Engineering manager domain knowledge interview questions

As asked

You are onboarding a backend engineer who has never used Datadog before. Explain the difference between metrics, distributed traces, and logs, and give a concrete example of when you would reach for each one to debug a latency problem in a microservices API.

Sample answer outline

A strong answer explains that metrics are numeric aggregations good for alerting on trends, traces capture the causal chain across service boundaries to find which hop is slow, and logs carry the raw event detail for understanding why a specific request failed. The debugging example should show the three being used in sequence: alert from a p99 latency metric, trace to find the slow downstream call, log on that service to read the specific error.

Expect these follow-ups

• How does Datadog's APM connect traces to the underlying infrastructure metrics?
• What is cardinality and why does it matter when choosing what to put in a metric tag?

As asked

Most engineering orgs have a leveling framework. What is it supposed to achieve, how do you use it in practice for promotions and calibration, and where does it break down?

Sample answer outline

Strong candidates describe the dual purpose of leveling frameworks: clarity of expectations and calibration across managers. Practical use includes using them in promotion write-ups as evidence against each dimension. Failure modes include treating levels as checklists, over-indexing on tenure, and company-specific level definitions that do not translate across organizations.

Expect these follow-ups

• How do you handle a report who is performing at level on most dimensions but below on one?
• What do you do when your company does not have a leveling framework?

As asked

Explain the difference between an SLA, an SLO, and an SLI. As an engineering manager, not an SRE, when and how do you engage with these concepts in your team's work?

Sample answer outline

Strong candidates accurately define each: SLI is the measured metric, SLO is the target, SLA is the contractual commitment. As an EM, they engage when setting reliability objectives for a service, defending the error budget concept to product when reliability investments are undervalued, and using SLO burn rate as an input to prioritization.

Expect these follow-ups

• How do you convince a product manager that an SLO breach should pause feature work?
• What happens when an SLO is set too tightly and the team burns out chasing it?

As asked

What are the four DORA metrics, what does each measure, and how do you use them in practice to assess and improve your team's delivery performance?

Sample answer outline

Strong candidates accurately name deployment frequency, lead time for changes, change failure rate, and mean time to restore. They describe how to baseline each metric, identify the constraint that is keeping the team in a lower performance bucket, and use them as inputs to improvement investments rather than targets to game. They also acknowledge that elite performance on all four simultaneously is rare.

Expect these follow-ups

• What is the most important DORA metric to improve first if your team is in the low performance bucket?
• How do you avoid teams gaming DORA metrics to hit a target?

As asked

Many engineering teams run sprint planning, standups, retrospectives, and reviews as ritual rather than with purpose. For each ceremony, what is its real purpose, and when is it appropriate to eliminate or significantly change it?

Sample answer outline

Strong candidates describe each ceremony's original intent accurately: planning to commit to a shared goal, standup to surface blockers not report status, retrospective to improve process, review to get feedback from stakeholders. They are comfortable recommending eliminating standups for a small senior team that communicates asynchronously, or replacing retrospectives with direct feedback channels when the ritual has become rote.

Expect these follow-ups

• What does a team losing value from a ceremony look like, and how do you diagnose which ceremony to fix?
• How do you handle pressure from leadership to maintain ceremonies that are not helping the team?

As asked

You do not own the CI/CD pipeline directly but your team's delivery speed depends on it. What do you need to understand about continuous integration and continuous deployment to make good decisions as an engineering manager, and what metrics do you monitor?

Sample answer outline

Strong candidates describe the EM's role as consuming and influencing the pipeline rather than building it. They track build time, flaky test rate, deployment frequency, and rollback frequency. They invest in trunk-based development culture, feature flags for safer deploys, and advocate for pipeline improvements to the platform team when build time exceeds a threshold that is costing developer productivity.

Expect these follow-ups

• How long should a CI pipeline take before you invest in speeding it up?
• How do you build a culture of merging frequently when engineers fear breaking the main branch?

As asked

What is the purpose of a skip-level meeting, how often should you run them, and how do you make sure the information you get is accurate rather than filtered by people telling you what they think you want to hear?

Sample answer outline

Strong candidates describe skip-levels as a way to calibrate their understanding of team health and identify blind spots in their direct reports' management. They run them quarterly, make clear that the conversation is not about the direct manager, and ask open-ended questions. They follow up on systemic issues without attributing them to specific individuals.

Expect these follow-ups

• What do you do when a skip-level reveals that a direct manager has a problem they have not disclosed to you?
• How do you ensure skip-levels do not feel like surveillance to your direct reports?

As asked

Amazon popularized disagree and commit as a leadership principle. What does it actually mean, where does it apply, and where does it become a problematic norm that silences legitimate dissent?

Sample answer outline

Strong candidates describe disagree and commit as appropriate for reversible decisions where debate has been had and a choice must be made. It breaks down when used to pressure people into committing to decisions made without their input, when applied to ethical concerns, or when it becomes a way for senior leaders to shut down debate prematurely. Candidates should distinguish it from compliance culture.

