick one.
Each answer opens the real screen.
0 of 4tried
0save a card without paying
The screen you land on shows here.
How it works today
foodpanda takes a new card in two places. Both cost money.
Top upRs. 250 or more
or
Place an orderfood in the cart
→
Card savedonly if the payment goes through
No payment, no card. There is no screen for "I just want to add it."
Why this matters
Most people who need to save a card aren't ordering right now.
Adding a card is a setup task, done on your own time. foodpanda has bolted it onto a purchase, so anyone who isn't buying food this minute has no way in.
The card expiredThe bank sent a new one. You want it ready before Friday, not during Friday.
Setting up for a parentYou're adding your card to their account so they can order. You're not hungry.
The bank blocked a cardA replacement arrived. Nothing to buy today, just a card to swap.
Moving off cashfoodpanda is pushing cash users to cards. Their first attempt happens with a hungry cart.
No time right nowAdd it on the train, use it at the weekend. Today, that's impossible.
Checkout is a transaction. A card needs a home.
Letting people add a card at checkout isn't wrong. Making it the only way is. Setup and spending are different jobs, and they want different moments.
The account: where a card should live
Add, verify, replace, choose a primary
- Add a card any time, with time for the OTP
- Replace the expired one without losing the rest
- See what's saved, what's verified, what's failing
→
Checkout: where a card is used
Pick, pay, done
- Choose from what's already saved
- Pay, with the bank check already behind you
- Add a new one if you must, and it lands in the account too
What I found
Five findings from the real app
Tested on my own account, repeated on a second account, and before touching anything I asked three friends where they would go first.
- 1People look in Payment methods. One card is there. Nothing adds another.
- 2Delete it, and only then does the app say where adding lives.
- 3Checkout trusts a method the account doesn't know about.
- 4Adding a card always costs money: a top up or an order.
- 5A card at checkout lives in memory, not in the account.
ObservedI saw it happen on my own account.
Confirmed on a second accountA friend repeated it on theirs.
Assumption, to confirmMy best reading. foodpanda would know.
ProposedDesigned, not shipped.
Finding 01
Swap, never add
People look in Payment methods. One card is there. Nothing adds another.
0Add buttons where all three looked
Before touching the app I asked three friends one question: "Your card expired, or you have a new one. Where would you add it?" All three said Account, then Payment methods. With a card saved, that screen shows the card, a Primary badge and one link: Change. Change opens the same list. You can pick a card or delete one. There is no Add.
Swap, never add
Finding 02
Goes to Home
Delete the card, and only then does the app say where adding lives.
Delete to addand lose the card you had
With no Add in sight, the next guess is "maybe it appears once the list is empty". So people delete the card they have. What appears is not a button but a message: save a payment method at checkout. It's the first time the app says where adding happens, and the price of finding out was the old card.
Save a payment method at checkout to view it here.Payment methods, once the list is empty. The only link goes Home.
Goes to Home
Finding 03
Ticked. Whose?
Checkout trusts a method the account doesn't know about.
1ghost method, ticked, nothing to inspect
Payment methods said nothing was saved. Checkout still had easypaisa ticked. Which easypaisa? No number, no name, and tapping the row shows nothing. You're about to pay with an account you can't see.
easypaisaPre-selected. Not in Payment methods.
Ticked. Whose?
Finding 04
Only after Rs. 250
Adding a card always costs money: a top up or an order.
Rs. 250top up, or a full order. Money moves either way.
Both routes that accept a new card charge you: a wallet top up of Rs. 250 or more, or placing an order. There is no way to give foodpanda a card without giving it money in the same breath.
Only after Rs. 250
Finding 05
Looks saved. Isn't.
A card at checkout lives in memory, not in the account.
16 digitstyped again, every time you don't pay
Say you have the card in hand, add it now, and plan to order tomorrow. Close the app: gone. Step out to Account and back: gone. Back out of Top up with "Save this card" ticked: gone. The card only exists once money has moved, and then it appears minutes later with no message.
Save this card for a faster checkout next timeThe checkbox at checkout. It doesn't, unless you pay.
