Google’s Senior Data Consultant role bridges the gap between raw data and strategic business decisions. Unlike pure analyst roles, you’ll be expected to translate complex analytical findings into executive narratives, design measurement frameworks for business initiatives, and partner with product and engineering teams to embed data-driven thinking into every decision. Interviews emphasize SQL fluency, stakeholder communication, business acumen, and the ability to handle ambiguous, open-ended problems.
This guide covers the most frequently asked questions across SQL, business case analysis, measurement design, and consulting communication — all based on real interview reports from Google’s gTech and Business Intelligence teams.
📚 Recommended Resources to Crack This Interview
2200 Most Asked Analytics Interview Questions
- 2,200+ questions · 23 topics
- For: those who want to master all analytics interview topics
Ace Any SQL Interview
- 220+ questions · SQL advanced, data models, DBMS
- For: anyone who wants to crack any SQL round (fresher to advanced)
Round 1 — SQL & Data Manipulation (BigQuery)
Google’s consulting SQL round focuses on writing production-quality BigQuery queries that answer business questions. Expect 3–4 questions in 45 minutes. Interviewers care about correctness, edge case handling, and code readability.
Question 1: Advertiser Spend Efficiency Analysis
You have a table ad_campaigns (campaign_id, advertiser_id, date, spend, clicks, conversions). Write a query to find, for each advertiser, the monthly spend, total conversions, cost-per-conversion (CPC), and rank advertisers within each month by CPC (ascending — lower CPC is better). Return only advertisers whose average monthly CPC is below $50.
WITH monthly_stats AS (
SELECT
advertiser_id,
DATE_TRUNC(date, MONTH) AS month,
SUM(spend) AS total_spend,
SUM(conversions) AS total_conversions,
SAFE_DIVIDE(SUM(spend), SUM(conversions)) AS cpc
FROM ad_campaigns
GROUP BY 1, 2
),
ranked AS (
SELECT
*,
RANK() OVER (PARTITION BY month ORDER BY cpc ASC) AS cpc_rank,
AVG(cpc) OVER (PARTITION BY advertiser_id) AS avg_monthly_cpc
FROM monthly_stats
)
SELECT
advertiser_id,
month,
total_spend,
total_conversions,
ROUND(cpc, 2) AS cost_per_conversion,
cpc_rank
FROM ranked
WHERE avg_monthly_cpc < 50
ORDER BY month, cpc_rank;
SAFE_DIVIDE to handle zero conversions gracefully — a bare division would NULL out on 0 conversions. The AVG(cpc) OVER (PARTITION BY advertiser_id) computes across all months for the HAVING-equivalent filter without a second CTE.- → How would you weight the average by spend volume instead of treating each month equally?
- → If an advertiser has months with zero conversions, should they be excluded from the ranking or ranked last?
- → Add a 3-month rolling average CPC column.
Question 2: Business Segment Revenue Attribution
You have contracts (contract_id, client_id, segment, start_date, end_date, contract_value) and invoices (invoice_id, contract_id, invoice_date, amount_paid). Write a query to show, for each business segment, the total contracted value vs total collected revenue in 2025, and the collection rate. Flag segments with collection rate below 80%.
WITH contracted AS (
SELECT
segment,
SUM(contract_value) AS total_contracted
FROM contracts
WHERE start_date <= '2025-12-31'
AND end_date >= '2025-01-01'
GROUP BY segment
),
collected AS (
SELECT
c.segment,
SUM(i.amount_paid) AS total_collected
FROM invoices i
JOIN contracts c USING (contract_id)
WHERE EXTRACT(YEAR FROM i.invoice_date) = 2025
GROUP BY c.segment
)
SELECT
ct.segment,
ct.total_contracted,
COALESCE(cl.total_collected, 0) AS total_collected,
ROUND(SAFE_DIVIDE(COALESCE(cl.total_collected,0), ct.total_contracted) * 100, 1) AS collection_rate_pct,
IF(SAFE_DIVIDE(COALESCE(cl.total_collected,0), ct.total_contracted) < 0.8,
'AT RISK', 'HEALTHY') AS status
FROM contracted ct
LEFT JOIN collected cl USING (segment)
ORDER BY collection_rate_pct ASC;
COALESCE handles segments with no invoices yet (collection = 0, not NULL).- → How do you handle contracts with multiple amendments (changes to contract_value mid-year)?
