The strategy I’d approach
Your first research project
already has a prime suspect.
The role asks for research into post-booking customer support in Flights, with the goal of reducing contact rates. Baymard scores Booking.com’s Accounts & Self-Service at −39.9 in the app. People contact support because the surface built to prevent that contact does not work. Here is the evidence, and how I would test it properly.
Four root causes, ranked
Everything I could find in the public record about Booking.com’s product experience reduces to four causes. The first one is, I think, the answer to the question this role is being hired to ask.
If you read one thing, read the orange boxes. That is the failure as a traveller actually meets it.
Ordered by relevance to the Flights post-booking brief
-
01
Self-service does not work, so people contact support
What happensAccounts & Self-Service is where a traveller changes a date, cancels, or chases a refund without speaking to anyone. Baymard scores it 54.0 on desktop, 13.7 on mobile web and −39.9 in the app. That last number is below the bottom of the scale, and it is the lowest score anywhere in the 649-guideline review.
What the user actually experiencesEVIDENCE App self-service −39.9 · mobile web 13.7 · desktop 54.0 · refunds and modifications are the top two themes in Flights reviews FIX → Treat self-service as the contact-deflection product it already is, and instrument it against contact rate rather than task completion.Your flight time moves. You open the app to change the booking, because that is where you booked it. There is no path that works, so you start a chat. The chat is a bot. You are now a contact in the contact-rate metric, and you did not want to be one either.
-
02
A three-way handoff that nobody owns
What happensA flight bought on Booking.com may be ticketed by a partner agent and operated by a carrier. Three parties, one traveller, and no declared owner of the case. Reviewers describe being passed between all three, with escalation requests refused and refunds running past ten weeks.
What the user actually experiencesEVIDENCE Booking.com Flights 1.5 / 5 across 371 reviews · refunds reported at 10+ weeks · “flexible” changes incurring fees that negate the flexibility bought FIX → One case, one owner, one thread, with ownership assigned by the platform rather than discovered by the traveller.“They constantly tried to send me to the airline, and both companies won’t contact the other.” Nobody says no. Nobody says yes. You are the only party with an interest in the case, and you have the least power in it.
-
03
The decision page collapses on the devices people decide on
What happensProperty & Room Detail Pages score 43.4 on desktop, 15.6 on mobile web and 10.5 in the app, across 151 guidelines. It is the page where the choice is actually made, and it is weakest on the two surfaces where most people make it.
What the user actually experiencesEVIDENCE Detail pages 43.4 desktop / 15.6 mobile web / 10.5 app · 151 guidelines · “misrepresented listings” a recurring Trustpilot theme FIX → Rebuild the detail page mobile-first against the guideline set. Under-informed decisions are a leading indicator of post-booking contacts.The page you use to decide is the page that tells you least, on the phone you are booking from. So you decide on less, and find out the rest after you have paid.
-
04
Persuasion that regulators have now put a price on
What happensUrgency messaging, drip pricing and account-gated rates have moved from growth tactic to legal exposure. Spain’s CNMC fined Booking.com €413 million in 2024; Hungary fined it around €7 million in 2020 over countdowns and “last room” messages; a Dutch class action followed in 2025.
What the user actually experiencesEVIDENCE CNMC €413m (2024) · Hungary ~€7m (2020) · Dutch class action (2025) · documented €75 of fees on a €211 base FIX → Research the honest version and measure it properly. If transparent pricing converts within tolerance, the argument is won with data rather than ethics.“Only 1 room left” made you hurry. Seventy-five euros of fees on a two-hundred-and-eleven-euro base made you feel tricked. Both are design decisions, and the second one is what people write the review about.
Protect these, and keep strengthening them
Four things are genuinely strong and should be built around, not traded against: homepage, navigation and taxonomy (84.4 desktop and 89.4 mobile web, among the best scores in the review), app site-wide design and interaction at 89.8, the app itself (4.8 across 2.1 million App Store ratings), and the sheer breadth of inventory that makes the Connected Trip plausible in the first place. Booking.com is excellent at getting people to a decision. The gap is everything after it.
User issues
What the public record says about the Booking.com product experience, with Flights separated out where the evidence allows, and what each finding translates into as an issue, a perception and a fix.
Three scoreboards, three different answers
The spread here is the widest I have seen on any product I have looked at, and the gap itself is the diagnostic: each measure captures a different moment in the relationship.
A 3.3-star gap between the app store and Trustpilot is not noise. It is a product that is excellent right up until the moment something changes, and then has almost nothing to offer. That transition point is exactly what this role has been asked to research.
The complaints
Eighteen, bucketed by severity and collapsed by default. Open a bucket, or filter to one.
Critical 7Drives contacts, refunds and regulatory exposure
−39.9 across 13 guidelines. The surface that exists to let people manage a booking without contacting anyone is the worst-performing thing in the review.
Reviewers describe waits of ten weeks and longer, some over a year, with the platform deferring to the airline and the airline deferring back.
Booking.com, the ticketing agent and the carrier each route the traveller to the next. Requests to escalate to a manager are reported as refused.
