Company interview guide

Booking.com Machine Learning Interview Guide (2026)

Case-driven ML and data science interviews on real travel-marketplace problems, with a coding track for ML engineers.

Booking.com's machine learning interview loop has 7 reported stages, from online test to culture fit and hiring manager; typical length: 3-12 weeks.

Roles covered: Applied Scientist, Data Scientist and ML Engineer. The question bank holds 3 questions reported from Booking.com interviews.

Try a free question from this loop's topics

Interview process length
3-12 weeks
Stages
7
Reports linked to bank questions
4
Bank report dates
Dec 2018–Jun 2022
Questions in bank
3 (0 free)
Last reviewed
October 2026

How the Booking.com interview loop works

The company describes an application review, a recruiter call, a business stage of several rounds with your future manager and team, then feedback within about a week; it puts the whole process at three to six weeks and says technical roles with tests and coding interviews can take longer. Candidates report anything from two weeks to three months. Scientist and data scientist loops: an online multiple-choice ML and statistics test (common in older reports, rarer since 2023), a recruiter call, then two case interviews (one on the spot, one prepared from a brief or a take-home and presented) and a final culture fit or hiring-manager round. ML engineer loops: an online coding test, a live coding interview, an ML system design interview and a final behavioral round, although some engineer loops in 2024 used the same two case interviews instead. One candidate was told that passed technical rounds stay valid for six months for similar scientist roles. Only about a dozen of our reports date from 2024-2026.

The Booking.com interview process, stage by stage

01Online test

Timed online test after applying: multiple-choice ML and statistics questions, sometimes with one coding problem

Many candidates got an online test before speaking to anyone. Data and scientist versions are multiple choice on ML, statistics and probability; one candidate had 12 questions in 30 minutes, another about an hour including a programming problem. A 2024 engineer version paired one string-processing coding task (rank hotels by keyword mentions in their reviews) with nine ML questions. Fewer recent scientist reports mention the test.

What it tests

  • When a given model works well or badly, and why
  • Regularization, model complexity, decision boundaries and validation
  • p-values and basic probability
  • Careful reading of a coding spec with small input traps

How to prepare

  • Practise answering applied ML multiple-choice questions at about two minutes each; reports call the time tight.
  • In the coding task, normalise case and strip punctuation before counting words; one candidate says that was the trap.
  • Revise bias-variance, regularization and evaluation metrics before anything exotic.

Online test in 13 data, scientist and engineer reports, mostly 2018-2022 plus one 2024 engineer report. The company's hiring page says technical roles may include tests.

02Recruiter call

About 30 min phone or video call with a recruiter or sourcer

A conversation about your background, motivation, location and expectations, and a walk-through of the rest of the process. Reports say recruiters explain the stages in detail and later relay feedback. In a few older data science loops a short hiring-manager call with basic technical questions (bootstrap, linear regression) came next.

What it tests

  • Motivation for Booking.com and the role
  • Relevant projects and experience in brief
  • Logistics: location, relocation, salary expectations

How to prepare

  • Ask which case format each later round uses (on the spot, prepared brief or take-home) and who will interview you.
  • Have a short summary of one project with a measurable business result ready.
  • Ask whether relocation is offered; several current postings say it is not.

Recruiter or HR call in 30 data, scientist and engineer reports, 5 of them since 2024. The company's hiring page describes this first call.

03Coding interview (engineer track)

Live coding interview in Python, usually one medium algorithm problem

ML engineer candidates describe a standard algorithm interview, typically one medium-level problem. Reported topics include graphs and trees and a recursive, knapsack-like task: plan a road trip through hotels within a budget. Scientist and data scientist loops rarely include it; one older data science loop instead had a two-hour hands-on modelling session in a notebook environment.

What it tests

  • Graphs, trees and recursion in Python
  • Search or dynamic programming under a budget constraint
  • Explaining your reasoning while you code

How to prepare

  • Drill recursion with memoisation and 0/1 knapsack variants.
  • Talk through the problem before coding; the hiring page describes the technical interview partly as a communication exercise.
  • Practise graph traversal and tree problems in plain Python.

Live coding round in 5 ML engineer reports from 2022-2024, consistent in style. The company's hiring page mentions coding interviews for technical roles.

