Question 1
Difficulty: medium
Can you walk me through how you would design a new Looker model for a business team that needs daily KPI reporting?
Sample answer
I’d start by clarifying the business questions first, because a good Looker model should reflect how the team actually makes decisions. I would identify the core KPIs, define the grain of the base tables, and confirm the source of truth for each metric before writing any LookML. From there, I’d build a clean Explore around a well-structured view, using persistent derived tables only where they solve a real performance or logic problem. I’d be careful about naming conventions, join relationships, and measure definitions so the model stays readable for both analysts and business users. I also like to include description fields and grouping logic so the content is easy to adopt. Once the model is built, I’d validate the numbers against known reports, then iterate with the stakeholders until the dashboard outputs match their operational needs and performance expectations.
Question 2
Difficulty: medium
How do you handle defining measures in LookML when different teams disagree on the calculation logic?
Sample answer
That happens a lot, especially when teams use the same metric name but mean slightly different things. My first step is to slow things down and get everyone aligned on the business definition, not the technical implementation. I’ll usually compare existing reports, documentation, and source data to pinpoint where the differences come from. If there are legitimate variants, I prefer to model them explicitly rather than force one definition into everything. For example, I might create a base measure for the canonical version and separate derived measures for departmental needs, with clear labels and descriptions. I also document assumptions right in the LookML so future developers understand why the metric was built that way. If needed, I’ll involve finance, operations, or product owners to make the final call. My goal is always consistency, transparency, and trust in the data, even when the answer is not the simplest one.
Question 3
Difficulty: hard
What is your approach to optimizing slow-running Looker Explores or dashboards?
Sample answer
I usually start by figuring out whether the issue is in Looker, the database, or the data model itself. I look at the SQL generated by Looker, check query patterns, and see whether the joins, filters, or table structure are causing unnecessary work. If I find that the Explore is scanning too much data, I’ll look at partitioning, clustering, summary tables, or PDTs depending on the warehouse. I also pay close attention to join order, symmetric aggregates, and whether a join is actually needed for the user’s common workflow. On the Looker side, I’ll reduce overly complex table calculations or unnecessary dashboard tiles if they’re slowing things down. I also like to test performance with realistic filters, since a dashboard that is fast on one filter set may be slow in real use. The key for me is to optimize with evidence, not guesses.
Question 4
Difficulty: medium
Describe a time when you had to troubleshoot a data discrepancy between Looker and another reporting tool.
Sample answer
I’ve dealt with this kind of issue by treating it like a structured investigation rather than jumping to conclusions. First, I confirm whether both tools are pulling from the same source tables and using the same filters, date logic, and grain. In one situation, the discrepancy turned out to be caused by one report counting distinct users while the other was counting rows after a join that duplicated records. I recreated both outputs in SQL, compared the intermediate steps, and isolated the exact point where the numbers diverged. After that, I documented the root cause and updated the LookML so the logic was consistent and easier to explain to stakeholders. I also walked the business team through the difference in plain language so they understood why the numbers had differed. I think the most important part is not just fixing the issue, but making the explanation durable so the same confusion doesn’t return later.
Question 5
Difficulty: hard
How do you decide when to use a persistent derived table versus a standard view in Looker?
Sample answer
I use a persistent derived table when it solves a meaningful problem, usually performance, reuse, or complex transformation logic that would otherwise be repeated in many Explores. If the dataset is large, expensive to query, or requires logic that should be standardized once and reused many times, a PDT can be a strong choice. That said, I try not to overuse them because they add maintenance overhead and can complicate freshness management. If the transformation is lightweight and the database can handle it efficiently, I prefer a standard view or a simpler SQL derived table. I also consider how often the underlying data changes, whether the business needs near-real-time results, and whether the team has a clear process for monitoring rebuilds. My goal is to choose the simplest architecture that still gives users fast, reliable numbers. I’d rather keep the model maintainable than introduce unnecessary complexity for marginal performance gains.