The single most common theme across 125,582 reviews: waits of forty minutes and more, agents who do not follow up, and AI systems in place of a human.
Travellers who paid for flexibility report change fees that cancel out the value of the option they bought. A product-definition problem before it is a service one.
Documented in one case as €75 of fees added to a €211 base. Now a regulated matter rather than a conversion tactic.
“Only 1 room left” and countdown messaging drew a fine in Hungary in 2020 and formed part of the CNMC case in Spain in 2024.
Serious 6Erodes trust and generates avoidable contacts
Across 151 guidelines on the two mobile surfaces. The decision page, weakest where the decision is made.
Rooms without windows, missing amenities, conditions well below what was advertised. Every mismatch becomes a support contact.
When a carrier cancels, reviewers report Booking.com declining responsibility and redirecting them to the carrier, who redirects them back.
40.4 desktop, 40.0 mobile web, 39.7 app, across 136 guidelines. Consistently middling on the highest-traffic task in the product.
Against 89.8 in the app. The app has solved something the desktop site has not, inside the same company.
App reviewers report completed stays with no Genius credit appearing, and no practical route to query it that is cheaper than calling.
Neutral 5Real, lower impact: watch rather than act
App reviewers describe rates well above what they remember paying. A market condition as much as a product one, but it shapes how every other failure is read.
Rates visible only after sign-in. Defensible as a loyalty mechanic, cited by regulators as a transparency problem.
The best-scoring transactional area in the review, and still only just over half. Solid rather than good.
The self-service problem is not universal. Desktop is passable, which makes the app score harder to explain and more worth researching.
Trustpilot notes that negative reviews go unanswered. A missed listening channel more than a service failure.
What it translates to
Seven themes. The perception column is what produces a 73% one-star distribution, and the fix column is where research earns its place.
| Theme | The issue | The perception it creates | What to fix |
|---|---|---|---|
| Self-service void | App accounts at −39.9, mobile web at 13.7 | “There is no way to fix this myself” | Rebuild self-service around the top contact drivers, measured on contacts avoided |
| Ownership vacuum | Three parties, no declared case owner | “Everyone is responsible so no one is” | Platform-assigned ownership, visible to the traveller from the first message |
| Decision starvation | Detail pages at 15.6 and 10.5 on mobile | “It wasn’t what the page said it was” | Mobile-first detail pages; under-informed decisions become later contacts |
| Flexibility mismatch | Paid flexibility undermined by change fees | “I paid extra for nothing” | Define the promise, then price it, then say so plainly |
| Fee surprise | Drip pricing, account-gated rates | “They hid the real price until the end” | All-in price early; test the conversion cost properly rather than assuming it |
| Manufactured urgency | Scarcity and countdown messaging | “I was rushed into it” | Replace with true scarcity signals; measure regret and refund rate, not just conversion |
| Support as a wall | Chatbot first, escalation refused, no follow-up | “There is nobody in there” | One case, one owner, one thread, with a stated escalation path |
How that compares
The honest finding first: the entire online travel agency category is rated appallingly, and Booking.com is the least bad of the majors. Bad post-booking service is a category norm rather than a Booking.com signature, which means it is available as a differentiator to whoever fixes it first.
Trustpilot rating, online travel compared
| Platform | Rating | Signature complaints | Relevant to Flights |
|---|---|---|---|
| Booking.com | 1.5 / 5 | Support unreachable; refunds delayed; listings misrepresented; drip pricing; urgency messaging | — |
| Airbnb | 1.3 / 5 | Host cancellations; refunds delayed; disputes decided against the reviewer; account suspensions | Same failure shape: a marketplace that will not arbitrate between two parties it introduced. |
| Expedia | 1.2 / 5 | Booking errors; hold times; refunds taking weeks; hidden charges | The direct OTA comparator, and the only one that answers a third of its negative reviews. |
| Airlines direct | varies | Own set of failures, but one accountable party | The real alternative for a frustrated flyer: booking direct removes the handoff entirely. That is what Flights is competing against. |
The read: Flights is not really competing with Expedia on price. It is competing with booking direct on certainty, and every unresolved handoff hands a customer back to the airline’s own website next time.
Design strategies
The same evidence read through four strategic lenses, because “reduce contact rate” is a target, not a strategy, and the difference between the two is exactly where a researcher earns their keep.
Business model innovation
Booking.com monetises the moment of commitment and, for most of its history, has not had to own what happens afterwards. Flights strains that model harder than accommodation does, because a flight changes far more often than a hotel room, and the change is where the whole cost sits.
Value creation
Choice and comparison at a scale no single supplier can offer, plus the Connected Trip thesis: one place for the flight, the room, the car and the rest. Flights is the pillar that makes that thesis credible.
Value capture
Commission at the point of booking, plus ancillaries and loyalty. Revenue lands at commitment; cost lands at change. The incentive gradient runs the wrong way for post-booking service.