04Live business case

60-75 min with two data scientists or ML scientists: often a past-project walk-through, then one or two cases solved on the spot

The core of the scientist and data scientist loop, often called a craft interview. Interviewers give an open problem from the travel marketplace and push you from framing to data, model, offline metrics and an online experiment. Older loops ran one such screen by phone and two more on site; recent reports describe one live case round next to one prepared case. Several candidates felt the interviewers had a preferred answer in mind, and A/B testing comes up in most write-ups.

What it tests

  • Turning a vague business goal into an ML or analysis problem with a clear target
  • Choosing data, features, a baseline and offline metrics
  • Designing an A/B test and naming the success and guardrail metrics
  • Handling cold start, ranking and trade-offs for travellers and partners
  • Presenting a past project and what you would do differently

How to prepare

  • Practise with a timer on hotel, pricing, recommendation and cancellation problems, speaking every step aloud; one candidate was marked down for silent pauses.
  • Give concrete, quantitative answers: name the label, the model and the metric instead of listing options.
  • Close every case with how you would test it online and what could bias the result.
  • Read the company's public ML work on recommendations, ranking and experimentation beforehand.

On-the-spot business or ML case in 28 data, scientist and engineer reports, 6 since 2024. The company's hiring page describes a business stage of rounds that test technical skills.

05Prepared case or take-home

Brief sent a few days ahead (slides, no code) or a take-home modelling case, then about 60 min presenting and taking questions

Most recent scientist loops add a second case you prepare in advance. Reported briefs include recommendation blocks for the home page, a destination recommender and a payment fraud model; one candidate had ten minutes to present before questions. Interviewers dig into your choices, so expect to defend data, labels and evaluation rather than read slides. Some job postings from 2025 describe three business interviews, one being a take-home business case.

What it tests

  • Structuring a recommender or fraud-detection solution end to end
  • Data and label choices, leakage and evaluation
  • Communicating a solution clearly in a short presentation
  • Defending trade-offs under follow-up questions

How to prepare

  • Keep the presentation short and leave most of the hour for questions.
  • Prepare answers on cold start, offline-online gaps and how you would A/B test the result.
  • For a fraud brief, cover class imbalance, label delay and the cost of false positives versus false negatives.

Prepared brief, presentation or take-home in 7 scientist, data and engineer reports, 3 since 2024; 2025 job postings also list a take-home business case for scientist roles.

06ML system design (engineer track)

One ML system design interview

ML engineer loops include a system design round on an ML product, for example hotel pricing, and how you would build and run its pipeline. One 2022 candidate was rejected at this stage; others say it was the expected ML design problem. Some 2024 engineer loops replaced it with the case interviews above.

What it tests

  • Steps of an ML pipeline from problem definition to deployment and monitoring
  • Designing pricing or ranking models for a marketplace
  • Serving, retraining and data pipelines at scale

How to prepare

  • Practise walking from objective and data to training, deployment and monitoring in a clear order.
  • Prepare one design for pricing and one for recommendations, both with an online test plan.

ML system design round in 5 ML engineer reports from 2022-2025; 2025 engineer job postings also list a system design interview.

07Culture fit and hiring manager

About 60 min behavioral round with the hiring manager, sometimes with a tech lead or product manager

The final round on values, collaboration and past work, usually with the hiring manager and sometimes a product or analytics manager. Candidates advise reading the company values first. A few reports say this round turned technical or became a deep dive on an unrelated past project, so be ready to go into detail on anything on your CV.

What it tests

  • Fit with the company values, such as customer first and owning outcomes
  • Times you failed or worked with sceptical stakeholders
  • Why Booking.com and why this team
  • Depth on any project on your CV

How to prepare

  • Map one story to each company value before the round.
  • Prepare a failure story with what you changed afterwards.
  • Expect follow-up on any CV project, not only the ones you choose.

Culture fit, behavioral or hiring-manager round in 23 data, scientist and engineer reports, 4 since 2024; 2025 job postings also list a final behavioral interview.

