Case study 01 · Webjet (ASX:WEB)

Unblocking a purchase-path dead-end worth half a million a month

Customers searching for an international flight landed on a screen where every returning flight appeared unavailable. This high-traffic flow needed a leader to deliver a derisked solution.

Roles I covered
UX researcher, interaction designer, prototyper, UI designer
Team
Me as UX lead, liaising with the product manager, product owner, UI dev team and Analytics
Duration
Picked up Mar 2016; four test rounds; measured post-launch 2017
Tools
SketchInVision UserTesting.comUI Pattern LibraryTableauGoogle Analytics
Outcome
+0.2% Transtasman Matrix, +0.4% Long-haul Matrix conversion – ≈ +$500k AUD revenue per month
01

A dead end that looked like a bug

Webjet has an unique user interface for selecting flights internally called 'the matrix'.

Depending on the underlying fare structure, some or all international return fares get marked with a small ‘R’ icon. Select one, and most prices in the returning matrix greyed out – because only a handful of returning flights were valid with that fare. Nothing explained this. Customers had to hunt the grid for the one matching price.

The legacy Webjet flight matrix. Paired fares are marked with a small R icon and, once selected, most returning prices grey out.
Before. The two questions people asked out loud in testing were “why does this have an ‘R’ next to it?” and ...
The legacy Webjet flight matrix. Paired fares are marked with a small R icon and, once selected, most returning prices grey out.
“why are all my returning flight options unavailable?”

The real problem

Customers concluded the site was broken or sold out, and left. This was a usability issue silently & continually impacting revenue.

02

What I inherited: three solutions, none built

The working method I’d implemented for product design projects at Webjet is called the Double Diamond. The method ensures a comprehensive understanding of the problem space is achieved before moving into the solution phase, and that ideation is separated from evaluation.

Double Diamond diagram: discover, define, develop, deliver, with divergent and convergent phases labelled.
The Double Diamond (British Design Council, 2005), used as the team’s working method.

For this project the product team had already discussed the issue at length and sketched proposals were sitting on the server from June 2014, eight months before I joined.

As the problem definition was clear (Discover & Define complete) and the team had already explored potential solutions (Develop), I decided to focus on validating the ideas (Deliver).

Initial ideation:

Hand sketch, solution A: duplicate the valid returning paired fares and shift them to the top-left of the returning matrix.
A. Duplicate available returning fares and lift them to the top of the results list (increase visibility).
Hand sketch, solution B: hide non-paired fares except a few at a similar price point or departure time.
B. Hide the unavailable fares, keeping only a few at a similar price or departure time (choice reduction).
Hand sketch, solution C: present the paired options in a modal with a fast route back to departing flight selection.
C. Put the available return fare options in a modal, with a fast route back if none of them suit.

The product team had mocked up Solution C in January 2015 but it stalled as it was a significant change to a revenue-critical flow, and nobody could say with confidence that customers would understand it.

The product team's January 2015 mockup of the paired flight modal, before the UX team existed.
The product team’s Jan 2015 mockup. Never implemented.

Decision: skip discovery, spend the time on validation

The problem was well understood and a solution direction already had internal agreement. Re-running discovery would have rendered no new information. What was missing was evidence that people could actually use it — so I moved the entire budget for this project downstream into prototype testing.

03

How I approached it

I picked the project up in March 2016, while working through a broader set of usability issues in the matrix results interface. First move was a mockup built to the UI Pattern Library standards that I’d established the year prior.

First UX team mockup of the paired flight modal, built to UI Pattern Library standards.
My first mockup, March 2016 – UI pattern-library conformant.
04

Prototype and task design

Prior to this project I secured an annual budget for remote usability studies. I hadn’t tested the Matrix, so my first scenario was a plain cheapest-flight task that would give us general feedback on this unique interface.

The second scenario focused on the new modal I’d created.

Tasks for usability studies
Remote usability study context, questions & tasks.

Writing unmoderated usability study plans is an art. You have to empower the participant with a relatable scenario, extract their mental model, and craft tasks that avoid leading the customer or encouraging unnatural behaviour. I used the opening task to capture their mental model — asking what factors would drive their choice before they saw any UI. Where the interface and the mental model diverge is where the usability problems often live.

The final task revealed what we’d been testing and invited them to share what they thought of it. That final 'reflection' prompt produced some of the most useful insights as it invited the participant to be a co-creator (and not slam the door once the study was complete).

Sketch file showing design artboards connected by hand-drawn flow lines.
Artboards in Sketch with my flow lines, covering every branch a participant could take.
InVision prototype with hotspots being placed over static artboard images to assign click actions.
Artboards exported and hotspotted in InVision so static images behave like a working site.

At the time, Invision App was the preferred tool for creating simple interactive prototypes. Now, Figma-integrated prototyping solutions are abundant.

Decision: test the worst case, not the happy path

My testing plan deliberately forced people into the modal and guaranteed it would not contain a flight matching their requirements. It was essential to test this failure path to ensure the solution would be effective in real-world use.

05

Four rounds, and the round that broke the deadlock

Each round: unmoderated remote studies through UserTesting.com, observations written up, changes made, retest.