The structural flaw
The party that earns the trust is Booking.com; the party that spends it is often a ticketing agent or a carrier. In accommodation that seam is narrow. In Flights it is the whole product, which is why root cause 02 is a Flights problem before it is a platform one.
The model-level move
Treat post-booking as part of the product rather than as the cost of it. A traveller who can change a flight themselves in ninety seconds does not contact anyone, does not write a one-star review, and books the next flight in the same place. Contact rate is not only a cost line, it is the best available proxy for whether the Connected Trip promise survives contact with reality.
The ten types of innovation
Doblin’s framework argues that winning innovations combine three to five reinforcing types, and that most organisations over-invest in product performance alone. Mapped to Booking.com Flights, the neglected types cluster tightly.
| Cluster | Type | Booking.com today | Opportunity |
|---|---|---|---|
| Configuration | Network | Strong: carriers, agents, properties, at global scale | Make the network legible to the traveller instead of invisible |
| Configuration | Process | Handoffs with no declared owner | Assign case ownership by design, not by escalation |
| Configuration | Profit model | Commission at commitment | Price flexibility as a real product with a real promise behind it |
| Offering | Product performance | Uneven: navigation 89, self-service −39.9 | Level the floor before raising the ceiling |
| Offering | Product system | Connected Trip in progress | The connection is only real if a change to one leg is handled across all of them |
| Experience | Service | Weakest area: the most common complaint in 125,582 reviews | The single biggest unclaimed differentiator in online travel |
| Experience | Channel | App excellent for booking, worst for managing | The same app has to hold both halves of the trip |
| Experience | Brand | Trusted to find, distrusted to fix | Certainty as the position, evidenced rather than asserted |
| Experience | Customer engagement | Persuasion under regulatory challenge | Rebuild the persuasion layer on true signals and prove the cost is survivable |
| Configuration | Structure | Not externally visible | Out of scope for public research |
The read: every Experience type is underperforming or failing, and that cluster is where every complaint in path 01 lives. A combination play across Service + Process + Brand is three reinforcing types, which is exactly what the framework says wins, and it is reachable from a single research programme.
Integrative thinking
Roger Martin describes this as facing the tension between opposing ideas constructively and, rather than choosing one, producing a resolution that contains both and is superior to either. Research lives on that skill: the job is usually to refuse the trade-off the room has already accepted. Three worth resolving here.
Tension 01
Conversion vs. trust
Model A: urgency and drip pricing demonstrably convert. Model B: they now carry nine-figure regulatory cost and a 73% one-star distribution.
Resolution: stop treating it as a values argument and make it an evidence one. Run the honest version as a proper experiment and measure it on lifetime value, refund rate and contact rate, not on session conversion. If the honest version holds, the debate is over.
Tension 02
Contact rate vs. contactability
Model A: fewer contacts is the goal, so make contacting harder. Model B: people contact because something is genuinely wrong, and blocking them turns a problem into a review.
Resolution: separate the metric into avoidable and unavoidable contacts. Drive the first to zero by fixing self-service; make the second faster and easier to reach. One number hides two opposite problems.
Tension 03
Research depth vs. delivery speed
Model A: product teams move weekly and cannot wait for a study. Model B: fast research produces confident answers to the wrong question.
Resolution: hold a standing evidence base so most questions are already answered, and reserve new fieldwork for what is genuinely unknown. Speed comes from having looked before you were asked, not from cutting the study short.
Design thinking, and where the squiggle actually goes
The squiggle running through this page is the argument: uncertainty is front-loaded, and clarity is earned by narrowing. The temptation here is to skip the divergent half, because the contact-rate target is already set and Baymard already names the weak surfaces. For two of the four causes that is correct. For the other two it would answer the wrong question quickly.
Where to converge immediately
Self-service and the detail page. These are solved problems with 649 documented guidelines pointing at the answers, and the scores are unambiguous. No ideation required: this is execution against a known standard, and treating it as an exploration would waste a quarter.
Where to stay divergent
Why people contact, and what honest persuasion costs. Neither is answerable from a benchmark. The first needs contact-driver analysis joined to behavioural data and interviews; the second needs a properly powered experiment. These are the two places where new research changes the decision rather than confirming it.
The gaps worth attacking
Every OTA draws the same generic complaints. The strategic question is which are worse at Booking.com, or uniquely available to it. Five hold up.
- 01
An 89 and a −39.9 in the same app
Site-wide interaction scores 89.8 in the app while self-service scores −39.9. The capability is demonstrably in the building. This is about where attention has been pointed, which is a far cheaper problem to solve than a capability gap.
- 02
The category’s worst-served moment is unclaimed
Every major OTA fails at post-booking. Nobody has taken it as a position. The first one to make change and cancellation genuinely self-serve takes the airline’s strongest argument away from it.
- 03
Flights changes more than hotels do
Schedule changes, cancellations and disruptions are normal in aviation and rare in accommodation. A post-booking model designed for hotels will keep failing in Flights, and Flights is where the Connected Trip is won or lost.
- 04
Regulators have priced the persuasion layer
€413m in Spain, a Dutch class action, an EU gatekeeper designation. The commercial question is no longer whether to change it, but what the honest version costs, and that is a measurable research question.
