Data Analyst Interview Questions in 2026: SQL, Cases & Dashboards

Data analyst interviews have gotten harder, more structured, and more product-aware over the last two years. Companies are no longer satisfied with candidates who can write a GROUP BY clause—they want analysts who can frame a business problem, build a defensible metric, and explain a dashboard to a VP without flinching. This guide covers every major question category you'll face in 2026, with real example questions, model answers, and the traps that eliminate most candidates. Whether you're targeting a $90K analyst role at a mid-size startup or a $160K+ senior position at a FAANG-adjacent company, the same fundamentals apply.

SQL Is Still the Filter Round—and It's Getting Harder

SQL remains the single most common first-round screen, and interviewers have stopped asking simple joins. In 2026, expect window functions, self-joins, and query optimization questions in even entry-level panels.

Here are the SQL topic areas ranked by how often they appear:

1. Window functions — ROW_NUMBER(), RANK(), LAG(), LEAD(), running totals with SUM() OVER ()
2. Aggregations with complex GROUP BY and HAVING clauses
3. Self-joins for cohort and retention analysis
4. CTEs vs. subqueries — when to use each and why
5. Query performance — indexes, execution plans, avoiding full table scans
6. Date/time manipulation — cohort calculations, 30/60/90-day windows
7. NULL handling — COALESCE, NULLIF, and how NULLs propagate in aggregations

A real example question: "Write a query to find users who made a purchase in their first 30 days but never purchased again after that."

This tests self-joins, date arithmetic, and NOT EXISTS logic simultaneously. Most candidates either over-complicate the join structure or forget to handle users with exactly one purchase. The clean answer uses a CTE to identify first-purchase dates, then a left join to find any subsequent purchases, filtering for NULLs on the second side.

Another common trap: "Which of these two queries is faster, and why?" where both queries return identical results but one uses a correlated subquery and one uses a window function. You need to know that correlated subqueries execute once per row—death on large tables—and window functions process the full set once.

Practice platforms worth your time: StrataScratch, DataLemur, and LeetCode's database section. Aim for 40–50 medium-difficulty problems before your first loop. Don't grind easy problems—they won't appear in 2026 interviews.

Business Cases Separate Analysts from Query Monkeys

Every serious company now runs at least one business case or analytical scenario round. This is where most technical candidates fall apart, because the question sounds open-ended and collaborative but is actually evaluating a specific framework.

"The best analysts don't just answer the question asked—they reframe it into the question that actually matters for the business."

The standard business case structure interviewers expect:

• Clarify the goal — What decision does this analysis support? Who is the stakeholder?
• Define success metrics — Primary metric, guardrail metrics, and why you chose them
• Identify the data — What tables/sources would you use? What are the known limitations?
• Describe the analysis — High-level methodology, segmentation approach, statistical considerations
• Anticipate confounders — Seasonality, selection bias, survivorship bias
• Recommend and caveat — A clear recommendation with explicit assumptions

Example question: "Our checkout conversion rate dropped 8% last Tuesday. Walk me through how you'd investigate."

Weak answer: "I'd look at the data and find what changed."

Strong answer: Start with scope — is this 8% relative or absolute? Is it sitewide or specific to a funnel step, device type, geography, or user segment? Then form hypotheses in priority order: product change or experiment rollout, infrastructure issue, external factor (competitor promotion, news event), data pipeline issue. Describe exactly which tables you'd query and what the breakdowns would look like. Finish with: "Before I conclude it's a real business problem, I'd verify the logging isn't broken."

Companies like Stripe, Shopify, and Airbnb run case interviews that are nearly indistinguishable from product sense interviews at PM-level. Prepare accordingly.

Metric Design Is the New Differentiator

In 2026, defining and defending metrics is a standalone interview category at top companies. They'll give you a product scenario and ask you to design a measurement framework from scratch.

Example question: "How would you measure the success of a new onboarding flow?"

Most candidates go straight to conversion rate. That's necessary but not sufficient. A complete answer covers:

• Primary metric: Completion rate of the onboarding flow
• North Star proxy: Time-to-first-value (the first moment a user gets tangible benefit from the product)
• Leading indicators: Step-by-step drop-off rates, time spent per step
• Guardrail metrics: Overall D7 retention shouldn't decrease; support ticket volume shouldn't increase
• Segmentation plan: New vs. returning users, mobile vs. desktop, acquisition channel

Also expect questions like "How would you detect if a metric is being gamed?" and "If your north star metric goes up but revenue goes down, what do you do?" These test whether you understand the difference between correlation and causation in your own measurement systems.

Dashboard and Data Visualization Questions Are Underrated

Candidates chronically underprepare for the visualization and dashboard design portion. Companies running Tableau, Looker, Power BI, or even internal tooling will ask you to either critique an existing dashboard or design one from scratch.

What they're actually testing:

• Do you understand who the audience is and what decision the dashboard serves?
• Can you distinguish between operational dashboards (real-time, anomaly detection) and strategic dashboards (trends, executive reporting)?
• Do you default to the right chart type, or do you reach for pie charts when you shouldn't?
• Can you articulate data freshness tradeoffs — hourly refresh vs. daily batch?

The most common mistake: Designing a dashboard with 15 charts because "more data = more insight." Interviewers at data-mature companies will push back hard on this. A strong answer names three to five KPIs maximum, explains why each one drives a decision, and acknowledges what gets cut and why.

If you're interviewing at a company with a known BI stack (easy to find on job descriptions or LinkedIn), spend time in that tool before your interview. A Looker-specific question about derived tables or persistent derived tables can catch you off guard if you've only used Tableau.