- → Which segment would you prioritize for a collection improvement initiative, and why?
Question 3: Support Ticket Escalation Patterns
You have support_tickets (ticket_id, client_id, created_at, resolved_at, escalated, priority). Find clients whose escalation rate (escalated tickets / total tickets) increased by more than 10 percentage points from Q4 2024 to Q1 2025. Return client_id, Q4 rate, Q1 rate, and the delta.
WITH quarterly AS (
SELECT
client_id,
CASE
WHEN created_at BETWEEN '2024-10-01' AND '2024-12-31' THEN 'Q4_2024'
WHEN created_at BETWEEN '2025-01-01' AND '2025-03-31' THEN 'Q1_2025'
END AS quarter,
COUNTIF(escalated = TRUE) AS escalated_count,
COUNT(*) AS total_count,
SAFE_DIVIDE(COUNTIF(escalated = TRUE), COUNT(*)) AS escalation_rate
FROM support_tickets
WHERE created_at >= '2024-10-01'
AND created_at < '2025-04-01'
GROUP BY 1, 2
)
SELECT
client_id,
ROUND(MAX(IF(quarter='Q4_2024', escalation_rate, NULL)) * 100, 1) AS q4_rate_pct,
ROUND(MAX(IF(quarter='Q1_2025', escalation_rate, NULL)) * 100, 1) AS q1_rate_pct,
ROUND((MAX(IF(quarter='Q1_2025', escalation_rate, NULL))
- MAX(IF(quarter='Q4_2024', escalation_rate, NULL))) * 100, 1) AS delta_pct
FROM quarterly
WHERE quarter IS NOT NULL
GROUP BY client_id
HAVING delta_pct > 10
ORDER BY delta_pct DESC;
COUNTIF is BigQuery-native and cleaner than SUM(CASE WHEN escalated THEN 1 ELSE 0 END). The MAX(IF(...)) pivot pattern is standard BigQuery — interviewers expect you to know it. The HAVING delta_pct > 10 filter applies after the pivot aggregation.Round 2 — Business Case & Consulting Communication
This round tests your ability to structure ambiguous problems, define success metrics, and communicate findings to a non-technical executive audience. Google interviewers look for MECE frameworks, proactive clarification, and concise storytelling.
Question 1: Declining Client Engagement
A Google Cloud enterprise client tells you: “Our data team’s reports are not being used by business stakeholders anymore.” How would you diagnose and solve this problem? Walk through your full consulting approach.
Step 1 — Clarify before structuring. Ask: What does “not being used” mean — are reports not opened, not acted upon, or causing wrong decisions? Which stakeholders? Which reports? When did this start?
Step 2 — Hypothesis tree (MECE).
- Discovery problem: Stakeholders don’t know the reports exist or can’t find them easily (distribution, portal usability).
- Relevance problem: Reports answer the wrong questions — built for old priorities, not current business decisions.
- Trust problem: Past data quality issues eroded confidence — stakeholders stopped trusting numbers.
- Usability problem: Reports are too complex, not actionable — 50-slide decks instead of 1-page summaries with clear recommendations.
- Competing solution: Stakeholders built their own shadow analytics in Excel/Sheets and no longer need the formal reports.
Step 3 — Diagnostic plan. Run a 2-week discovery: usage analytics from the BI platform (who opened what, when), 5 stakeholder interviews (decision-makers, not analysts), and a content audit mapping each report to a current business decision.
Step 4 — Recommendation structure. Present findings as: Here’s what we found, here’s why it matters, here’s the 30-60-90 day fix. Prioritize quick wins (rename reports, add executive summary pages) before structural changes (rebuild data products around decision flows).
