How do you instrument a marketing funnel versus a product-led growth loop?

WHAT THIS TESTS: Whether you understand that a funnel is a linear one-way street where energy dissipates, while a growth loop is a compounding system where output becomes input. The interviewer wants to know if you can architect telemetry and data models for attribution versus viral mechanics, not just slap the same analytics library on two different products.

A GOOD ANSWER COVERS: Four engineering distinctions in order. First, instrumentation scope: funnels instrument sequential user events across AARRR stages like page views, signups, and purchases within a single session or identity, while loops instrument cross-user events such as invite sends, invite acceptances, user-generated content creation, and referral-driven signups that link a new user back to a referrer. Second, metric design: funnel engineering optimizes stage-to-stage conversion rates, channel-level CAC, and drop-off percentages, whereas loop engineering optimizes viral coefficient, loop cycle time, and the ratio of organic acquisition to paid acquisition because each user acquired generates future users. Third, data architecture: funnels rely on session-based attribution and last-click tracking, but loops require persistent identity resolution to connect invitees to referrers, often across devices and anonymous sessions, plus graph-style data models to measure network effects. Fourth, dashboard feedback loops: funnel dashboards aggregate historical drop-off rates, while loop systems need near-real-time telemetry showing whether the current cycle is accelerating or decelerating so product teams can tune mechanics immediately.

COMMON WRONG ANSWERS: Treating both systems as simple conversion-rate optimization problems. Saying you would use the same event taxonomy and attribution model for both. Ignoring the identity resolution problem in loops. Focusing only on front-end tracking without discussing how backend systems must attribute downstream revenue to the original viral trigger. Claiming that CAC always rises with scale in loops, which misses the compounding effect described in the canonical model.

LIKELY FOLLOW-UPS: How would you calculate and instrument viral coefficient in a B2B product with long sales cycles? What happens to your data model when a user enters through a paid ad but then triggers a referral loop, how do you attribute LTV? How do you detect a broken loop versus a funnel bottleneck using telemetry? When should a startup abandon funnel thinking entirely?

ONE CONCRETE EXAMPLE: Consider a file-sharing product. In a funnel model, you instrument upload, share-link creation, and premium upgrade as a linear sequence, measuring drop-off between each step and optimizing checkout flow. In a loop model, you instrument the share-link creation as a viral event, track when recipients open the link as acquisition events tied to the original user ID, measure how many of those recipients then upload and create their own share links within seven days to calculate cycle time, and build a real-time dashboard showing whether one share generates more than one new share over that window. If the ratio falls below one, the loop is broken and you are back to linear spend.

Source: LeanPivot.ai