Identity document verification

Booking's Know your partner form rejected roughly half of all submissions with no explanation. This redesign uses OCR to catch document mismatches in-session - so partners can fix errors before they ever hit submit.

The TEam

2 designers, 1 product manager,
5 engineers, 4 marketers

Responsibilities

User experience, interface design,
simplyifying omnichannel journeys,
user research & testing,

TimelinE 

January 2020 - October 2020

Team

2 Product Managers, 1 Engineering Manager, 5 Engineers,
Vendor partner

My Role

Lead Designer

Responsibilities

UX strategy, Systems mapping, User Research, Interaction design, Cross-functional alignment

Year

2026

WHY

Context

At Booking.com, every accommodation partner must complete Know Your Partner (KYP) - a regulatory compliance process that includes verifying their identity document.

Partners submit ID documents with zero real-time feedback. Any typo, data mismatch or expired document leads to rejection - discovered days later, with no explanation of what went wrong.

Business goal: Improve pass rate by 15% by catching document errors before submission, instead of relying on manual analyst rejection after the fact check weeks later.

WHAT

Business problem

The major business problems we were tackling were:

  • Almost 40% of partner dcoument submissions were getting rejected
  • 31 % of the rejections came from data comparison failures - between what users were typing vs. the document they were sharing.

User problem

Our accomodation partners going through the regulatory KYP process were facing 2 major problems:

  • Partners were both submitting and retrying blindly:
    Partners couldn't see why a document failed and had to re-upload blindly only to fail days later
  • Partners hitting recoverable vs. unrecoverable errors:
    Some partners have simple typos, while others have expired documents (a hard block requiring a new document) - these needed fundamentally different resolution paths.

WHAT

My role

I drove end-to-end ownership from discovery to validation:

  • Strategic autonomy: Drove product vision and problem framing - design shaped the PRD, not the other way around.
  • Systems thinking: Mapped every failure outcome, mismatch combination, and recovery path into a deterministic system.
  • Cross-functional alignment: Aligned PM, engineering, vendor, compliance, analysts and research teams on phasing and dependencies.
  • Data-driven execution: Designed the research methodology, ran post-launch validation, fed findings into Phase 2.

The How

Driving the process to the solution

The How / Discover

1. Understanding the problem space

The problem looked like "partners upload bad documents."
But I reframed it: the system has no intelligence layer between upload and analyst review. Hence, partners have zero feedback loop.

"What if we could catch mismatches the moment a document is uploaded, and let partners fix them before they ever hit submit?"

The How / DEFINE

2. Mapping the failure landscape

Three validation outcomes
‍

Mismatch scenarios mapped:
By diving deeper into the data, I mapped every combination across data mismatch and rejection scenarios- including name, date of birth and document expiry- and identified how each scenario changes on retry, as well as which resolution paths block one another.

Experience principles as decision framework:
‍I defined 4 experience principles BEFORE designing screens. Used them to evaluate and kill weaker options:

The How / DESIGN

Design decisions

01

Editable extraction table

When OCR extracts wrong data, partners correct in-context."Use ID data" lets them accept ID data over their typo in one click.

02

Single-side capture

No front or back tab. System determines what's needed based on document type - reducing cognitive load and errors.

03

Fallback

If OCR fails more than 3 times, graceful degradation to manual entry with pre-filled extracted data. Never trap.

04

In-session quality coaching

Screenshots, blurry images, expired docs caught BEFORE submission. Not discovered by analyst days later.

THE HOW / VALIDATION

Validating with partners

Designed the research methodology: With the help of research team, we conducted 10 moderated sessions, 7 countries, controlled mismatch scenarios on both mobile and desktop.
‍
What worked well:

01

Two-column comparison immediately understood by all the participants

02

Use ID data button was universally clear and users found it intuitive

03

8/10 median ease of use rating on the new experience from the partners

THE HOW / DEVELOP

We moved feedback from days later to moment-of-upload

Partners can now fix errors before they submit - which means fewer rejections, fewer repeat cycles, faster time to verified.

CONCLUSION

Outcome

While project priorities shifted internally, the MVP got successfully launched in May 2026 driving a great first impression.

01

Business Impact - The redesign resulted in a huge success and almost 71% partners were able to self-correct the mismatches real-time in session -> leading to lower submissions for manual review

02

User Impact - Partners no longer wait days to learn a submission failed. Errors that used to require a full resubmission cycle and an analyst's manual rejection are now caught and resolved in the same session they occur in.

03

Strategic Impact - This wasn't just about catching document errors faster. It was about making sure a partner never has to wonder what went wrong - because the system tells them, in the moment, while there's still time to fix it.