Helping customers bank across Discover and Capital One.

The customer lived in two banks at once

The acquisition could not behave like a single switch. Card, bank, and loan accounts lived on different systems and had to move at different times. That meant a customer with several Discover products could have one account at Capital One while another remained at Discover for months.

Capital One projected that 3.2–3.7 million customers would pass through this state across the migration. Internally, we called it the seam: a temporary but consequential gap in which “I bank with one company” quietly stopped being true.

The smallest cohort carried the largest risk

Research, product analytics, call-center themes, and post-conversion survey data pointed to the same four sources of friction:

• Dual-ecosystem servicing. Customers had to manage accounts across both apps.
• Uncertain support. It was unclear which company owned a problem when something went wrong.
• Notification overload. Both sides generated alerts and required actions.
• Payment confusion. Customers did not always know where or how to pay each account.

The payment problem created a business case as well as a customer one. The migration carried an estimated ~$250M payment-loss exposure, and the in-seam population was the most vulnerable: a customer who cannot tell which app owns which bill is more likely to miss a payment.

“It wouldn’t let me move both cards… so I haven’t used that converted card lately. I’m kind of waiting to see how it goes.” — In-seam research participant

From an ambiguous segment to a shippable scope

I led the strategy and solution within HCX Design, owning the problem reframe, concept direction, Wave 1 MVP scope, and alignment across Bank, Card, Payments, Content, Research, and Technology. The work moved through four linked activities:

• Make the segment concrete. I anchored the work on one representative multi-product customer and mapped the low-fidelity end-to-end journey.
• Triangulate the evidence. Interviews, analytics, call themes, and survey data kept a single anecdote from driving the design.
• Pressure-test feasibility early. The experience depended on new backend detection that could identify customers with accounts still waiting for a later wave.
• Align across lines of business. The solution had to behave as one coherent experience across multiple teams and two brands.

Bridge the seam. Don’t rebuild the bank.

The useful insight was that we were migrating accounts, not customers. The experience did not need to recreate Discover inside Capital One. It needed to preserve the customer’s mental model of one relationship while the underlying systems were mid-move.

Three principles shaped the direction: reassure rather than add to the fear of loss; make the safe payment path the easy path; and build the smallest intervention that removed the most friction. “Small but mighty” became a strategy, not an apology for the deadline.

The Ghost Tile and Companion Sheet

On the Capital One home screen, below a customer’s real accounts, I introduced a Ghost Tile: an account-shaped Discover element with no balance and no standard account controls. Its job was quiet and specific—acknowledge that Capital One knew about the remaining Discover accounts, set expectations for a future move, and provide a door back to servicing.

Tapping the tile opened a dismissible Companion Sheet that explained what would happen next, linked customers back to Discover to service their remaining accounts, and kept Discover-specific help one tap away. I deliberately left account-level personalization out of Wave 1 so the MVP could ship without waiting for more data plumbing, while preserving a pattern that later waves could enrich.

The other side of the seam

The bridge also reached into the Discover experience. A post-conversion sheet showed essential read-only information and deep-linked customers into Capital One servicing and payment flows. Migrated accounts were visually separated, and the fuller treatment ran for roughly 60 days on a timer that teams could adjust as real behavior emerged.

The live experience changed behavior

Wave 1 served 56.9K in-seam customers. By the late-April 2026 read, the experience had generated 120.3K tile views, 59.3K deep-link clicks into Capital One servicing, and 22.3K payments submitted through deep-linked pathways. Customers were using the bridge repeatedly, and traffic moved far more often toward Capital One than back to Discover.

On-time payment rates stayed stable through the 60-day suppression window. Delinquency trended lower among in-seam customers who engaged than among comparable single-card customers, an early directional signal consistent with the hypothesis. Capital One logins also held after the Discover-side links disappeared, suggesting customers had adopted the destination rather than only relying on the temporary bridge.

What made it hard

• The experience could not exist until the data did. Identifying an in-seam customer had to become a Wave 1 requirement.
• Two ecosystems had many owners. The customer saw one journey where the organization saw Bank, Card, Payments, Content, and two applications.
• The deadline did not move. The migration date forced clarity about what mattered enough to build.
• Anxiety was part of the task. Fewer alerts and clearer words often mattered more than adding another feature.

What I’d carry forward

Scope to the seam, not the whole system. Make the safe path the easiest path. Build temporary bridges so they can flex with behavior. And use the first wave to earn the next: Wave 1 was intentionally simple, but every decision left room for richer communication and personalization without rebuilding the pattern.