← Back to Directory

Thumbtack interview notes, page 1 of 3

Mid-sized

Most of these Thumbtack write-ups come from Data Scientists (19), then Product Managers (3). The most reported rounds are the technical phone screen (18), the product sense round (2), the cross-functional panel (1) and the system design round (1).

22 interview notes · updated Jul 2026

Thumbtack·Product Manager·Onsite - Product Sense / Strategy

Apr 2026
Thumbtack PM interview, just one open-ended product design question. Pretty light on details but the prompt itself has some interesting angles if you think it through.
  • Design a birthday-related product.

“I went in thinking this would be easy and kind of coasted on surface-level ideas at first.”

View Post

Thumbtack·Data Scientist·Technical Phone Screen

Apr 2026
Thumbtack DS interview focused entirely on a product metrics case for a new 'Instant Book' feature. It was a deep, multi-part question that covered everything from north-star metrics to rollout strategy and risk mitigation. Felt more like a PM case than a typical DS interview, which threw me off a bit.
  • Thumbtack is launching an 'Instant Book' feature where customers can immediately book a pro at a pre-set price without waiting for quotes. Design the full measurement framework: north-star metric, secondary objectives, diagnostic metrics, instrumentation plan, rollout strategy, success criteria, and key risks with mitigations.
  • How would you define and instrument the metrics to distinguish organic bookings from Instant Book conversions, handle multi-device sessions, and avoid double-counting when a request receives both a quote and an Instant Book?
  • What are the main risks of rolling out Instant Book and how would you monitor and mitigate them?

“This was essentially six questions wrapped into one.”

View Post

Thumbtack·Data Scientist·Technical Phone Screen

Apr 2026
Thumbtack data scientist interview that was basically a graduate-level stats exam disguised as a conversation. Heavy on A/B testing mechanics, sample size math, and diagnostic reasoning. Walked out feeling like I'd just taken a stats final.
  • Define statistical power for a two-proportion A/B test and rank the main levers that increase power from most to least practical impact, covering effect size, variance, sample size, allocation ratio, alpha, variance reduction techniques, stratification, and test duration.
  • Given a baseline conversion of 5%, a target relative lift of 7%, two-sided alpha of 0.05, and 80% power with equal allocation, compute the required sample size per variant and the minimum test duration assuming 80,000 eligible users per day with 10% post-randomization attrition. Show your work.
  • Redo the sample size calculation assuming a variance reduction technique (like CUPED) achieves a 30% relative variance reduction. What changes in the required sample size and test duration?
  • How does switching to a 90/10 traffic split (90% control, 10% treatment) affect power at fixed total traffic compared to an equal split? Give both intuition and a quantitative comparison.
  • Your test returns a statistically significant negative 2% lift, opposite to your expected positive 7%. Walk through a step-by-step diagnostic process before drawing any conclusions, covering sample ratio mismatch, instrumentation audits, bot and demographic imbalances, novelty effects, outlier handling, segment-level breakdowns, guardrail metrics, and peeking risk.
  • After completing diagnostics on a surprising negative result, propose a decision framework for when to ship the change, when to iterate with a follow-up test, and when to simply rerun the original test. Be specific about what conditions justify each path.

“This felt like a warm-up but it wasn't.”

View Post

Thumbtack·Product Manager·Technical Phone Screen

Apr 2026
Thumbtack product analytics question, probably a phone screen or onsite case round. Just the one question but it's a meaty one if you haven't prepped metric investigation frameworks.
  • Thumbtack's revenue is down 4% week over week. How would you investigate what's causing it?

“This is the kind of question where you can spiral fast if you don't slow down.”

View Post

Thumbtack·Data Scientist·Onsite - Cross-functional / Panel

May 2026
Thumbtack DS interview with a structured presentation exercise that felt more like a performance review than a traditional interview. The bar for rigor was real.
  • Walk a mixed panel (PM, engineer, DS) through a past data science project in about 7 minutes, covering the business problem, the decision it was meant to inform, and the north-star metric you chose.
  • As part of the same presentation, explain your data sources, the key assumptions you made, and what risks those assumptions introduced.
  • Describe the modeling or analysis approach you used and explain why you ruled out the alternatives you considered.
  • Present your results including confidence intervals and any sensitivity analyses you ran.
  • Compare the projected impact you forecasted before launch to the actual shipped impact, and explain how you validated the results post-launch.
  • Name two failures or deliberate trade-offs you accepted during the project and explain why you made those calls.
  • Close your presentation with a 60-second roadmap for the next iteration of this project.