Statistics and Experimentation Questions Are Back in Full Force

The A/B testing question never truly left, but in 2026, interviewers have raised the floor significantly. You cannot get through a senior analyst loop at a product-led company without solid experiment design knowledge.

Core topics you must be able to discuss fluently:

• Statistical significance vs. practical significance — a result can be significant at p < 0.05 and still not be worth shipping
• Sample size calculation — minimum detectable effect, baseline conversion rate, power (typically 80%), and why underpowered tests are dangerous
• Novelty effects — why week-one experiment results are often misleading
• Network effects and interference — when user behavior in control affects treatment (common in social or marketplace products)
• Multiple testing problem — running 10 variants simultaneously inflates Type I error
• Metric sensitivity — why you sometimes need to transform metrics (log scale for revenue) to reduce variance

Example question: "You ran an A/B test and got p = 0.04. Your stakeholder wants to ship immediately. What do you say?"

The answer isn't "yes, ship it" and it isn't "wait for more data" as a reflexive hedge. It's: check the sample size against your pre-registered power calculation, check that you didn't peek at results early (which inflates false positive rate), confirm the effect size is practically meaningful for the cost of the change, and verify there are no guardrail metric regressions. Then make a clear call.

Salary Benchmarks and Role Tiers in 2026

Know your market before you walk into an offer negotiation. Analyst comp has compressed slightly from the 2021–2022 peak but remains strong at senior levels.

United States (USD, base salary):

• Entry-level Data Analyst (0–2 years): $75,000–$100,000
• Mid-level Data Analyst (2–5 years): $100,000–$135,000
• Senior Data Analyst (5+ years): $130,000–$170,000
• Staff / Lead Analyst or Analytics Engineer: $160,000–$200,000+

Canada (CAD, base salary, major metros):

• Entry-level: $65,000–$85,000
• Mid-level: $85,000–$115,000
• Senior: $115,000–$150,000

FAANG and FAANG-adjacent companies (Google, Meta, Amazon, Stripe, Airbnb) pay 20–40% above these bands in total compensation once RSUs and bonuses are included. Startups often pay below band and compensate with equity—model that equity conservatively.

If a recruiter asks for your salary expectations before you have an offer, deflect once: "I'm focused on fit right now—I'm confident we can align on compensation if we get there." If they push again, give a range anchored to the top of the market band for your level.

Behavioral Questions at the Analyst Level Are About Influence, Not Just Execution

Analyst behavioral questions aren't just "tell me about a time you cleaned a messy dataset." At mid and senior levels, interviewers want evidence that you've influenced decisions, pushed back on bad analysis, and moved stakeholders with data.

The STAR format (Situation, Task, Action, Result) is fine as a skeleton, but the differentiator is in the Action and Result. Weak results sound like: "The stakeholder appreciated my work." Strong results are quantified and business-connected: "The analysis led the team to deprioritize the feature, saving approximately 3 engineer-months that were redirected to a higher-impact project."

Prepare stories for each of these scenarios:

• A time you disagreed with a stakeholder's interpretation of data and what happened
• A time your analysis was wrong and how you caught it (or how someone else caught it)
• A time you simplified a complex finding for a non-technical audience
• A time you had to make a recommendation with incomplete data
• A time you proactively identified a business problem no one had asked you to investigate

Don't have five stories that are all about being right. One story where you were wrong and handled it well is more credible than five stories where you were the hero.

Next Steps

If your first interview is in the next two to four weeks, here is exactly what to do:

1. Solve 20 SQL problems on DataLemur this week — prioritize window functions and retention analysis patterns. Do them without hints. Review the editorial only after you've submitted.
2. Record yourself answering one business case question out loud — use the framework from this guide. Watch it back. If you hear yourself say "it depends" without following it with a concrete answer, redo it.
3. Pick one dashboard you use in your current role (or a public one on Tableau Public) and critique it in writing — what three things would you change and why? This sharpens your visualization instincts faster than any course.
4. Nail down your salary number before any recruiter call — use Levels.fyi, LinkedIn Salary, and the bands in this guide to set your anchor. Know your walk-away number.
5. Prepare your five behavioral stories and write them down — not as scripts, but as bullet-point outlines you can reconstruct naturally in conversation. Practice them with a friend or into a voice memo.

Data analyst interviews reward candidates who are specific, structured, and honest about uncertainty. The analysts who get offers aren't the ones who know the most SQL tricks—they're the ones who make interviewers feel confident that they'll make good decisions with ambiguous data under real business pressure.

Related guides

• Data Scientist Interview Questions in 2026: Stats, SQL & Cases — Crack DS interviews in 2026 with this no-fluff guide to statistics, SQL, and case study questions—plus concrete prep steps for the week ahead.
• Data Engineer Interview Questions — Pipelines, SQL Optimization, and Warehouse Design — Data engineer interviews test practical judgment: modeling data, moving it reliably, optimizing SQL, and designing warehouses that analysts and products can trust. This guide covers the 2026 questions, answer patterns, and senior-level signals.
• Data Modeling Mock Interview Questions in 2026 — Practice Prompts, Answer Structure, and Scoring Rubric — A 2026 data modeling mock interview guide with schema prompts, relationship modeling, tradeoff examples, scoring rubric, drills, and a 7-day prep plan.
• SQL Mock Interview Questions in 2026 — Practice Prompts, Answer Structure, and Scoring Rubric — Prepare for SQL interviews with realistic 2026 prompts, clean answer structure, scoring criteria, and worked query patterns for analytics, product, marketplace, and data roles.
• Android Engineer Interview Questions in 2026 — Kotlin, Jetpack Compose, and Android System Design — Android interviews in 2026 test Kotlin, coroutines, Jetpack Compose, lifecycle, offline behavior, and release judgment. This guide gives the questions and answer patterns that show native Android production maturity.