Question 6
Difficulty: easy
How would you support a business user who wants to build their own dashboard but keeps creating inconsistent metrics?
Sample answer
I’d approach that as both a support and enablement opportunity. Instead of taking over their work, I’d help them understand which fields are safe to use and which metrics need to be standardized. I’d review the dashboard with them, point out where inconsistencies are coming from, and explain the logic in a way that connects to their business outcome. If they’re using multiple similar measures, I would simplify the Explore or create curated fields so the right choice is more obvious. I also think templates, saved Looks, and well-named folders can reduce mistakes significantly. When someone is self-serving analytics, guardrails matter as much as flexibility. I’ve found that a short training session plus clear documentation often prevents repeated issues. My goal is to make them more successful without turning every request into a dependency on the BI team. That balance is important in a healthy Looker environment.
Question 7
Difficulty: medium
How do you validate that a new Looker dashboard is accurate before it goes live?
Sample answer
I validate in layers rather than relying on one quick check. First, I confirm that the underlying LookML fields map correctly to the source system and that the joins behave as intended. Then I compare a handful of core metrics against trusted SQL queries or existing reconciled reports, ideally across different date ranges and filters. I also test edge cases, such as null values, missing records, duplicated records, and unusual filter combinations, because those are often where problems show up. If the dashboard includes business logic like rolling windows or attribution rules, I make sure those calculations are tested with known examples. I like to involve the stakeholder in the final review so they can confirm the output matches the real-world use case. Before launch, I also check labels, descriptions, ordering, and dashboard performance. Accuracy matters most, but usability and clarity are part of a good release too.
Question 8
Difficulty: hard
Tell me about a time you had to balance strict governance with the need for self-service analytics.
Sample answer
I’ve found that governance works best when it enables self-service instead of blocking it. In one project, multiple teams needed access to customer data, but there were concerns around inconsistent logic and sensitive fields. I helped create a curated Looker model with approved dimensions and measures, while keeping raw or sensitive fields out of the default Explore experience. For power users, we documented a controlled path to request additional access when there was a legitimate business need. I also added clear field descriptions and naming standards so users could trust what they were selecting. That approach reduced confusion without taking away flexibility. The key was to make the governed path feel useful, not restrictive. When users see that the curated model is actually easier and more reliable than building everything themselves, adoption improves naturally. I think good BI governance should protect the data while still helping the business move quickly.
Question 9
Difficulty: medium
What steps do you take when a Looker dashboard needs to be redesigned for executives who want fewer metrics and faster access to insights?
Sample answer
For executives, I focus on signal over detail. I would start by identifying the handful of metrics that actually drive decisions at that level and remove anything that is operational noise. Then I’d redesign the dashboard so the most important trends, exceptions, and comparisons are visible within seconds. I would likely use concise titles, conditional formatting, and a logical visual hierarchy so the story is obvious at a glance. I’d also check performance carefully because executive dashboards need to feel responsive. That may mean simplifying explores, limiting unnecessary dimensions, or pre-aggregating data where appropriate. I usually keep drill paths available so leaders can go deeper if they want, but the default view should be clean and focused. I’d test the design with a few actual users and adjust based on their feedback. In my experience, executive reporting works best when it is simple, fast, and clearly tied to outcomes.
Question 10
Difficulty: medium
How do you stay effective when business requirements keep changing during a Looker implementation?
Sample answer
I try to build in adaptability from the beginning, because changing requirements are part of the job rather than an exception. I start with a clear version of the business problem, but I avoid locking everything down too early if the team is still exploring what they need. I use modular LookML, reusable dimensions and measures, and documentation that makes it easy to adjust without breaking the whole model. When requirements change, I ask whether the change is a true business shift, a clarification of the original need, or just a preference for presentation. That helps me respond appropriately instead of endlessly reworking the solution. I also keep stakeholders updated so they understand tradeoffs in scope, timing, and impact. Good communication matters just as much as technical skill here. The best implementations I’ve worked on were the ones where we stayed flexible but still disciplined about definitions, testing, and release management.