Questions and themes reported most often

  • A/B testing and measuring the impact of a product change (8 reports, live business case)
  • Design a recommender: destinations, home-page blocks, or hotels with no users yet (7 reports, prepared case or take-home)
  • Identify and highlight good hotel deals or best value for money (4 reports, live business case)
  • ML and statistics multiple choice: regularization, p-values, model complexity (4 reports, online test)
  • Predict how many days until a hotel is fully booked (2 reports, live business case)
  • Failure stories, sceptical stakeholders and why Booking.com (2 reports, culture fit and hiring manager)
  • Prioritise which property descriptions to translate (1 reports, live business case)
  • Payment fraud detection model as a take-home (1 reports, prepared case or take-home)

What Booking.com asks about

  • Business and ML case framing34%
  • Experimentation and causal inference18%
  • Recommendation and ML system design16%
  • Coding and algorithms12%
  • ML and statistics fundamentals10%
  • Behavioral and values fit10%

Booking.com interview prep checklist

  • Solved three travel cases aloud in under 30 minutes each, ending with an A/B test plan
  • Designed a hotel recommender that handles cold start and can be tested online
  • Can explain how to pick a good-deal badge and prove it helps bookings
  • Built a short slide deck for a prepared case and practised defending it
  • Engineer track: solved medium graph, tree and knapsack-style problems in Python
  • Revised regularization, bias-variance, p-values and evaluation metrics
  • Prepared one story per company value, including a failure

Tips

  • Ask the recruiter early whether your track uses case interviews, a coding round or both; scientist and engineer loops differ.
  • Practise open travel-marketplace cases aloud: deals, rankings, recommendations, pricing, cancellations and translations.
  • For every case, go from goal to label, model, offline metric and an A/B test with guardrails.
  • Be concrete and quantitative; candidates say solving the case with a specific, measurable plan is what counts.
  • Revise experimentation thoroughly: sample size, novelty effects, interference between travellers and partners, and what to do without a test.
  • Rehearse a ten-minute presentation of a prepared case and leave room for questions.
  • Engineer candidates should still drill medium algorithm problems, recursion and knapsack-style search.
  • Read the company values and have one story for each before the final round.
  • Use the written feedback many candidates receive; one candidate was told passed technical rounds stay valid for six months.

Practice the Booking.com loop

Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.

Booking.com interview FAQ

How many interview rounds does Booking.com have for ML roles?

Candidates report 7 stages: Online test, Recruiter call, Coding interview (engineer track), Live business case, Prepared case or take-home, ML system design (engineer track) and Culture fit and hiring manager. The company describes an application review, a recruiter call, a business stage of several rounds with your future manager and team, then feedback within about a week; it puts the whole process at three to six weeks and says technical roles with tests and coding interviews can take longer. Candidates report anything from two weeks to three months. Scientist and data scientist loops: an online multiple-choice ML and statistics test (common in older reports, rarer since 2023), a recruiter call, then two case interviews (one on the spot, one prepared from a brief or a take-home and presented) and a final culture fit or hiring-manager round. ML engineer loops: an online coding test, a live coding interview, an ML system design interview and a final behavioral round, although some engineer loops in 2024 used the same two case interviews instead. One candidate was told that passed technical rounds stay valid for six months for similar scientist roles. Only about a dozen of our reports date from 2024-2026.

How long does the Booking.com ML interview process take?

Reported timelines run 3-12 weeks.

What does Booking.com ask in ML interviews?

By reported question mix: Business and ML case framing 34%; Experimentation and causal inference 18%; Recommendation and ML system design 16%; Coding and algorithms 12%; ML and statistics fundamentals 10%; Behavioral and values fit 10%. The most repeated themes: A/B testing and measuring the impact of a product change; Design a recommender: destinations, home-page blocks, or hotels with no users yet; Identify and highlight good hotel deals or best value for money.

How should I prepare for the Booking.com ML interview?

Ask the recruiter early whether your track uses case interviews, a coding round or both; scientist and engineer loops differ. Practise open travel-marketplace cases aloud: deals, rankings, recommendations, pricing, cancellations and translations. For every case, go from goal to label, model, offline metric and an A/B test with guardrails.

Are there free Booking.com interview questions?

LeetML has 3 questions reported from Booking.com interviews. The stage sections above link the free practice problems for the topics each round covers; the reported-question set itself is in the app.

Related interview guides