- 05
The contact-rate brief is unusually well posed
Most research briefs arrive vague. This one names an outcome, a domain and a stage of the journey. That is rare, and it means the first study can start on week one rather than after a scoping exercise.
Interface
Usability and accessibility as measured rather than felt: Baymard’s 649-guideline review across three surfaces, and the usability updates it produces.
Baymard UX performance by surface
Where the points are lost
| Assessment area | Desktop | Mobile web | App | Read |
|---|---|---|---|---|
| Homepage, Navigation & Taxonomy | 84.4 [11] | 89.4 [10] | 69.1 [10] | The strongest area in the review, and unusual in the industry. Getting people to a decision is genuinely well solved. |
| Site-Wide Design & Interaction | 38.1 [13] | 48.7 [27] | 89.8 [23] | A 51-point spread across surfaces for the same discipline. The app has solved something the desktop site has not. |
| Checkout Process | 47.8 [79] | 55.3 [71] | 56.9 [69] | 219 guidelines, the largest pool in the review, and the most consistent. Solid rather than good. |
| Travel Search | 40.4 [48] | 40.0 [47] | 39.7 [41] | 136 guidelines sitting at exactly middling on every surface. The highest-traffic task in the product. |
| Property & Room Detail Pages | 43.4 [52] | 15.6 [50] | 10.5 [49] | The decision page, collapsing on the two surfaces where most decisions happen. |
| Accounts & Self-Service | 54.0 [21] | 13.7 [15] | −39.9 [13] | The lowest score in the review, below the bottom of the scale, on the surface most people manage bookings from. This is the contact-rate problem, measured. |
Accessibility
This is the area I can speak to with the most direct experience, and it is also where I have to be most careful about what I can and cannot claim from outside.
What I could and could not check
Booking.com publishes an accessibility statement, which is good practice and more than many travel platforms manage. I could not retrieve it or run my own automated scan from this environment, so I am not going to characterise its contents or claim findings I did not verify. What follows is what the usability evidence implies, flagged as inference rather than audit.
Where the risk concentrates
The European Accessibility Act has applied since June 2025, and travel booking services sit squarely inside its scope. Under EN 301 549 the expectation is WCAG 2.2 Level AA across the booking journey, including the app. The areas scoring worst here, self-service, forms and detail pages, are the same components that carry the heaviest accessibility obligations.
What I would do in week one
Baseline properly: an automated pass across the booking and manage-booking journeys, a manual WCAG 2.2 AA audit of the highest-traffic flows, and assistive-technology testing with real users on the change-and-cancel path. Then prioritise by transaction impact rather than issue count, and land fixes in design-system components so each one ships once.
Why this belongs in a research role, not beside it
Accessibility failures and contact drivers are usually the same failures. A form field without a proper label, an error message that does not announce itself, a change flow that cannot be completed without sighted mouse use: each produces an abandoned task, and an abandoned task in post-booking becomes a support contact. Studying them together is cheaper than studying them twice, and it lets one programme of work answer to both a legal obligation and a commercial target. That is the argument I have made to product teams for three years, and it is the one I would make here.
Usability updates, in the order the data supports
- 01
Rebuild app self-service around the top contact drivers
−39.9 across 13 guidelines. Change a date, cancel, chase a refund, get a receipt, correct a name: each one completable in the app, with state visible without contacting anyone, and with accessible labelling and error handling designed in rather than retrofitted.
Directly addresses the role’s stated first project, and closes the lowest score in the review. - 02
Give the traveller a named owner for every case
One case, one thread, one accountable party, decided by the platform. Status visible without asking. The three-way handoff is not a service failure, it is an information-design failure with a service consequence.
Closes root cause 02 and the loudest theme in Flights reviews. - 03
Rebuild the detail page mobile-first
15.6 mobile web, 10.5 app, across 151 guidelines. Structured attributes, honest imagery, complete fare and fee conditions, and change and cancellation terms stated before purchase rather than discovered after it.
Under-informed decisions are a leading indicator of post-booking contacts. - 04
Port the app’s site-wide patterns to the web
89.8 in the app against 38.1 on desktop for the same discipline. The better implementation already exists inside the company, on a different surface.
Highest points-per-effort in the list: the work is done, on the wrong platform. - 05
Bring travel search off the 40-point plateau
40.4, 40.0 and 39.7 across 136 guidelines. Remarkably consistent, which suggests a shared root cause rather than three separate ones, and makes it a good candidate for a single diagnostic study.
- 06
Baseline accessibility to WCAG 2.2 AA across the manage-booking journey
Automated pass, manual audit of the highest-traffic flows, assistive-technology testing on change and cancel. Prioritised by transaction impact and fixed in components.
Overlaps almost entirely with fixes 01 and 03, so it costs far less run together than run separately.
Two independent instruments, a 649-guideline benchmark and 125,582 public reviews, point at the same surface: the place where a traveller goes to change something. That agreement is the strongest signal in this document, and it is the surface this role has been asked to research.