1

The pattern-library mockup

What I tested

  • Modal on fare selection, ‘R’ icon retained, icon description and instructions present

What happened

  • People were confused about why they were seeing the modal, and what to do if they didn’t like the options — the same failure as the flow we were replacing
  • “Was the total $254, or 2 × $254?”
  • The ‘R’ icon meant nothing. Nobody read the description or the instructions

Read

  • Steve Krug’s insight from his book Don’t Make Me Think — “people don’t read, they scan” — was evidenced loud & clear.
Round one modal: R icon, small description text, dense instructions.
2

Big explainer strip

What I changed

  • Added a prominent strip across the top explaining the ‘R’ and what to do if the options didn’t suit
  • Prefixed the second price with ‘+’

What happened

  • The ‘+$254’ killed the total-versus-extra confusion outright
  • “Return fare” didn’t resonate — to customers every flight they were booking was a return fare. I needed a different word
  • Prominent or not, people still didn’t read it. Too wordy
Round two modal with a wide explanatory strip added above the flight list.
3

More direction (the wrong lever)

What I changed

  • Added more direction to the explanation and italicised the next steps

What happened

  • Still unread. Still confused about why the modal appeared
  • Too much to absorb at a glance — I had responded to “they don’t understand” by writing more words, rather than 'chunking' information delivery
Round three modal with expanded explanatory copy and italicised next steps.
4

New word, split message

What I changed

  • Renamed the concept: ‘R’ for Return Flight became ‘P’ for Paired Flight
  • Split the explainer into two separate parts — what this is, and what to do if you don’t want it

What happened

  • “Paired flight” landed. It was a new term, so people had no competing prior meaning to fight
  • Separating explanation from next-step meant each was short enough to be scanned rather than skipped
Round four modal using the P for Paired Flight label with explanation and next steps split into two parts.

The insight that unlocked it

Three rounds of failure were all the same failure: I was trying to explain an unfamiliar concept using a familiar phrase. “Return fare” already meant something to customers, so my explanation was competing with their existing definition and losing. Inventing “paired flight” gave the concept a label with no prior meaning attached — and suddenly the explanation had somewhere to land. The nomenclature outlived the project and became the standard term across Webjet’s flight products.

06

The design that shipped

Round four validated the concept. The remaining work was craft: tightening vertical space, and handling some messy real-world cases the mockups had overlooked.

  • Higher-contrast next-step instruction
  • Compressed vertically so more flight options sat above the fold
  • Flight details pop-over on hover, as per Matrix convention
  • Peeking for three or more options (show half a row plus the scrollbar so people know to keep going)
The final paired flight modal in production: P badge, short two-part explanation, departing flight summary, and a list of selectable returning flights with a peeking row.
Shipped. Short explainers, comprehensive flight information, and a peeking row signalling there is more upon scroll.
07

What it was worth

Event tracking went in with the interface, so we could read the conversion shift by matrix type (which changes depending on the searched flight route).

Conversion rate chart for Hybrid Matrix showing a step up after implementation.
Hybrid Matrix (mixed one-way and paired, so the modal is seen less often): +0.2% conversion — about +$250,000 TTV per month across international and trans-Tasman.
Conversion rate chart for International Matrix showing a larger step up after implementation.
International Matrix (every fare is paired, so the modal is unavoidable): +0.4% conversion — another +$250,000 TTV per month.

Outcome

≈ +$500,000 AUD in total transaction value every month, from resolving a single usability dead-end. The uplift was twice as large where the modal was unavoidable.

The research also travelled. The paired-flight insights and terminology fed straight into the mobile flights redesign that ran alongside it, which saved that project several rounds of relearning the same lesson on a smaller screen.

Mobile flight results applying the paired flight terminology to the departing leg.
Applied on mobile. Same language, same explanation pattern.
Mobile returning flight screen showing the paired flight explanation and a recovery link.
Including the recovery path — “select a different departing flight”.
08

Future enhancements

Once the initial implementation was complete, additional usability enhancements were identified, which were added to the product backlog:

  • Make opening and closing the modal instant, with no page scroll. A platform limitation at the time, that distracted customers and added to the cognitive load.
  • Turn dismissed paired fares purple — the web-standard visited-link colour (purple) is a strong visual cue that the option has been seen and rejected — so people don’t reopen options they’ve already rejected. Cheap, and it removes real cognitive load.
  • Combine matrix columns for same-flight paired fares. Two fares departing at the same time on the same carrier don’t need two columns; merging them reduces the horizontal sprawl that makes the matrix hard to scan in the first place.
Annotated note: paired flights departing at the same time with the same carrier should be grouped into a single modal.
The column-merging proposal, annotated for the next team.
09

What I learned

Four things I took from this

  • People don't read - they scan. Chunking information into smaller, scannable pieces improves comprehension and decision-making.
  • Don't be afraid to coin your own terms. The “Paired flight” term worked precisely because nobody arrived with a definition of it.
  • Leverage the familiar. Building using the internal UI Pattern Library accelerated protoype production & allowed me to focus on the user experience.
  • Measure your impact. Having a reliable baseline and the ability to segment based on what users experience helps justify future spend on Product research.

Get in touch

Happy to walk through the research files, not just the outcome.

Looking for a full-time, in-house role on a product team that wants to build with its customers.