Claim
The right object of attention for product growth is the marginal user — the person who wants to take your desired action but has not yet, in the worst conditions you support. Find them in high-traffic, low-conversion segments. Watch them use the product. The data will show you where they drop; only watching shows you why.
Mechanism
Average-user optimisation pulls product decisions toward the existing user base, which has already self-selected through the friction the team is trying to remove. Power-user optimisation pulls toward feature depth that few new users need. The marginal user — high-intent, low-success — is where the largest unconverted demand sits. Fixing their experience lifts everyone downstream because the friction that blocks them also taxes everyone else, more lightly.
Conditions
Holds when:
- The team has access to real marginal users (recruiting them, watching them).
- The product has heterogeneous conditions of use (geography, device, network) where marginal cases concentrate.
- The team can distinguish "wants to convert but cannot" from "does not want the product."
Fails when:
- The team mistakes power users for marginal users and over-invests in advanced features.
- "Marginal user" becomes a politically convenient label for any segment the team already wanted to chase.
- The marginal user's problem is so context-specific (one country, one device class) that the fix does not generalise.
Evidence
"When you fix marginal user problems, you fix everyone downstream. Focus on making that person's experience stupid easy."
Adriel watched someone sign up for Facebook in India. The user used their legal name; nobody in real life called them that; friend requests went unaccepted. The funnel data blamed the friend-request acceptance rate. Only watching the user surfaced that the registration name field was the actual bottleneck.
— Adriel Frederick on Lenny's Podcast, 2026-04-28
Signals
- Conversion lifts in the segment researched, plus collateral lifts in adjacent segments.
- Product decisions reference specific marginal-user observations, not aggregate funnel metrics.
- Team rituals include regular sessions watching real users in real conditions.
Counter-evidence
For mature products with high conversion across all conditions, marginal-user research returns diminishing signal. The discipline is most powerful at scale-up stage where the next 10x of growth depends on the next 10x of friction reduction.
Cross-references
- Lead with understand-work, not identify-and-justify — data informs understanding, not the reverse — the broader frame
- Growth needs both lead bullets and cannonballs — the laziness trap is shipping only lead bullets — the same operator's portfolio rule