Research
Everything above is desk research, and desk research is a hypothesis generator, not an answer. This is how I would actually run the first project: post-booking customer support in Flights, with the goal of reducing contact rates.
The framing matters more than the method. “Reduce contact rate” is a target that can be hit by making contact harder, which would raise every other number that matters. So the first job is to split the metric into contacts that should never have happened and contacts that should be easy, and then never confuse them again.
Find out what people are actually contacting about
Secondary and behavioural first. It is cheaper, faster, and it tells you what to ask in the field.
- Contact-driver analysis: code support transcripts and tickets into a taxonomy of reasons, sized by volume and cost
- Join that to behavioural data: where in the manage-booking flow the contact was preceded by an abandoned self-service attempt
- Heuristic evaluation of the change, cancel and refund paths on all three surfaces, against the guideline set already in play
- Literature and competitor review: what the airlines and the other OTAs do at the same moment
- Stakeholder interviews with CS leads, Product, and the partner-facing teams, to surface constraints before proposing anything
Go and watch it happen
Primary research, aimed at the specific drivers the framing phase ranked highest.
- Depth interviews with travellers who contacted support in the last 30 days, sampled by driver rather than by convenience
- Usability testing of the current change and cancel flows, including with assistive technology users
- A survey powered to size the drivers across markets, so the qualitative story carries a number with it
- Diary study across a disrupted trip, because the moment of change is not reproducible in a lab
Make it change something
A finding that does not alter a decision was not worth gathering. Landing is part of the method, not the afterthought.
- Findings paired with design directions, so the recommendation is actionable in a backlog rather than admirable in a deck
- A pre-registered success measure agreed before the work ships: avoidable contacts, not total contacts
- A standing evidence base so the next question is half-answered before it is asked
- Follow-up sessions with the teams implementing it, because the report is not the deliverable, the change is
The method mix, and why each one is there
No single method answers this question. The contact-rate brief needs volume and reason, behaviour and motivation, and the honest answer sits in the overlap. Each method below covers a specific weakness in the others.
Contact and journey data joined together. Tells you how often and where, at full scale.
Blind to why. Will happily explain a behaviour that never happened.Travellers who contacted support recently, sampled by driver. Tells you what they were trying to do and what they believed.
Small n, and memory is reconstructive. Needs the numbers beside it.The change and cancel flows, observed. Tells you exactly where the task breaks, including with assistive technology.
Artificial stakes. Nobody in a lab has actually missed a flight.Powered to size the drivers across markets, so the qualitative story arrives with a number attached.
Only as good as the framing that preceded it. Never run first.Heuristic evaluation, literature, competitor and desk research. Cheapest evidence available, and the fastest route to a good question.
Generates hypotheses, not answers. Mistaking one for the other is how this page could mislead you.Contact drivers: the hypotheses I would test first
Seven hypotheses, drawn from the public evidence above, each with the signal that would confirm it, the method that would test it, and what would have to move for the fix to count. This is the table I would want to be wrong about in week four.
| Hypothesised driver | What you would see | How I would test it | What has to move |
|---|---|---|---|
| Self-service dead end | Contacts preceded by an abandoned manage-booking session within the hour. | Behavioural join, then usability testing of that exact path. | Avoidable contacts per thousand bookings. |
| Ownership confusion | Contacts that open with “who do I even talk to”, or arrive after contact with another party. | Transcript coding plus depth interviews with three-party bookings. | Repeat contacts on the same case. |
| Expectation gap on flexibility | Contacts from travellers who bought a flexible fare and were charged to use it. | Transcript coding, then a comprehension test of the fare terms as displayed. | Contacts citing flexibility, and refund requests on those fares. |
| Under-informed booking | Contacts about conditions that were technically disclosed but not seen. | Detail-page usability testing with eye on what was actually read. | Post-booking contacts in the first 48 hours. |
| Status anxiety | Contacts that ask for information rather than action: is it confirmed, has it been refunded. | Diary study across a disrupted trip; sizing survey. | Information-only contacts, which should be the easiest to remove. |
| Accessibility exclusion | Tasks that cannot be completed with assistive technology, appearing as contacts rather than as errors. | Assistive-technology testing on change and cancel, plus a WCAG 2.2 AA audit. | Task completion parity between assistive and non-assistive users. |
| Escalation blocking | Contacts repeated because the first attempt could not reach anyone able to decide. | Transcript analysis of repeat contacts; interviews with CS agents. | First-contact resolution, and the ratio of unavoidable contacts resolved once. |
Four ongoing research practices
A single project answers one question. These are the loops that make the second question cheaper than the first, which is how a research function of three people covers a domain this size.
Practice 01
The standing evidence base
Every study, finding and contradiction filed against the journey stage it belongs to, so the next question starts half-answered. The fastest research is the research you already did and can find.
RETURNS → Turnaround on a new question drops, without anyone cutting a study short.Practice 02
Contact-driver tracking
The driver taxonomy maintained as a live measure rather than a one-off analysis, split into avoidable and unavoidable from the start. It becomes the shared language between Research, Product and Customer Service.
