Skip to main content
Jorge de Lima Jorge de Lima

Case study 01 · Booking.com for Business

Designing trust at the entry point

Redesigning the B4B registration flow, from a written value hypothesis to a shipped, measured result.

Booking.com for Business registration, step 1 of 2

Hypothesis ledger

Written before design · resolved after ship

Hypothesis

If we reduce time and error rate during onboarding, users will complete registration, build trust, and convert to active users.

Success criteria

  • −50% task completion time
  • −20% error rate

Status

Confirmed — both exceeded

Error rate down 53% against a 20% target; task time down 46% against a 50% target.

Context & challenge

Booking.com for Business was focused on user acquisition through trust. The hypothesis: friction in the onboarding flow was a direct blocker to conversion. A user who can't register themselves is a user who doesn't adopt the platform, and doesn't generate revenue. The strategic goal was growth. Onboarding was identified as the key lever.

Role & contribution

As Senior Product Designer I owned the end-to-end redesign of the B4B registration flow, from research to a shipped MVP, working alongside Anna and Jacqui (Product Managers), James (Product Owner), and Lucas and Chris on Engineering.

Product

Anna, Jacqui, James

Research

Scott, Keryn

Engineering

Lucas, Chris

Research & insights

Working with Scott (Lead Researcher) and Keryn (Senior Researcher), we ran 20+ moderated usability sessions to find out not just that users were failing, but why. Four patterns kept surfacing.

  1. 01Error messages told users something was wrong without telling them how to fix it, and sat in a position on the screen users weren't reading.
  2. 02Users arriving from the leisure site kept trying to sign in with the same email they used there, and the flow didn't account for it.
  3. 03Users lost track of where they were in the process, with no visible sense of what came next.
  4. 04The handoff between the legacy system and the new one broke visual consistency and slowed load time, undermining trust before the flow had even started.

Rather than jumping straight to a redesign, we fixed what we could inside the legacy system first, rewriting error messages in clearer language, and tested those fixes to confirm the hypotheses before committing to the larger rebuild.

Legacy registration form before and after the error handling fix
Legacy error handling — the alert banner detached from the field it referred to, against the inline, actionable state we tested in its place.

That sequencing is also where the aggressive descoping and the parallel MVP-against-control structure came from: we removed features that wouldn't deliver immediate onboarding value, and launched an MVP in parallel with the live control so outcomes could be attributed directly to the new design, not external noise. We also built AI-powered prototypes to run usability tests faster and at higher fidelity than traditional wireframes.

Design strategy & approach

We restructured the experience around progressive disclosure, clear state feedback, visible progress markers, and language simplification. Rather than compressing the flow, we optimised cognitive load and clarity per step.

Legacy self-registration form
BeforeLegacy registration step
Redesigned registration step, 1 of 2
AfterRedesigned registration step

Solution

  • Redesigned step logic with clearer grouping of information
  • Improved validation states with actionable error messaging
  • Introduced clearer progress tracking
  • Reduced redundant input friction
  • Standardised UI patterns to align with the platform design system

The experience shifted from transactional to guided.

Completed registration flow: details, goals, and account creation
Final UICompleted registration flow — details, goals, account creation

Results & impact

Error rate

minus 53%

18.5% → 8.7% · target −20%

Task completion time

minus 46%

40+ sec → 21.5 sec · target −50%

Both success criteria were met or exceeded. The error rate target was a 20% reduction; we delivered 53%. The time target was a 50% reduction; we delivered 46%.

Reflection & learnings

This project reinforced that onboarding is not a form problem, it's a confidence architecture problem. It also informed internal standards around error messaging and progressive disclosure patterns now reused in other flows. Future iterations could incorporate adaptive onboarding based on company size and role segmentation.

Next case study

Tips to getting started

Booking.com for Business →