Looks saved. Isn't.
Finding 01 · Payment methods
The pattern
Five findings. One pattern.
Five findings. One pattern.
A card only exists if money moves.
| Finding | What the person can't do | What it costs |
|---|---|---|
| 01 · A card, and no Add | Add a card where they look for it | A guess |
| 02 · Delete to find out | Keep the card they have | The old card, to learn adding lives at checkout |
| 03 · A ghost in the picker | See which account is about to pay | Trust in the screen that handles money |
| 04 · Every add is a payment | Add a card without spending | Rs. 250, or a full order |
| 05 · Memory, not the account | Add now, use later | 16 digits, typed again |
| Card setup has been bolted onto a purchase. Everything else follows from that. | ||
Every route
Let's walk every way in.
The task is the same each time: add a card and have it verified, without paying for anything. Three routes exist today. Pick one to see what happens and why it fails. None of them lets you do it.
What's at stake
Every card expires. So every customer hits this.
Not an edge case. A certainty, on a schedule the bank sets. Here's what it costs when it goes wrong, and what has to change.
For the person
Setup at the worst moment.
You find out whether your card works only when you're trying to buy dinner. If it fails: pay cash if you have it, or abandon the order. And the only way to check again is the next checkout.
For the order
A failed card is a lost basket.
Restaurant, dishes and slot already chosen. Payment is the most expensive place to lose someone.
For the business
Cards on file are the engine.
Repeat orders, one-tap checkout and top ups all need a working card already there. foodpanda is actively moving cash users to cards.
What has to change
Add a card anywhere, any time
Without a top up, without an order, without deleting the one you have.
What has to change
Verify without taking money
The bank check the card already goes through, moved to a calm moment.
What has to change
One list checkout can trust
What the account shows is what checkout offers. No ghosts, no surprises.
Before designing
What a payments team would say, and how I'd answer
I don't think this is an accident. A free add-card form creates real problems. The proposal has to solve them too.
A payments team would say
Free card forms are how fraudsters test stolen numbers.
A saved card that doesn't work becomes a failed order later.
Tying cards to top ups is how pandapay grows.
Our provider only stores a card after a charge.
The proposal
Every new card goes through the bank's own check, OTP included, and adds are capped per account per day.
New cards stay Unverified until the bank or a first payment confirms them. Checkout can tell the difference.
Top up stays one tap away. A card that already works makes topping up easier, not harder.
Visa and Mastercard mandate a zero-amount account verification in place of small charges. It's a network feature, not a wish.
The copy fix
The message is wrong. Fix that first.
It sends people to checkout to save a card, and checkout doesn't save it unless they pay.
Today
Payment methods
Save a payment method at checkout to view it here.
Back To Home
Back To Home
Points to a step that only saves the card if you pay. Links somewhere else.
Honest
Payment methods
Pay for an order with a card and it's saved here once the payment goes through.
Start an order
Start an order
Says when the card is kept. The link goes where the message points.
Nobody should have to buy dinner to tell foodpanda their card number. So card setup moves into the account.
The findings changed the design: not a better checkout form, a home for cards.
Add it.
Verify it.
Forget it.
One Add button in the wallet, one bank check that takes no money, and a status that tells checkout what it can trust. Everything else on the screen stays where it was.
pandapay › Payment methods, the only change
Payment methods+ AddThe only new button
Your saved methods 3
Mastercard•••• 3133
ⓘ Primary›Visa•••• 5000
ⓘ Unverified›+ Add
Opens the same type list checkout uses. The card is saved the moment you tap Add card.
Unverified
Saved, not yet checked. A Rs. 0 check with the bank, or the next order, clears it.
Why this one
Three reasons it beats the alternatives
1
The bank check already exists.
Every first payment on a card already triggers the bank's OTP. The proposal moves that check to a calm moment. It adds nothing new for the bank.
2
People already look there.
All three friends I asked went to Payment methods first. Putting Add there means nothing to learn and nothing to announce.
3
Nothing new to build.
foodpanda's own card form, its own type rows, its own list. One button, one status, one sheet.
