# Eight product teams on one Home zone, with nearly twice the click-through

PayPal · Nov 2023 – Mar 2024, measured 18 months later
Author: Henrique Senra (https://henriquesenra.com)
Web version: https://henriquesenra.com/cases/paypal-financial-products-zone

> I led the redesign of the Financial Products zone on the PayPal app's Home screen, so that eight product teams could ship on one zone without asking the Home team. Its click-through rate is 27%, against 14.5% for the zone it replaced.

## Key outcomes

- Click-through rate of 27%, against 14.5% for the replaced zone, over the same seven days.
- About 80% of all Dashboard zone impressions, on a screen with about 50 million daily impressions.
- 50 or more requested instrumentation fields cut to the 4 or 5 that were shared.
- Revenue gained: click-through measured. Revenue figures are domain-team projections.

Tags: Platform, Federated governance, Roadmap, Home

## TL;DR

- **Problem.** Eight product teams needed room on the first carousel of the PayPal app's Home screen, and the zone they shared had become unreadable after years of uncoordinated changes.
- **My role.** I led the multi-team effort and owned the structure, the schema and the refusals. Ranking models and the business cases stayed with each domain team.
- **Decision and cost.** Home owns a parent component and each team owns its children. I refused nearly every instrumentation request and every complex eligibility rule at the parent. Machine-learning ranking was postponed.
- **Outcome.** Click-through is 27%, against 14.5% for the zone it replaced. Business end: revenue gained. The click-through is measured. The revenue figures belong to the domain teams and are projections.

## Context

In November 2023 a new initiative inside PayPal, called Quantum Leap, set out to speed up checkout and the app experience, with checkout as the one key metric. Results were due as early as January, with users reaching the new experience in March. For the Home and Navigation team, where I was the platform product manager, that meant rebuilding the whole Dashboard (the Home screen's list of zones) nearly from scratch.

The Financial Products zone was the hardest part. A zone is a horizontal container on the Dashboard. This one replaced the Financial Snapshot zone, the first carousel a user sees when opening the app, which held cards mostly about balances. Different domains governed those cards, each with its own product goals and legal requirements, and years of changes by several teams without central governance had left hundreds of assets, each with its own logic.

Three problems had to be solved together.

- **Information architecture.** The new zone had to show financial information and also guide the user to an action, called a Next Best Action. The carousel had a limited number of tiles, and the number of possible actions had no limit.
- **Screen space.** The Dashboard had already gone from 25 zones to 18, and domains still competed for the space above the fold.
- **Federated governance.** Every domain team was used to its own data source. There was no time to change that, so the Dashboard risked timing out while waiting for their answers.

The zone and its cards are configured in PStudio, PayPal's internal no-code platform for personalization, whose messages go live without a new app release. The platform team that built PStudio wrote the backend support for the hierarchy described below. Home Analytics handled instrumentation and the Home backend team handled rendering.

The zone ended with eight domains: Balance, Credit, Pay Later, Giving, Crypto, Savings, Cashback and Credit Card. The final configuration had eight context-specific interactions with about 120 unique messages, localized for about 200 countries.

## What I owned

I ran Financial Products myself. The idea was that each team would own its child components, and the Home team would build a layer of parent components that reacted to what came below. Our job, in my words, was enablement and keeping the structure healthy.

What I did:

- I brought the problem to the table, because I knew first-hand that the old zone had become unreadable.
- I collaborated with each domain to transfer knowledge, then took their questions and feedback back to the PStudio team so the solution could be iterated.
- I led many of the conversations, because I could speak business, engineering, data and design.
- I paired with one Home engineer and we built each node of the date-level messages, called interactions. That included knowledge transfer so the domain teams could maintain their own setup, and a lot of saying no to protect the data schema.
- I onboarded the largest teams first, and what we built for them helped the next ones. One team had not considered a state for users eligible but not enrolled, and another reused the pending state created for Balance.