“The tricky part is calibrating for three different audiences at once.”

View Post

Thumbtack·Data Scientist·Technical Phone Screen

May 2026
Thumbtack data science interview that went deep on experimentation design for a marketplace matching model. The question was essentially one giant multi-part case and it covered pretty much every corner of A/B testing you can think of.
  • You have a new matching model ready to ship. Design a full A/B test for it, covering primary metrics and guardrails for both sides of the marketplace, randomization strategy, power analysis, variance reduction and validity checks, a pre-registration document with stop/go criteria, and how you'd analyze heterogeneous lift without p-hacking.

“This was a lot.”

View Post

Thumbtack·Data Scientist·Onsite - System Design / Architecture

May 2026
Thumbtack data scientist interview, system design round focused entirely on a single gnarly streaming problem. The question was dense and covered a lot of ground, felt more like a distributed systems eng interview than a DS one.
  • Design a streaming system that tracks new vs returning users per calendar month, emitting counts and percentage shares at month end. The system must handle up to 1B distinct users, 50K requests per second, 8 GB RAM, late-arriving events up to 7 days out, duplicates, and limited per-user state. Cover your data structures and memory footprints, exact vs approximate tradeoffs with error bounds under 0.5 percentage points, late event and watermarking strategy, deduplication, time and space complexity, and fault tolerance.

“This one wrecked me a little.”

View Post

Thumbtack·Data Scientist·Technical Phone Screen

May 2026
Thumbtack data scientist interview with a Monte Carlo simulation problem. Pretty much a coding screen dressed up as a stats question, which I didn't fully expect going in.
  • Simulate the Monty Hall problem using Monte Carlo methods to estimate the win probability for both the 'stay' and 'switch' strategies, given n rounds per strategy. Use a fixed random seed and output both probabilities rounded to 2 decimal places.

“I knew the theoretical answer going in (1/3 vs 2/3) so my first instinct was to just print those values and call it done.”

View Post

Thumbtack·Data Scientist·Technical Phone Screen

May 2026
Got a SQL question for a Data Scientist role at Thumbtack that was more involved than I expected. Single query, no dynamic SQL, and they wanted you to verify your output against sample data too. Felt like a legit test of whether you actually understand window functions and conditional aggregation or just know the syntax.
  • Given a users table and a requests table, write a single PostgreSQL query (no dynamic SQL) that returns, for every calendar month present in requests, the counts and percentage shares of requests from NEW versus RETURNING users. A user is NEW in any month where that month equals their first-ever request month; otherwise they are RETURNING. Output should include month, new request count, returning request count, new share pct, and returning share pct, ordered by month ascending. Also verify your query against the provided sample data.

“This one took me a minute to untangle.”

View Post

Thumbtack·Data Scientist·Technical Phone Screen

Jun 2026
Technical phone screen for a Data Scientist role at Thumbtack. The questions were heavily focused on Python internals and data engineering fundamentals, more CS-heavy than I expected for a DS position. Left feeling like I'd done okay on the high-level stuff but probably got too vague on some of the implementation details.
  • Compare Python lists and dicts across append, insert, lookup, update, and delete operations. What are the average and worst-case time complexities, and what are the memory and ordering implications in CPython 3?
  • Write concise code showing how to append an element to a list and update a value in a dict.
  • What is the difference between JSON and CSV formats, and when would you choose JSON over CSV? Consider nesting, schema changes over time, interoperability, and compression.
  • Show Python code to stream-read a JSON Lines file line by line using json.loads, read a CSV using csv.DictReader, and use pandas read_csv with chunksize to compute the sum of a numeric column without loading the full file into memory.
  • How would you handle malformed rows, missing or NaN values, bad encodings, and numeric overflow when processing large files? Also, what chunk size would you use for a 10 GB file on a machine with 16 GB RAM, and is there a non-pandas streaming alternative?

“I knew the O(1) amortized append for lists and O(1) average lookup for dicts, but fumbled a bit on worst-case dict behavior (hash collisions, O(n)).”

View Post