Rematch
Same wallet, now with Add. Tap the same four places.
Pick one.
Same question, new wallet.
Today0 of 4save a card without paying
Proposed0 of 4save a card without paying
The screen you land on shows here.
Proposed
Card setup moves into the account. This is the system behind it.
An Add path, a verification step the bank already performs, and a set of states checkout can trust. Designed in Figma from foodpanda's own components. For each decision, what I chose and what I dropped.
Decision 1
Put Add where people already look.
Both Payment methods screens get Add. The Account screen itself doesn't change.
✓
ChosenAdd on both Payment methods screensOne shared list.
✕
DroppedA new "Add card" entry in AccountA third place for the same job.
✕
DroppedMake the checkout route clearerStill the wrong moment.
The wallet shows the primary method plus anything needing attention. The full list is one tap away.
Decision 2
Save first. Verify after.
Tap Add card and it's saved, marked Unverified. Nothing typed is ever lost again. Skip the check with Not now, and the card is still saved and ready at checkout.
✓
ChosenSave straight away, as UnverifiedThe status says what's left to do.
✕
DroppedVerify before savingLate OTP or a bank "no", and the card is gone again.
Decision 3
Check with the bank. Take no money.
A zero-amount authorisation asks the bank if the card works online. Nothing shows on the statement.
✓
ChosenVerify now, Rs. 0 checkAbout 30 seconds, sometimes an OTP.
✓
Also chosenOr verify with the next orderThe first successful payment verifies it.
✕
DroppedSmall charge, then refundDays to show, two statement lines, support tickets.
Checked on a live order: the bank asked for an OTP on the first payment, not the next. That one-time check is all the proposal moves.
Decision 4
Keep the rows people know.
The "Add new" rows from checkout take you straight to the right form. One tap, nothing to learn.
✓
ChosenThe same rows as checkoutCard, JazzCash, easypaisa.
✕
DroppedOne "Add a payment method" buttonTidier, one step more.
At checkout, radios appear only next to methods that exist in Payment methods. Nothing unlinked is ever pre-selected.
Decision 5
When the bank says no, keep the card.
Most failures are fixed in the bank app in a minute. The card stays, with the usual fixes and a Try again.
✓
ChosenKeep it Unverified, show common fixesTry again, or use another method.
✕
DroppedRemove the card, start over16 digits again for a bank setting.
And at checkout
Checkout, today and proposed
Three of the five findings happen on this one screen. With the account fixed, checkout only has to read from it.
123Today
1A method I never saved, already ticked.
2Saved cards and "add new" look identical. The chevron is the only way to add.
3Whatever you type on this screen lives here. Leave, and it's gone.
123Proposed
1Radios only next to methods that exist in Payment methods. Nothing unlinked is ever pre-selected.
2"Add new" is its own group. Adding here saves to the account, Unverified; the order verifies it.
3One list, read from the account. The primary card is already chosen, so most orders never open this screen.
The point of all of it
A declined card is fine. Finding out at checkout isn't.
The proposal doesn't stop banks declining cards. It changes when you find out, and what you can do about it.
Today
At checkout
Food chosen. You assumed the card would work.
- The app just failsNo reason, no fix, no card
- Pay cash, if you have it, or abandon the order
- The only way to check the card again is another checkout
- Type the 16 digits again next time
Proposed
In the wallet
A quiet evening. Nothing in the cart.
- The reason, on screenOnline payments off, wrong expiry, bank block
- Open the bank app, flip the switch
- Tap Try again
- The card stays saved the whole time
Don't make people fix their card at the exact moment they need it most.
The system
One list, four states
Checkout reads the same list as Payment methods. Nothing more.
! Unverified
Saved, not checked
Tap to verify now, or it verifies on the next order.
Primary
Verified, default
The first verified card becomes primary. No extra badge.
Couldn't verify
Bank said no
The card stays. Try again when the bank app is sorted.
✓ Verified
Just passed
Done, or set it as primary. The list behind has already updated.
The flow
Add on Tuesday. Pay on Friday.
