How would you use telemetry and logs to refine SAM calculation?

WHAT THIS TESTS: This question tests whether you can bridge engineering telemetry and business strategy to produce a bottom-up SAM estimate. Interviewers want to see that you understand SAM is the portion of TAM you can realistically serve based on product fit, distribution, and technical constraints, and that you know how to operationalize that filter using internal data rather than analyst reports alone.

A GOOD ANSWER COVERS: First, define the SAM boundary by starting with TAM and narrowing it. Explain that telemetry helps you find who is actually using the product and how. Second, map API usage and feature flag data to segment fit. For example, if enterprise accounts heavily use a specific endpoint while SMBs do not, that signals which segments align with your current capabilities. Third, use performance metrics and system logs to identify serviceability limits. High latency in certain regions, frequent integration timeouts, or low availability in specific zones reveal where you cannot yet reliably deliver service, so those populations should be excluded from SAM even if they show demand. Fourth, rebuild the number bottom-up by counting qualified active accounts or organizations that pass both the fit and feasibility filters, then multiply by a realistic average selling price or usage-based revenue per account. Mention that this should be cross-checked with sales coverage and support capacity so you do not overstate reach.

COMMON WRONG ANSWERS: Confusing SAM with TAM by claiming all API consumers are part of the market. Treating raw log volume as revenue potential without filtering for bots, free tiers, or trial accounts. Ignoring technical constraints and suggesting that global traffic automatically means global serviceability. Proposing a purely top-down approach that applies a fixed percentage to an industry report without using your own logs to validate segment behavior.

LIKELY FOLLOW-UPS: How would you distinguish signal from noise in noisy logs? What would you do if telemetry showed high usage in a region where you have no sales team? How would you validate that a feature flag correlation actually indicates willingness to pay rather than just curiosity? When does it make sense to expand SAM versus shrink it based on these signals?

ONE CONCRETE EXAMPLE: Suppose your platform has one thousand accounts hitting a new analytics API. Telemetry shows eight hundred are in North America and Europe with sub-one-hundred-millisecond latency, while two hundred are in Southeast Asia with frequent five-second timeouts and low feature flag adoption for advanced exports. Performance logs also reveal that the Southeast Asian traffic is mostly from free-tier trial workspaces. A refined SAM calculation would count the eight hundred proven high-fit accounts, apply a conservative expansion factor for similar accounts in those regions, and set aside the Southeast Asian cohort until infrastructure and paid conversion improve. This yields a SAM grounded in actual serviceability rather than hope.

Source: gwi.com