- → Stakeholders say they trust the data but still don’t use the reports. What does that tell you?
- → How do you measure whether your intervention actually improved report adoption?
- → What if the real problem is that the data team’s charter is misaligned with business priorities?
Question 2: North-Star Metric Design for a B2B SaaS Client
A Google Cloud consulting client runs a B2B SaaS project management tool. Their CEO asks you: “What one metric should we obsess over?” Design a north-star metric framework for them. Include the metric, why it’s the right one, its limitations, and how you’d track it.
Recommended North-Star Metric: Weekly Active Teams (WAT) — the number of teams (not individual users) that complete at least 3 collaborative actions (task assignments, comments, or status updates) in a 7-day window.
Why this metric:
- Team-level, not user-level: B2B value is delivered to teams, not individuals. A power user on an inactive team means nothing for retention or expansion.
- Collaborative actions only: Solo task creation doesn’t prove team adoption. 3 collaborative actions signals genuine workflow integration.
- Weekly cadence: Project management tools are used weekly, not daily — daily metrics would show artificial churn on weekends.
- Predictive of revenue: Teams with WAT > 2 consecutive weeks have 3× higher renewal rates (typical benchmark to validate with client data).
Limitations to be transparent about:
- WAT can be inflated by “demo teams” set up for procurement — filter to teams with ≥3 real users and ≥30 days post-onboarding.
- A team with 1 active user creating tasks for others inflates WAT without true adoption.
Tracking plan: Track WAT weekly in a Looker/BigQuery dashboard with cohort breakdown by team size, industry, and customer success rep. Set leading indicator threshold: if a team misses WAT for 2 consecutive weeks, trigger a CS outreach automatically.
- → The CEO pushes back: “We should track revenue, not engagement.” How do you respond?
- → How do you connect WAT to NRR (Net Revenue Retention)?
- → What counter-metrics would you track alongside WAT to prevent gaming?
Question 3: Executive Presentation Scenario
You analyzed 6 months of data and found that a client’s marketing spend has a negative ROI in 3 of their 5 channels. You have 10 minutes with the CMO. How do you structure your presentation and what do you NOT say?
Structure (Pyramid Principle — conclusion first):
- Lead with the insight (60 seconds): “3 of your 5 channels are destroying value. We recommend reallocating 40% of your budget to the 2 high-performing channels by Q3 — our model shows this adds ₹X in incremental revenue.”
- Evidence (3 minutes): One slide with channel-level ROI table. Use a simple red/green heatmap — no data dump. Show confidence intervals so they understand it’s not a point estimate.
- Recommendation with trade-offs (3 minutes): What to cut, what to scale, and what risks to acknowledge (channel cannibalization, seasonality adjustments, data lag).
- Clear ask (1 minute): “We need your sign-off to proceed with the reallocation model — can we align on this today?”
What NOT to say:
- Don’t say “the data shows” 8 times — say “we found” or “the finding is.”
- Don’t present methodology in the room — have it in an appendix if asked.
- Don’t hedge every point with caveats — one caveat slide in the appendix is enough.
- Don’t end with “any questions?” — end with a clear recommended next step.
Round 3 — Measurement Design & Experimentation
Senior consultants at Google are expected to design measurement frameworks for client initiatives — not just run analyses but define what success looks like, how to measure it, and how to communicate it to stakeholders who distrust statistics.
Question 1: Designing a Measurement Framework for a CX Initiative
A Google Cloud client is launching a new customer onboarding experience for their enterprise SaaS product. They want to know if it’s working. Design a full measurement framework: success metrics, leading/lagging indicators, data sources, and evaluation timeline.