Three chapters, each in its own colour. No top up, nothing typed twice, and the card is already there when the food is.
Tap a chapter, scroll sideways, or use the arrows
Before and after
Drag it.
One handle moves both screens. Today the wallet offers Change, which only swaps cards. Proposed adds Add; swapping happens on the row itself.
The wallet
The checkout picker
Elsewhere
Other apps let you add a card without paying
Most take a small hold and release it. A Rs. 0 check is the cleaner version of the same idea, and it's what the card networks now ask for.
Hold, released
Careem
Profile › Cards and accounts
a small temporary authorization hold of amount (not a charge)
Hold, voided
Uber
Wallet › Add payment method
usually $1, $2, $5
Add before remove
DoorDash
Account › Payment
required to first add a new card before removing the existing card
Pay first
foodpanda today
Checkout, or inside Top up
Saved only after a successful payment
Rs. 0 check
foodpanda proposed
Wallet › Add
Saved at once as Unverified. A Rs. 0 account verification with the bank, or the next order, verifies it
Proposed
How it could backfire, and how I'd know
Risk
Fraudsters use the free form to test stolen card numbers.
How I'd know
Adds per account spike; verification failures cluster on new accounts.
What I'd do
Cap adds per day, OTP on every add, lock adding after three failed checks.
Risk
Unverified cards pile up and still fail at the first order.
How I'd know
Share of first orders placed on an Unverified card, and their decline rate.
What I'd do
Nudge to verify before checkout; run the Rs. 0 check in the background after adding.
Risk
A card can't be checked with the bank.
How I'd know
Verification failure rate by bank.
What I'd do
The card stays saved as Unverified, and the next order verifies it. Never a charge and refund.
Kill criteria
If first-order declines go up after launch instead of down, Add stays but saving waits for verification. The account keeps the card either way.
What success would look like
Signals to watch. Nothing has shipped.
If card setup moves into the account, these are the numbers that should move, for the person and for foodpanda.
| Signal | What I'd watch |
|---|---|
| Cards saved outside checkoutPeople set up on their own time, not mid order. | Adds from Payment methods vs adds from checkout |
| Fewer first-order declinesThe bank check happened before the food. | Decline rate on a card's first payment |
| Fewer cash fallbacksA working card is already there. | Switches to cash after a card failure |
| Less card re-entryNo more typing 16 digits twice. | Repeat card entries per account |
| More cash users become card usersfoodpanda's own goal, with a first attempt that can succeed. | Share of cash-on-delivery users who save a card before their next order |
| Fewer "where did my card go" ticketsNothing disappears any more. | Payment-method support contacts per 1,000 orders |
Five-person task test
"Your card expired. Add the new one so it's ready next time." Half on today's app, half on the prototype. Time to find Add; who gives up or starts an order to do it.
To confirm with the payments team
- Why did checkout only offer a new card once my list was empty? A friend's account with three methods still got the option, so it isn't a limit.
- Which easypaisa was pre-selected? A fallback or my deleted link. Either way it should show whose it is, or not be selected.
- Do banks support Rs. 0 checks? Yes. Rs. 0 account verification is a Visa and Mastercard requirement, so no money is ever taken to check a card.
Key takeaways.
Four things this study changed about how I work.
Setup is not a transaction.
Adding, verifying and replacing a card are account jobs. Checkout should read from the account, not stand in for it.
Now I ask which screen owns the data before I draw the form.
Verify, don't charge.
The bank already checks every new card. The networks mandate a Rs. 0 check for exactly this. A Rs. 250 top up was never the only way to know a card works.
Now I look at what the rails already do before I add a step.
Design for where people look.
Three out of three went to Payment methods. The fix belongs where the mental model already points, not where the data happens to live.
Now I start from the first click, not the data model.
The point
A new card should be ready before the food is.
Card setup moves from the checkout into the account, with a verification step the bank already does and states checkout can trust. Seventeen screens, ending where the problem started: at checkout, with the card already there.
Next, from my case studies
Designing and shipping a luxury property website, end-to-end.
That was a self-initiated study. Here's one of my full case studies.
