DoorDash Data Scientist candidates commonly report a first stage of SQL plus a product case, followed by a loop of case studies and a behavioral or product partner conversation. Diagnosing metric drops and experiment design come up repeatedly.
$201K
Avg. Base Comp
$292K
Avg. Total Comp
Not reported
Typical Rounds
Not reported
Process Length
Candidates describe the DoorDash Data Scientist interview as a short first stage followed by a case-heavy loop. The first stage pairs SQL with a product case. Reports describe it as about an hour in total, or as two back-to-back 30-minute halves. One candidate worked through four related SQL questions on the same tables that got progressively harder, so watch edge cases and pace yourself on the later questions. Another describes a rolling-window aggregation and a day-7 retention query where the signup date had to be inferred from activity data. That candidate was also told to tighten up on CTEs and window functions. Calling out your assumptions explicitly is worth practicing.
The case half is where most reports focus. A recurring prompt is diagnosing a drop, such as fewer successful deliveries in Los Angeles or orders falling in a specific location. Break the problem into supply, demand and restaurant-side causes, and prioritize hypotheses before proposing solutions. Separate short-term fixes from longer-term ones. Another reported prompt asks you to define success metrics for each side of the three-sided marketplace for a fast-delivery feature and plan its rollout. In that case the candidate was pushed on A/B testing in depth, including sample sizing and multi-arm bandits.
Candidates who reached the later loop describe three or four product case interviews plus a behavioral conversation with a product manager or product partner. One candidate found the cases overlapped. Another said the questions were simpler than expected and rewarded clarity of thought over technical complexity. Outcomes vary: one candidate cleared the technical screen but stalled at team matching, and another did not receive an offer. Practice structuring answers out loud, writing down your structure, naming metrics for each side of the marketplace, and stating tradeoffs explicitly.
Synthesized from 48 candidate reports by our editorial team.
Had an interview recently?
Share your experience. Unlock the full guide.
Real interview reports from people who went through the Doordash process.
Recruiter screen, after that it's 1 hours of interview that consist of 30 min analytical design, 30 min case question, if you pass that then it's 3 rounds of case question, 1 business partner interview, 1 hiring manager interview
Questions asked: Your current work, specific projects, how do you handle conflicts at work, what are you looking for in the next role, what's important to you. A subscription and order table is given in the sql round that contains subscriptions information, orders, delivery information. Tables ● subscription_units Definition: A subscription unit is a one-month unit for a given subscription_id (one row per subscription unit).
○ subscription_unit_id integer — unique id for each subscription unit (one row per subscription month) ○ subscription_id integer — id for the subscription (grouping multiple subscription units) ○ user_id integer — id for the user ○ subscription_type STRING — free_trial, monthly, annual ○ subscription_unit_start_date TIMESTAMP ○ subscription_unit_end_date TIMESTAMP ○ last_subscription_unit_id integer — ID of the prior subscription unit if this row is a renewal; NULL if first month ● orders ○ order_id integer ○ user_id integer ○ is_completed BOOLEAN — was the order completed or not ○ is_pick_up BOOLEAN — was the order for delivery to customer or customer pickup ○ order_placed_at TIMESTAMP ○ delivered_at TIMESTAMP - actual delivery time ○ estimated_delivered_at TIMESTAMP — original estimated delivery_time ○ market_id integer ○ line_of_business STRING — restaurant, grocery, convenience, etc ○ merchant_id integer ○ device_type STRING — ios, android, web ○ subtotal INTEGER — cents ○ delivery_fee INTEGER — cents Question: The DashPass team suspects that customers who experience late deliveries are less likely to renew their monthly subscription. Using the available tables, how would you investigate this? Walk me through your analysis framework. AI is allowed for the interview, the interview is mainly to evaluate your personal analytical reasoning and problem solving abilities.
Share your own interview experience to unlock all reports, or subscribe for full access.
Sourced from candidate reports and verified by our team.
Topics based on recent interview experiences.
Featured question at Doordash
Select the 2nd highest salary in the engineering department
| Question | |
|---|---|
| Monthly Customer Report | |
| Experiment Validity | |
| Button AB Test | |
| Month Over Month | |
| Average Order Value | |
| Over-Budget Projects | |
| Longest Streak Users | |
| Netflix Retention | |
| WAU vs Open Rates | |
| Network Experiment Design | |
| Daily Retention Summary | |
| Instagram TV Success | |
| Delivery Estimate Model | |
| Group Success | |
| Random Bucketing | |
| Post Composer Drop | |
| Christmas Dinner Ingredient Optimization | |
| Google Maps Improvement | |
| Uber User Journey | |
| Comparing Search Engines | |
| Marketing Channel Metrics | |
| Biggest Tip | |
| Valid Anagram | |
| Cancellation Fees | |
| Hurdles In Data Projects | |
| Recruiting Leads | |
| Forecasting New Year Revenue | |
| Food Delivery Times | |
| Success Measurement |
Synthesized from candidate reports. Individual experiences may vary.
One candidate reports a recruiter screening before the technical interview. Expect a conversation about your background and the role. This step is not described in every account.
Candidates describe SQL as one half of the first-stage screen, either as a 30-minute block or as half of a 60-minute interview. One candidate worked through four related questions on the same tables that got progressively harder. Another saw a rolling-window aggregation and a day-7 retention query and was advised to tighten up on CTEs and window functions.
The paired case often asks you to diagnose a metric decline, such as a drop in successful deliveries in Los Angeles or orders falling in a specific location. One candidate also reports an A/B testing question focused on ads. Another describes a feature-launch case covering metrics for each side of the marketplace and the rollout. Structure and clear prioritization of hypotheses appear to matter.
Candidates who advanced describe several product case interviews: three in one report and four in another. One candidate found the cases overlapped. Another said the questions were simpler than expected and rewarded clear, structured reasoning. The number of sessions may vary by candidate and level.
Reports mention a behavioral round with a product manager, or a session with a product partner, alongside the cases. Outcomes vary; one candidate reported clearing the technical screen but not advancing because of team matching.
