PayPal Developer Portal & AI Authorization
From 45k to 135k monthly developers — and +$1M merchant revenue.
Context
PayPal's developer audience spans solo integrators to enterprise platforms. The legacy portal optimised for documentation breadth, not first-integration success — and merchants in the US and EU were losing revenue to declined authorizations they couldn't diagnose or fix.
Problem
PayPal's developer portal was the front door for integrators, but discovery, onboarding and tool adoption were fragmented. Developers churned during integration; merchants lost revenue to declines they couldn't explain.
Approach
- 01
Ran 30+ developer interviews, a competitive teardown of Adyen, Square and Stripe, and a diary study with SMB merchants.
- 02
Mapped a single SMB × Developer journey to find the 7 moments that earn or lose trust.
- 03
Redesigned the portal IA and shipped an in-product walkthrough; rebuilt credentials, webhooks and error surfaces.
- 04
Designed an AI authorization-rate tool that explains every decline in plain language and proposes the fix.
Leadership & scope
Senior IC leading design across the Developer Portal redesign and the AI Authorization workstream. Embedded with Applied ML and the merchant-facing payments org; reported into the Director of Developer Experience.
Direct: 2 designers + 1 content designer. Cross-functional: ~18 (eng, PM, applied ML, developer relations, localization).
Director of Developer Experience, VP Payments, Head of Applied ML, regional merchant leads (US + EU), legal & compliance for AI explainability.
Coached 2 mid-level designers through the redesign; built a journey-mapping playbook adopted by the wider DevX org; ran weekly portfolio crits open to the broader design team.
- 01
Rejected a 'docs-first' homepage proposal — research showed developers wanted recent transactions and error logs at the door, not marketing copy. Defended this against marketing leadership.
- 02
Pushed Applied ML to expose model confidence and a plain-language reason on every declined transaction, instead of a generic 'try again' fallback. That decision is what unlocked the $1M revenue lift.
- 03
Forced a single design system across sandbox and live modes — eliminated the 'two dashboards' problem that had been a top-5 developer complaint for two years.
AI-specific design decisions
Explain every decline in plain language
Merchants saw cryptic decline codes (e.g. 'DO_NOT_HONOR') with no path to remediation. Authorization rate was bleeding revenue silently.
Every decline surfaces a plain-language explanation, the model's confidence in that explanation, and a ranked list of fixes (retry timing, alt-funding, AVS tweak). Merchants can click any fix to simulate the rerun.
Required slowing down some checkout flows by 80–200ms to fetch the explanation; merchants overwhelmingly preferred the latency to the lost revenue.
Confidence display without false precision
Showing '94.2% confident' implies a precision the model doesn't have, and invites merchants to argue against rounding.
Three-band confidence (High / Medium / Low) with a tooltip explaining what each band means and the historical accuracy of that band on the merchant's own traffic.
Loses fine-grained ranking inside a band; gained dramatically in trust and reduced 'why did your AI say…' support tickets.
Human-in-the-loop on auto-retry rules
Auto-retrying declined transactions can violate network rules and trigger fines. Letting the AI do it autonomously was a non-starter for legal.
AI proposes retry strategies; merchant approves a policy (per BIN, per region, per amount band) once, then the AI executes within those bounds and logs every decision for audit.
Slower initial activation (merchants must read and accept policy), but turned a legal-blocked feature into a shippable one.
Evaluation loop merchants can see
Merchants didn't trust an opaque AI deciding their revenue. Trust was the whole adoption barrier.
Built a weekly 'AI scorecard' surface: predictions made, accepted, overridden, and the resulting auth-rate delta vs. the no-AI control group on the same merchant's traffic.
Required running a permanent holdout cohort (slight revenue cost) — paid for itself by removing the 'is this even working?' objection in every renewal call.
Showreel
Showreel — end-to-end walkthrough of the new developer portal & AI authorization surface.
Research
Research
Before a single pixel moved, we ran 30+ developer interviews, a competitive teardown of Adyen, Square and Stripe, and a diary study with SMB merchants. The data made the brief: developers trusted PayPal the brand but distrusted the portal — and that gap was costing merchants real revenue.
Where the old portal lost developers
- Outdated design system65%
- Confusing information architecture58%
- Slow page performance42%
- Accessibility gaps30%
Quantified from 30+ developer interviews and a heuristic teardown. The top two are design-fixable — the bottom two are tickets we filed with platform.
Where PayPal still won
- Brand trust90%
- Active developer forums54%
- Recognized documentation as a weak spot47%
- Willing to recommend to peers38%
Trust was the moat. Documentation was the leak — and the redesign concentrated on closing it without breaking the brand promise developers already believed in.
Competitive
Who set the bar — and where
| Competitor | Strengths | Weakness |
|---|---|---|
| Stripe | DocsCode snippetsTime-to-first-call | Pricing complexity at scale |
| Square | SubscriptionsPOS integrationOnboarding | Limited global coverage |
| Adyen | Global coverageAuth ratesEnterprise tooling | Integration ease for SMBs |
| PayPal | Brand trustMerchant basePayouts breadth | Dashboard & developer surface |
Stripe set the bar on documentation. Square owned subscriptions. Adyen led on global coverage. PayPal had to out-design all three on the dashboard surface.