Framework: Input → Process → Output → Outcome hierarchy
Leading Indicators (measurable in week 1–4):
- Onboarding completion rate: % of new users who complete all onboarding steps within 14 days
- Time-to-first-value: days from account creation to first meaningful action (first project created, first team member invited)
- Setup step abandonment rate: where users drop off in the onboarding flow
Lagging Indicators (measurable in month 2–6):
- 30-day retention rate by onboarding cohort
- 90-day NRR (Net Revenue Retention) — are customers expanding or churning?
- Support ticket volume in first 60 days — lower is better
- NPS at 60-day check-in
Data sources: Product event logs (Mixpanel/Amplitude), CRM (Salesforce), Support ticketing (Zendesk), Billing system for NRR
Evaluation design: If possible, A/B test the new vs old onboarding (50/50 split for 8 weeks). If not (too risky to expose 50% to old flow), use a difference-in-differences approach comparing cohorts before/after launch with matched controls from a similar customer segment.
Reporting cadence: Weekly leading indicators for the first 2 months → monthly lagging indicator review for months 3–6 → final 6-month readout with recommendation on full rollout or iteration.
- → The CMO wants to declare success after week 2. How do you handle this?
- → Onboarding completion is up 20% but 90-day retention didn’t change. What does that mean?
- → How do you control for the fact that the new onboarding launched during your client’s busy season?
Question 2: Attribution Modeling for B2B Sales
A client’s sales cycle is 6–18 months long, with 8–12 touchpoints across marketing, SDR outreach, demos, and proposals. Their current CRM attributes 100% of closed revenue to the “last demo.” Why is this wrong, and how would you design a better attribution model? What data would you need?
Why last-demo attribution is wrong:
- It systematically undervalues top-of-funnel activities (content, events, SDR outreach) that created awareness and intent.
- It creates perverse incentives — marketing stops investing in demand generation because they get no credit.
- With 6–18 month cycles, the “last demo” is often just the final confirmation of a decision made 6 months earlier.
Better attribution options (in order of sophistication):
- Linear: Distribute credit equally across all touchpoints. Simple, fair baseline. Better than last-touch but treats all touchpoints as equal.
- Time-decay: Weight recent touchpoints more heavily. Reasonable for short sales cycles. Less appropriate for 18-month B2B deals.
- Position-based (U-shaped): 40% to first touch, 40% to last touch (deal close), 20% split among middle touchpoints. Good for B2B where the first meeting and the final proposal are most influential.
- Data-driven (Shapley value): Use machine learning to assign credit based on the actual incremental contribution of each touchpoint to deal closure probability. Requires 500+ deals to train reliably.
Data needed: Full touchpoint sequence per opportunity (CRM + marketing automation), timestamps, channel, deal value at close, deal outcome (won/lost), and at least 12–18 months of historical closed deals.
Recommendation: Start with position-based attribution (implementable in 2 weeks with existing CRM data). Build toward Shapley value over 6 months once you have enough clean deal history.
- → The sales team pushes back: “Demos close deals, that’s just reality.” How do you respond?
- → How do you handle touchpoints from the same channel at different stages of the funnel?
4-Week Prep Plan for Google Senior Data Consultant
Build your SQL foundation (window functions, CTEs, BigQuery-specific syntax), refresh consulting frameworks (MECE, Pyramid Principle, STAR method), and read TDM’s interview pattern posts on analytics consulting and measurement design.
Work through the 2200 Most Asked Analytics Interview Questions ebook — cover SQL, business case, and metrics design sections. At the end of Week 2, book a mock interview at topmate.io/nitin_kamal to test your case structuring under time pressure.
Complete the Ace Any SQL Interview ebook (link) — cover all 220 questions without shortcuts. Focus on BigQuery syntax: COUNTIF, SAFE_DIVIDE, DATE_TRUNC, MAX(IF()) pivots. Book second mock interview at the end of Week 3.
Practice presenting analytical findings out loud. Run timed case simulations: 20 min to structure a business case, 15 min to write a SQL query, 10 min to build an executive narrative. Book final mock interview at the end of Week 4.
Target: ~950 questions + 3 mock interviews = ready for any Google consulting interview.