RETURNS → The team can see whether a shipped fix actually removed the contact it targeted.Practice 03
Continuous contact with travellers
A standing slot of sessions with real travellers that anyone in the multidisciplinary team can attend. Watching one person fail a task changes more minds than any deck, and it distributes user advocacy instead of centralising it in me.
RETURNS → Product, UX and Engineering start arguing from evidence rather than from preference.Practice 04
Method development
The role asks for testing new ways of gathering insight. I would treat that as a real practice: one new tool or mixed method per quarter, run on a live question, with an honest written verdict on whether it earned its place. Including where AI-assisted analysis holds up and where it quietly invents a pattern.
RETURNS → The craft improves on evidence rather than on enthusiasm.How I would know it is working
Five measures. The first one is the one I would insist on, because it is the difference between reducing contacts and hiding them.
- 01
Avoidable contacts, not total contacts
The metric split into contacts that should never have happened and contacts that should be easy. Driving the first to zero while making the second faster. One number hides two opposite problems, and optimising the wrong one makes the product worse.
- 02
Self-service task completion, including with assistive technology
Measured on the real change-and-cancel paths, with parity between assistive and non-assistive users treated as the pass mark rather than a stretch goal.
- 03
Decisions changed
How often a finding altered what a team built, versus how often it confirmed what they had already decided. The only honest measure of whether research is influencing or decorating.
- 04
Time to a defensible answer
Falling, as the evidence base fills. Rising turnaround usually means the team is re-running work it already has.
- 05
Who else talks to travellers
Whether Product, UX and Engineering are attending sessions and citing users unprompted. If user advocacy lives only with the researcher, the researcher has become a bottleneck rather than a multiplier.
The thing I actually care about
I have spent six years turning research into things teams can act on the same week, in two-hour sessions with product owners, developers and business analysts who did not commission the work and did not have to agree with it. What I care about is not the study, it is the moment someone changes their mind because they finally saw what a person was actually trying to do. Everything above is a structured way of getting more of those moments, more often, with better evidence behind them.
Method, and what to distrust
Sources used
- Review corpora: Trustpilot (Booking.com, Booking.com Flights, Airbnb, Expedia), Apple App Store.
- Regulatory record: Spain’s CNMC decision, the Hungarian competition authority, the Dutch class action, and reporting on EU dark-pattern enforcement.
- Heuristic audit: the Baymard review of Booking.com, Q4 2023, 649 guidelines across 63 topics and three surfaces, benchmarked against 184 major ecommerce sites.
- Role context: the published Booking.com vacancy for Researcher II, Flights.
- Frameworks: Doblin’s Ten Types of Innovation (Keeley), Roger Martin’s integrative thinking.
Known biases and limits
- Trustpilot self-selects for grievance. A 1.5 is not a satisfaction score; it measures how bad the worst experiences are and how motivated people are to record them.
- App stores self-select for satisfaction, and are usually rated before the trip, not after a problem.
- The Flights sample is small. 371 reviews is enough to show a pattern and not enough to size one. I have treated it as directional throughout.
- The Baymard review is Q4 2023. Some of it will have moved. Where I quote a score I am quoting that review, not today’s product.
- I could not verify accessibility directly. No scan, no access to the statement from this environment. The accessibility section is explicitly inference and a proposed method, not findings.
- This is outside-in research. No internal data, no contact-driver taxonomy, no roadmap, no constraints. Everything here is a hypothesis that a fortnight inside the team would sharpen or kill, and path 04 is written on that assumption.
Sources
- Trustpilot: Booking.com
- Trustpilot: Booking.com Flights
- App Store: Booking.com
- Trustpilot: Airbnb
- Trustpilot: Expedia
- Baymard Institute: UX benchmark
- Dark patterns on Booking.com: manipulation tactics and legal consequences
- Cybernews: Booking.com sued over dark patterns
- CNN: campaigners on Booking.com’s sales tactics
- Hospitality Today: Europe vs Booking.com
- Booking.com: accessibility statement (not retrievable from this environment)
- Doblin: Ten Types of Innovation
- Roger Martin: The Opposable Mind (integrative thinking)
How I fit this position
Six years of research
that had to change someone’s mind.
The role asks for three to five years of UX research delivery in a consultancy or in-house setting, autonomy across the full research lifecycle, both primary and secondary methods, and the ability to influence multidisciplinary teams. That is a description of what I have done at Baymard every week for six years: scope it, run it, and then sit in a two-hour session with product owners, developers and business analysts who did not commission the work and do not have to agree with it.
The first project named in this brief, post-booking support in Flights, is the kind of question I have spent my career answering: find where the task breaks, size it, and hand back something a backlog can act on.
How I fit this position
What the role asks for, how I would approach it, and where I would be learning rather than arriving expert. The research in the first tab is the working sample.
How I would allocate the capacity
Commitments only mean something if they add up. This is the whole hundred per cent for a researcher embedded in one domain, and what happens to it when a study is in the field.