Expect these follow-ups

• How do you tell the difference between legitimate disagree-and-commit and coercion?
• What do you do when you have committed to a decision you disagreed with and it is going wrong?

As asked

How do you define the difference between managing and leading, and which is more important in your current role? Can you give an example of a time you were managing when you should have been leading, or vice versa?

Sample answer outline

Strong candidates distinguish management (process, delivery, people operations) from leadership (direction, culture, inspiration, change). Both are needed. They give a concrete example of over-indexing on one at the wrong moment: for example, focusing on sprint hygiene during a period when the team needed a clear long-term direction.

Expect these follow-ups

• How do you recalibrate when you realize you have been in the wrong mode?
• Does the balance shift based on team maturity or company stage?

As asked

Your organization has 200 microservices and engineers frequently do not know who owns a service when responding to an incident. Walk me through how you would use the Datadog Service Catalog to improve ownership clarity, what metadata you would require for every service, and how you would enforce that teams keep it up to date.

Sample answer outline

A strong answer describes defining a service in service.datadog.yaml with team, tier, contacts, and runbook links, enforcing schema validation in CI, linking the Service Catalog to PagerDuty schedules so on-call contact is always current, and using the catalog's SLO and error-budget views to surface reliability health per team. The candidate should mention using catalog completeness metrics to drive adoption rather than just mandating it.

Expect these follow-ups

• How would you handle a service owned by a team that has been disbanded?
• What is the difference between a service and a component in the Datadog Service Catalog schema?

As asked

A team running a payments API wants to adopt SLOs. Walk me through how you would define a meaningful SLO for them, how to implement it in Datadog using monitors and SLO widgets, how to set up error budget burn rate alerts, and how to use the burn rate data to make decisions about engineering velocity versus reliability work.

Sample answer outline

A strong answer identifies a good SLI for payments (success rate of charge API calls, excluding retried duplicates), sets a reasonable SLO target based on historical performance (not aspirational), implements it as a metric-based SLO in Datadog tracking rolling 30-day compliance, configures a fast burn rate alert (5x budget consumption rate over 1 hour triggers page) and a slow burn rate alert (2x over 6 hours triggers ticket), and uses the remaining error budget to decide when to freeze non-critical feature work.

Expect these follow-ups

• What is the difference between a metric-based SLO and a monitor-based SLO in Datadog, and when would you use each?
• How do you handle planned maintenance windows in your SLO calculation?

As asked

It is 2am. You are on call and you get a page: Notion page loads are failing for 30% of users, error rate is climbing, and your Slack is lighting up. Walk me through exactly what you do in the first 30 minutes.

Sample answer outline

A strong answer follows a clear incident response playbook: (1) acknowledge the alert and open an incident channel, (2) assess impact with dashboards (error rate, affected regions, error type distribution), (3) check recent deploys and roll back the most recent one if there is a strong correlation, (4) mitigate before fully diagnosing (if possible, route traffic around the failing component), (5) communicate status to users and stakeholders every 15 minutes. The candidate should discuss how to triage between a Postgres outage, a bad deploy, a network issue, and a third-party provider failure using the right signals (DB connection pool metrics, APM traces, Cloudflare status). A great answer also covers post-incident: a blameless postmortem within 48 hours and tracking action items to completion.

Expect these follow-ups

• How do you decide whether to roll back or push a hotfix?
• The on-call rotation is burning people out. What systemic changes would you propose to fix the underlying reliability issues?

As asked

Your Snowflake bill has grown from $50K to $400K per month in one year and your CFO wants a plan to reduce it. Walk me through how you would analyze where the credits are going and what levers you would pull to reduce cost without hurting performance.

Sample answer outline

Start with ACCOUNT_USAGE.WAREHOUSE_METERING_HISTORY to find which warehouses consume the most credits, and QUERY_HISTORY to find the top credit-consuming queries. Common levers: right-size warehouses (many teams default to LARGE or X-LARGE when SMALL would do), enable auto-suspend at 60 seconds or less, check for always-on warehouses with no traffic overnight, identify long-running queries that could be optimized, audit automatic clustering costs on tables that do not need it, and check storage costs from excessive Time Travel retention settings. Finally, implement resource monitors with hard credit caps per warehouse.

Expect these follow-ups

• How would you implement a chargeback model so each team sees and is accountable for their own Snowflake spend?
• If the main driver of cost is a single ETL pipeline running every hour, what are the options?

As asked

Walk me through the high-level software and control system architecture that enables a Falcon 9 first stage to execute a return-to-launch-site landing. What are the key computational constraints and failure modes in that loop?

