JPMorgan Chase · DCE CoBrand Card
Marriott Bonvoy
Benefit Add-Ons
Designing a scalable platform that lets Marriott cardholders personalize their card with premium benefit add-ons — increasing relevance, driving engagement, and unlocking new revenue streams.
Project scope
Overview
The opportunity
Cardholders don't always feel they get the most value from their existing card benefits. Research showed that the ability to 'add on' more relevant benefits — benefits better aligned to individual lifestyles — was a compelling proposition that customers were willing to pay for.
This project explored how to build that capability, using the Marriott Bonvoy card as the proof-of-concept. The goal was to design a scalable UX for adding benefits to a credit card, then learn from this pilot to scale across the broader card portfolio.
Design Challenge
How might we create an experience that makes 'adding on' benefits clear and easy to understand — to the extent it's worth paying for?
Research & Strategy
Mapping the cardholder journey
Before designing screens, we mapped the end-to-end journey a cardholder goes through — from first hearing about benefit add-ons to renewing or changing them. This gave us a shared framework for identifying where design could have the most impact.
Who we designed for
Understanding the cardholder mindset
I didn't even know I could add things to my card. I thought the benefits were just… fixed when I signed up.
— Chase cardholder, mid-tier rewards segment
Quantifying what drives add-on choice
When I could see exactly what I'd pay versus what I'd get back, it felt like a no-brainer. The math made sense.
— Participant, DCE preference study
Six opportunity areas to explore
Three principles to guide concept development
Show the math. Users need to see exactly what they pay and what they get — not a vague promise of "savings."
Don't assume cardholders understand how add-ons work. A brief, honest explanation converts skeptics into buyers.
An add-on recommendation tied to an upcoming trip lands 3× better than a cold browse experience.
Five moments that define the add-on experience
Concept A · Simplicity & Value
Streamlined and direct
A minimal benefit list with no onboarding preamble. Focused on a direct comparison with existing card incentives to make the upgrade value immediately legible.
Screens
Straight to the list — value front and center
×
Concept B · Education & Management
Context before commitment
Onboarding screens with more detailed recommendations, educational content about how benefits work, and the option to explore and add more benefits before paying.
Screens
Educate first, then let users decide
BOUNDLESS
Concept A — Streamlined
Action-first approach minimizing steps between discovery and add-on enrollment.
Concept B — Contextual
Education-first approach building comprehension before commitment.
Concept Testing · 5 Sessions
What testing told us
We tested both prototypes with 5 existing Marriott cardholders simultaneously — comparing approaches to find what to keep and what to combine.
Synthesis
The merged direction
Neither concept won outright — but together, they pointed at the right answer. We combined the clarity of Concept A's value-first list with the confidence-building of Concept B's education and the hub management model both concepts hinted at.
A streamlined browse experience that leads with value, backed by an education layer for those who want it — and a management hub that makes the purchase feel like the beginning of a relationship, not the end of a transaction.
Final Design · 10-Screen Flow
The complete experience — from first touch to ongoing management
The final design brings together the merged direction into a complete end-to-end experience — from onboarding education through benefit browsing, purchase, and ongoing management.
Final Design · Complete Screen Library
All 10 screens across the purchase flow, onboarding, and concept explorations.
BOUNDLESS
×
×
BOUNDLESS
×
×
Vision State · What's Next
From catalog to personalized For You
The 5-session concept test surfaced one finding strongly enough to be the next product hypothesis: contextual recommendations outperformed cold-browse by roughly 3× in our 5-session concept test. The current direction is a clean catalog — the next iteration applies that finding at scale.
An AI-personalized "For You" surface uses cardholder behavior to rank the catalog by predicted utility. The existing browse view stays as a deliberate alternative path for serendipity.
Rank by predicted utility
A model trained on transaction history, Bonvoy stay data, and seasonal patterns ranks each cardholder's add-on catalog. The most-likely-to-be-used add-on rises to the top with a confidence match score.
Plain-language reasoning
An LLM generates a one-line "why" for each recommendation, grounded in the cardholder's actual spend ("You spend 3× more than average on dining in NYC"). Concrete, falsifiable, defensible.
Proactive nudges in-context
When the model detects a Bonvoy property booking or a travel-pattern change, it proactively surfaces the matching add-on as a dashboard nudge — not buried inside a catalog visit.
The personalized view leads; the existing catalog browse is one tap away. No information is hidden — the change is what surfaces first.
"Based on your last 6 months: 12 hotel nights · 18 dining purchases · NYC" makes the input data legible. The model isn't a black box — the cardholder can see what fed it.
Top match (96%) gets the hero treatment. Reason line is grounded in the user's actual behavior ("3× more than average on dining in NYC"), not a generic "you might like." Dismiss button trains the model.
Trust affordances: explainability surface for the "why," and explicit user control to pause categories or reset the model. Failure state: low-confidence output reverts to category browse with no AI ranking.
Inputs are exclusively things the cardholder did with their card. No demographic, no inferred preference. Behavioral signals are correctable; lifestyle inferences are insulting.
"Your dining spend is 3× higher … 47% happens in NYC." The chain spend → rank in one sentence. Pattern borrowed from Spotify "Made for You" attribution, rendered for financial services.
Most products bury opt-out in Settings → Privacy → Personalization. Promoted because confidence to engage in the first place depends on knowing they can step away.
Stating the threshold makes the failure mode predictable, which is a precondition for trusting the success mode.
The system refuses to fake personalization for new cardholders. Showing low-confidence guesses would burn trust before the model has a chance to earn it. The "For You" tab is dimmed; Browse is the active default.
Cardholders who land here see exactly what they'd see today. No new pattern to learn. The AI layer is purely additive — the failure mode of the new system is the success mode of the previous one.
-
Amazon "Recommended for you"The gold standard for ranked retail-style recommendations; mirror the row-of-cards pattern in a financial-product context.
-
American Express OffersDirect competitor surface; their personalized merchant offer ranking is the closest peer pattern in cards.
-
Uber One benefitsThe "you've used X of Y" personalized utility framing is a strong reference for showing AI-driven relevance without being preachy.
My Contributions
What I worked on
End-to-End Flow Design
Mapped discovery, purchase, and benefit management flows — ensuring the experience felt like account management, not a shopping cart.
Concept Development
Designed both Concept A and B — exploring different approaches to simplicity vs. education to stress-test which model resonated most with users.
Value Communication
Designed clear, honest interfaces that present benefit details, costs, and terms — building the case for value without pressure or hidden complexity.
Legal & Compliance
Collaborated with legal and compliance teams to integrate disclosures thoughtfully — making required information feel helpful, not intimidating.
Prototype & Synthesis Support
Built and iterated on interactive prototypes for both concepts, and helped synthesize concept testing findings into the merged direction.
- Aligned 3 product teams on shared add-on framework
- Presented concepts to VP stakeholders for sign-off
- Partnered with legal, compliance, and engineering on constraints
- Synthesized findings from 2 rounds of user testing
Key Learnings