Steady state
45 · 25 · 20 · 10 = 100
A fifth of the time on communication is deliberate and, in my experience, the part most often under-resourced. A finding nobody acts on cost exactly as much to produce as one that changed something.
In the field
60 · 20 · 15 · 5 = 100
During fieldwork delivery takes the load, and I would say so in advance rather than quietly missing other commitments. Method development is the first thing to pause and the first thing to restore.
-
01
Drive research end to end, independently
At Baymard I scope, run and deliver evaluations of large ecommerce and travel products on my own: objectives, method, execution, analysis and communication. Each one produces up to fifty prioritised recommendations grounded in ongoing user research, and each recommendation ships with a mockup showing how the issue could be solved, because a finding that a team cannot act on has not finished being research.
COMMITMENT → 45% of capacity on delivery, with the first study framed in week one rather than after a scoping quarter. -
02
Mix primary and secondary methods, and know what each one cannot tell you
My daily work is expert and secondary method: heuristic evaluation, literature and desk research, competitor analysis. My earlier work was primary and qualitative at the other extreme: ethnographic research across the Western Cape to design government services for rural communities, and two healthcare systems taken through iteration with the clinicians who used them. I co-authored two published papers on personas for rural users of digital services and presented at an international conference in Rio de Janeiro. Path 04 sets out the method mix I would bring to the contact-rate question, and what each method is blind to.
-
03
Be the user’s advocate inside a multidisciplinary team
Six years of consulting has meant persuading people who do not report to me, did not ask for the finding, and have a roadmap already. What works is not conviction, it is bringing the evidence and the design together so agreeing is the easy option. I would also open sessions to the team rather than reporting back from them, because watching one traveller fail a task changes more minds than any deck, and it distributes advocacy instead of centralising it in me.
-
04
Connect disparate data and distil complex datasets
The contact-rate question is a joining problem before it is a research problem: support transcripts, behavioural data from the manage-booking flow, survey sizing and interview transcripts each answer a different part of it and none answers it alone. I have built structured processes for capturing customer feedback from chat and support and feeding it into product prioritisation before, at Allan Gray and at Luno, and that is essentially the same pipeline this project needs.
-
05
Test new ways of gathering insight
The role explicitly asks for this, and I would treat it as a real practice rather than an aspiration: one new tool or mixed method per quarter, run on a live question, with an honest written verdict on whether it earned its place. That includes where AI-assisted analysis genuinely accelerates coding and where it confidently invents a pattern that is not in the data, which is a question I care about a great deal.
SEE → Path 04 is this thinking applied: a full research plan for the post-booking contact-rate project.
Two honest gaps
Flights and travel. I have no aviation or OTA experience. The brief lists it as preferred rather than essential, and everything in the first tab was built from a standing start in public sources, which is the closest thing to evidence I can offer about how quickly I get up to speed in an unfamiliar domain.
Statistical software. The role names SPSS, R or JMP. My quantitative work has been survey design, sizing and analysis rather than modelling in those environments, and I would be learning the toolchain rather than bringing it. I would rather say that now than discover it in week three. I have picked up every tool this industry has thrown at me since Flash, and I learn quickly, but that is a claim about aptitude, not a claim about experience.
About me
I hold two viewpoints at once, and keep a close pulse on both.
I can hold two viewpoints, keep a close pulse on both, and clearly articulate, facilitate and iron out any creases at either level.
Roger Martin calls it integrative thinking. I call it listening: sadly a skill not many have these days. I believe good leadership takes careful listening, and that listening is a practice that requires constant work. In a role whose entire authority is credibility, it is also the only thing that actually moves anyone.
About me
One thing worth knowing, and it happens to be the thing this whole document is built on.
I have a remarkable ability to hold two viewpoints, keep a close pulse on both, and clearly articulate, facilitate and iron out any creases at either level.
Roger Martin calls it integrative thinking. I call it listening: sadly a skill not many have these days. I believe good leadership takes careful listening, and that listening is a practice that requires constant work.
Building the function, not just the work
In three companies where I was lead product designer, I was the first design team member: I created the design processes, hired my colleagues, and in two companies hired my own managers, who later became head of product design. I was always obsessed with making the design process as smooth as possible as the industry quickly adapted from Adobe (as early as Flash), to Sketch, to Figma, and now to AI. I am driven by curiosity and learn immensely quickly.
Leading change by getting everyone on the same page
With a track record of leading organisational change, I have been described as incredibly inspiring and optimistic. I have a natural ability to reverberate keenness, and I am always seeking how we can all be on the same page and maintain absolute clarity around our goal posts.
What I have not done traditionally
I do not have “traditional” line-management experience, and I have not sought it. I have mentored and hired throughout my career, including hiring a Head of Product, but the path I have chosen is depth of craft rather than headcount. For an individual contributor research role that is not a caveat, it is the point.
Presenting, writing and consulting
I thrive in presentation mode. I love talking about complexity and chaos in a way that gets everyone invested and on board, and hand in hand with that, I thoroughly enjoy documentation, reports, presentations and pitches. I have spent the past six years consulting with companies across the globe on how they can improve their product, through two-hour deep dives, followed up with team implementation calls and expert advice calls. I have also presented academic research at conferences from Brazil to Zimbabwe.