What was not mine. Each domain team owned its ranking and scoring models and its business case. Of the results I was partly responsible, but the business result belongs to the products and not necessarily to Home. Monitoring guardrail metrics was not my job either, since the team had engineering and QA people for that.

## Options

```mermaid
flowchart TD
  H["Home team owns the parent"] --> P1["Parent: Balance"]
  H --> P2["Parent: Credit"]
  H --> P3["Parent: Pay Later and the other domains"]
  P1 --> C1["Child cards, owned by the Balance team"]
  P2 --> C2["Child cards, owned by the Credit team"]
  P3 --> C3["Child cards, owned by each domain team"]
```

The final design had three parts: a standard Dashboard zone, messages at the component layer, and messages at the data level. The component schema supported pending, zero and enrolled states, plus error and loading states. The team stitched the header and the Next Best Action together, and the instrumentation with them. To handle the slow data sources, the backend resolved them asynchronously, with the skeleton and the components loading in parallel.

Alternatives I rejected, and what each cost.

| Option | Why I rejected it | Cost accepted |
| --- | --- | --- |
| Two stitched cards (header plus action) with no parents | It would recreate the mess of the old zone. | More structure to maintain on the Home side. |
| Complex eligibility rules at the parent | Each rule meant calls to several platforms and hurt latency. Simple, fast rules stayed at the parent, complex ones moved to each domain's children, and a parent hides when it has no child to show. | Some teams were uncomfortable with it. |
| Machine-learning ranking in the first version | It would widen the scope in a short window. | It was postponed, not dropped. |
| Instrumentation requests from teams | Teams asked for 50 or more fields, and only 4 or 5 were shared. Most of the rest paraphrased the same thing under another name. | I refused all but the ones that proved they needed parent-level instrumentation. At child level teams could work as they wished. |

I also refused every request to use the EDS (external data source) call for parent eligibility, for latency reasons, with one exception. That decision has its own case, linked below. I refused many requests for ownership of the layers Home owned too. One of them, from Credit, asked for a separate Credit Card parent. I tried to deny it and could not, and that parent was created.

## Results

Eighteen months after launch, the zone accounts for about 80% of all Dashboard zone impressions. Its click-through rate is 27.14% for the last seven days, against 14.53% for the Financial Snapshot zone over the same seven days. The comparison is a snapshot of users who had one experience or the other. The Home screen has about 50 million daily impressions.

The structure outlived the project. It was still in use, with one more Credit product added. Teams maintained and expanded their own data-level assets without consulting Home, and the Home team experimented with a different design, better titles and better performance.

The business effect sits with the domain teams. Experiments on cards in the zone project about $300M in branded checkout payment volume and $21M in margin, from cashback card cart-math tests. These figures came from the zone's A/B tests, composed of each card's individual results, and the domain teams own them. They are projections, and I did not measure them.

What I did not measure: app crash rate, the number of calls against the ramp percentage and the Home build latencies were tracked by the team, not by me. I have no revenue figure that I can attribute to the zone itself.

## Friction and what I would change

Legacy is the biggest answer. PayPal does not force app updates, so a long tail of users stays on old versions and teams keep carrying duplicated configurations. Rules grow tangled, and the real behavior often differs from the documented one. Objects get created, teams change, the owner leaves, and nobody knows why something behaves the way it does. With the Crypto team I piloted about half a dozen total overhauls that never happened.

Dropping complex rules at the parent caused discomfort for the teams that wanted them.

I would do two things differently, and both are mine. I would write objective escalation criteria for architectural vetoes, so the structure is not split along the org chart in ways that later break ranking. And I would tie every platform migration to an explicit plan to retire the old configurations. I would also record the discussion sessions and keep the transcripts.

## Limits

- The 80% share comes from an internal chart that counted unique users. I call it impressions here, as I did when I first wrote this case up.
- The click-through comparison is a snapshot, not a controlled test of the redesign alone.
- Revenue and checkout figures are domain-team projections, not measurements, and they are not mine.
- Internal screenshots and the names of colleagues stay private.