Sample answer outline

A strong answer covers the guidance, navigation, and control (GNC) stack: inertial navigation plus GPS fusion, the powered descent guidance algorithm (PDG, derived from fuel-optimal convex optimization), aerosurface and engine throttle actuation, and the real-time constraints that make this a hard problem (millisecond control loops, fault tolerance, limited compute aboard). Candidates should mention that the landing burn must solve a constrained trajectory in near-real-time and that sensor failures must be detected and handled within the loop.

Expect these follow-ups

• What would happen if the onboard computer lost GPS lock 30 seconds before landing?
• How would you test a guidance algorithm change before it ever flew on hardware?

As asked

What is Tuckman's model of team development, and how does it change how you manage a team that is in the storming phase versus one that is in the norming phase?

Sample answer outline

Strong answers describe forming, storming, norming, and performing accurately and show they have used this model practically. In storming they establish clear decision-making processes and address conflicts early. In norming they codify the team's emerging norms formally. In performing they protect team autonomy and reduce manager-as-bottleneck dynamics.

Expect these follow-ups

• What can cause a high-performing team to regress into storming?
• How do you accelerate a team through forming into norming faster?

As asked

What are the four fundamental team types in the Team Topologies model, and when would you choose a platform team over an enabling team in your org design?

Sample answer outline

Strong candidates describe stream-aligned, platform, enabling, and complicated-subsystem teams accurately. A platform team is appropriate when multiple stream-aligned teams have the same infrastructure need and the demand justifies a product team approach. An enabling team is better when the goal is to teach and then step back, not maintain an ongoing service. Candidates connect the model to their actual experience.

Expect these follow-ups

• What is the cognitive load model in Team Topologies, and how do you apply it in practice?
• When does a team become too big to keep as a single stream-aligned team?

As asked

What is psychological safety as defined by Amy Edmondson's research, how does it differ from comfort or niceness, and what is the evidence for why it matters in high-performing engineering teams?

Sample answer outline

Strong candidates describe psychological safety as the belief that speaking up will not result in punishment or humiliation, and distinguish it from group harmony or absence of conflict. They reference Edmondson's 1999 research on hospital work teams, where higher-performing teams reported more errors because they felt safe surfacing them, not because they made more mistakes. They may also cite Google's Project Aristotle, which found psychological safety to be the strongest predictor of team effectiveness. They note that psychologically safe teams surface problems faster, iterate more, and have lower error rates in complex tasks.

Expect these follow-ups

• Can a team have too much psychological safety?
• How do you measure psychological safety without making the measurement itself uncomfortable?

As asked

Salesforce has over 40 certifications available through Trailhead. If you were hiring a Salesforce developer, how much weight would you give their certifications versus hands-on experience? What certifications would you expect a senior developer to hold?

Sample answer outline

Strong answers are balanced: certifications show baseline knowledge of the platform and are useful signals especially for candidates without much prior Salesforce work history, but hands-on experience building real solutions matters more. A senior Salesforce developer should minimally hold Platform Developer I (PDI), and ideally Platform Developer II (PDII), which requires a practical coding exam. Application Architect and System Architect are meaningful for senior roles. The candidate should also mention that Trailhead badges supplement but do not replace real project experience.

Expect these follow-ups

• What would you look for in a code review from a certified but inexperienced Salesforce developer?
• How would you get someone up to speed on the Salesforce platform who comes from a strong Java background?

As asked

Starship is designed for rapid full reusability: both the booster and the ship are caught, inspected, and reflown quickly. From a software perspective, what are the hardest problems that rapid reusability creates that do not exist for an expendable rocket?

Sample answer outline

A strong answer identifies: health monitoring and prognostics software that tracks component wear across flights and predicts maintenance needs, configuration management that tracks per-vehicle software and hardware state across reflights, the challenge of incorporating inspection data from the previous flight into go/no-go decisions, and the scheduling and logistics software for a high-tempo launch cadence. The candidate should note that expendable rockets have no concept of per-vehicle flight history in their software stack.

Expect these follow-ups

• How would you design a vehicle health database that persists across flights and feeds into the pre-launch readiness check?
• What telemetry data from the landing would most help you decide if the vehicle is ready for reflight?

As asked

During atmospheric re-entry, a vehicle experiences extreme thermal loads. From a software perspective, what does that mean for sensor data integrity, onboard computer operation, and the communication blackout period?

Sample answer outline

A strong answer covers the ionization blackout (plasma sheath around the vehicle absorbs RF, cutting communications for several minutes), which means the vehicle must operate fully autonomously during that window with no uplink possible. Onboard software must be designed to complete re-entry without ground intervention. Thermocouples and heat shield sensors may produce noisy or out-of-range readings that the GNC software must handle gracefully. The candidate should note that this is why pre-planned autonomy and thorough pre-entry state verification matter.

Expect these follow-ups

• How do you design the software that handles the communication blackout if something goes wrong during that window?
• What telemetry do you log locally to reconstruct what happened during blackout for post-flight analysis?