I still read the code
I have a history of front-end development: HTML, CSS and PHP, and the new possibilities now available to us are just the start. I am keeping a close and very geeky eye on all the new realms as our industry evolves further.
Where you can see it in this document
Design strategies treats three of Booking.com’s problems as genuine tensions rather than trade-offs: conversion versus trust, contact rate versus contactability, and research depth versus delivery speed. Each resolves into a position that keeps both sides rather than picking one. That section is the argument, not a summary of someone else’s.
Why it produced this research
Booking.com has an app rated 4.8 by 2.1 million people and a Trustpilot score of 1.5 from 125,582. Most readings pick a side: Booking.com is excellent, or Booking.com is terrible. The useful answer holds both at once, because both are true of the same product at different moments, and the moment they diverge is precisely what this role has been asked to research.
The choice, as I see it
The risk with hiring for the résumé is inheriting the playbook that came with it. Methods already run somewhere else, applied again here, without anyone asking whether they answered the question the first time.
OR
You could hire someone dynamic, engaging and curious, who has already done a week of the work before the interview.
Get to know the human behind all this
Irani
Sangham
Human on planet Earth, mother, and a terrible pun pusher.
Get to know the human behind all this
A little about me that has nothing to do with design.
South Africa to the Netherlands, via a great deal of learning
I was born and raised in South Africa. On weekends I was either hiking the wild mountains or surfing in the ocean.
In 2022 my husband and I moved to the Netherlands. He got a job here, so we decided to give it a try, and I was six months pregnant with my son at the time. Navigating a country move while pregnant is no small feat.
I have two children, and life since has been full of learning new things and upskilling incredibly quickly, in a completely new country, on the literal other side of the planet from my familiar environment and support system. It has been the most difficult and challenging time, but I have enjoyed making every day just a little bit easier by consistently trying my best, and those skills are directly transferable to this role.
I am the primary market
A mother, constantly looking for kids’ clothes every season, books in two languages, and of course any classic vintage items for myself along the way. The research in the first tab is not only desk work. It is my own weekly experience of the product.
Fun to have around, allegedly
I have been told I am very fun to have around and that I have a knack for making people laugh. I do love a good and bad joke, and I have disastrous puns, but I am not sorry about that.
I have been an individual contributor by choice, and I intend to stay one
You will notice I have not held roles titled manager or director. That is deliberate, and it is not a gap. For most of my career I have steered towards depth of craft rather than headcount: being the person who does the work, sets the bar and brings others up to it, rather than the person who owns the org chart. Six years ago I made that explicit, moving into a consultative role at Baymard and in my own capacity, partly to put my personal life at the centre while moving country and having two children, and partly because it is simply the work I am best at.
This is an individual contributor role, supported by another researcher and a Research Manager, and that is exactly where I want to be. I would rather go deeper into the craft than wider into an org chart. Six years of consulting has been a deliberate choice to keep doing the work rather than managing people who do it.
Some things I like
Competitive swimming
As a child my goal was the Olympics. I trained four hours a day and represented my country on numerous occasions, from age eight to fourteen.
Shaolin Kung Fu
From fourteen, I trained for fifteen years. I taught it, and competed for South Africa.
Ambidextrous, and juggling
Balls, and life. Both ongoing.
Coffee, and sourdough
A long-standing coffee nerd, and recently obsessed with sourdough and its many applications.
The growing brain
A lot of my time goes into toddler activities and investigations at the moment. The growing brain is fascinating to me.
Mountains and ocean
Hiking and surfing were the shape of every South African weekend, and the habit has not left.
Testimonials
She earned the complete trust of my entire team.
Lauren Hunt · on working together at Luno
Four colleagues and two clients, on what it is actually like when I am in the room.
Irani has had a massive impact on my career as a product designer. She has always made time to give me advice, resources and guidance to further my career, and continues to do so. She inspires through being extremely knowledgeable and capable while remaining immensely humble. She earned the complete trust of my entire team at Luno, and I would jump at the opportunity to work with her again.
Irani is very good at holding complexity and clarity, opposing points of tension, in her head simultaneously. She can empathise with the user and understand the problem while recommending a simple solution for the team to implement.
Irani, thank you for making an incomparable difference. You are an integral part of the team and possess a quiet composure, low-key witticism (and a vast supply of puns), and a warmth that makes it a pleasure to work with you. Not only that, but over time you have improved the work flow of the team by continually coming up with clever, innovative ideas.
Kudos going straight to Irani for organising an awesome summit. Practical, insightful, kept to schedule and beautifully facilitated.
Irani has made accessibility tangible and actionable for us. We have a clear list of where to start, removing the confusion and overwhelm. Thank you, Irani.
Irani has an exceptional ability to hold a room and get people to be productive. I felt like I had ideas flowing, and that any idea I put down was a good contribution. Great facilitation and run-through of our product improvement possibilities.