Accenture · Primly Community

Accenture data scientist interview (SQL + case + stats): what actually came up

ds_dmitri · 4 replies

just finished my Accenture data scientist loop for an Applied Intelligence role (US, hybrid, mid-senior level). took about 3 weeks from recruiter screen to offer. figured I'd write it up since there's not much recent DS-specific info out there.

the process had 4 rounds:

recruiter screen (30 min): standard. experience summary, why consulting, salary range check. nothing surprising.

technical screen (60 min): this is where SQL showed up hard. I got three SQL problems escalating in difficulty. first was a simple aggregation with GROUP BY. second involved window functions, specifically LAG() to compute period-over-period changes. third was a self-join to find something like "customers who bought product A but not product B within 30 days." medium-hard leetcode SQL equivalent. I also got a stats question: given two groups with different sample sizes, walk me through how you'd test if their conversion rates are significantly different. they wanted me to name the test, explain the assumptions, and talk about what happens when sample size is small.

case study (60 min): they gave me a business scenario. a retail client wants to reduce churn. I had to talk through what data I'd want, how I'd define the target variable, feature ideas, model choice (I said gradient boosting, they asked why not logistic regression first), and how I'd measure success in deployment. this felt like a consulting case crossed with a DS interview. they pushed on tradeoffs a lot.

behavioral + stakeholder (45 min): all STAR format. questions like "tell me about a time you had to explain a model result to a non-technical executive" and "describe a project that didn't go as planned." one question I didn't expect: "how do you handle a client who disagrees with your analysis?" classic consulting angle.

no ML theory deep-dives (no backprop, no math-heavy stuff). the SQL and the case were definitely the highest-weight rounds based on debrief feedback.

offer was for a consultant-level DS role, not a pure "data scientist" title. pay structure is a bit different from product companies -- more base-heavy, smaller variable. if you're coming from a startup or big-tech DS role, expect that adjustment.

happy to answer follow-ups.

4 replies

analyst_ana

thank you for this. the SQL difficulty level is really helpful to know. did they give you a coding environment or was it whiteboard-style?

ds_dmitri

shared Google Doc, no IDE. they said they care more about approach than syntax but I'd still practice writing clean SQL without autocomplete. the window function one tripped me up briefly because I had a typo I didn't catch.

consultant_cam

the case format you described is very typical for Accenture's Applied Intelligence practice. they're staffing consultants who happen to know data science, not pure researchers. the "client disagrees with your analysis" question is a tell. they want to see if you can hold your ground diplomatically, not just fold or get defensive.

ml_mike

this matches what I heard from someone who did the Chicago office loop last year. SQL-heavy, case-heavy, relatively light on pure ML theory. makes sense for consulting work where you're building dashboards and presenting findings, not training transformers.