Themes
Two themes ran through every interview
API quality & velocity
Developers measured us in time-to-first-API-call, error rates and how fast we shipped fixes. Anything that slowed them down — including the portal itself — was a tax on our reputation.
Developers as decision-makers
Developers weren't just implementers — they were the ones recommending payments providers to SMB founders. Earning their trust was earning the merchant.
These two themes became the lens for every prioritization call: if a feature didn't measurably move Performance or Trust, it didn't ship.
Goals
Seven measurable goals
Each goal mapped to a single owner and a quarterly checkpoint. The 3× traffic and +$1M revenue lift became the project's North Star metrics.
Journey
Two journeys. One map.
- · Compare providers
- · Evaluate docs
- · Match developer
- · Create app
- · Get keys
- · Scope project
- · Sandbox tests
- · First call
- · Webhook setup
- · Go live
- · Monitor
- · Diagnose
End-to-end journey across SMB actions, developer–SMB interactions and developer actions. The seven numbered moments are where the redesign concentrated its budget.
Opportunities
Four How-Might-We bets
How might we match SMBs with the right developer at the right moment?
→ Developer–SMB matching surface inside the portal.
How might we make integration complexity visible up front?
→ Effort-and-scope estimator on every product page.
How might we let merchants explore checkout visually before they commit?
→ Interactive checkout sandbox with live previews.
How might we give developers a scoping tool their PMs trust?
→ Shareable project scope with auto-generated estimates.
Each HMW pinned to a specific pain point — each one became a feature workstream that landed inside the year.
Before & after
Before & after
Landing → Dashboard. The old apps & credentials page asked developers to read marketing copy before getting to keys. The redesign opens with a personalised welcome, recent transactions, error logs and monitoring — the four things developers actually return for.
Webhooks events, before & after. The legacy table buried failures inside dense sidebar navigation. The new design surfaces status, response time and a resend action in one row — 37% more developers used the tool within a quarter.
Process
How we built it
Developer journey mapping
Mapped merchant authorization workflows end-to-end, identified friction points in onboarding, and built empathy maps and personas to align product, research and engineering on the highest-leverage developer moments.
Portal redesign + guided walkthrough
Streamlined the developer portal IA and built an in-product walkthrough so developers could learn and try new features without leaving context. Shipped in 4 months — traffic grew from 45,000 to 135,000 monthly visits.
AI authorization tool
Designed an AI-driven tool that surfaces why a transaction was declined and how to fix it — improving merchant authorization rates and lifting revenue by $1M for US and EU merchants. Led the AI chat app that cut payment processing time by 14% and added $230k in revenue.
Tooling adoption
Iterated on developer tools with usability testing and analytics — usage rose 30% in a single quarter while developer-side errors dropped 62%.
Where the portal was leaking trust
Developers measured the brand by the dashboard, not the marketing site. Buried errors, dense webhooks and unclear sandbox-to-live transitions were costing both adoption and merchant revenue.
What we could not break
Brand trust and an active developer community were the foundations. The redesign had to modernise the surface without disrupting the credibility developers already gave PayPal.
Storyboard · 10 frames
Ten frames
How the redesign moved the work, scene by scene.
Mapping two journeys at once
SMB merchants and the developers who serve them never lived on the same map. We synthesised 30+ interviews into one shared journey — and found the seven moments where the portal could earn trust or lose it.
Where developers used to land
A dense marketing page with no clear path to keys, sandbox or working code. Time-to-first-API-call averaged days, not minutes.
A welcome that does the work
Solution-led entry: pick what you're building, get the right SDK, sandbox and guide in one screen. Traffic 3×'d in a quarter.
Credentials, demystified
One-screen sandbox: live keys, return URLs and 12 feature toggles with plain-language explanations beside each switch.
From sandbox to production
A clear, colour-shifted live mode with the same mental model — so developers never relearn the dashboard when revenue starts flowing.
Errors, surfaced not buried
Error log moved from a 4-click hunt to the home dashboard. Time-to-diagnose fell sharply.
AI explains the 500
Click any failure and get the full request, the response and an AI-written explanation of what to fix — lifting auth rates by enough to add $1M in merchant revenue.
API health at a glance
Twelve months of request volume and error rate, filterable per app. The chart is the headline; the table is the proof.
Reward the calm
When nothing is on fire, the dashboard quietly cross-sells the next product — payouts, subscriptions, online checkout.
One responsive system
Every screen above is built from the same tokens, grid and components — shipped across XXL → S in one library.
Gallery
Selected screens
End-to-end developer experience
Journey map through dashboard, sandbox, live, observability and the AI diagnostic.
Outcomes
What shipped, and what changed.
Monthly developers nearly tripled — 45k → 135k in the first four months.
Estimated +$1M merchant revenue lift attributed to the AI authorization tool.
Developer tool usage up 30% in a single quarter, sustained across regions.
Developer integration errors down 62%; AA WCAG compliance across portal surfaces.