Most users never travel the whole way from novice to expert; they settle in between

What it is

Novice and expert are often treated as the two ends of one gradual spectrum, with the implicit assumption that most users will eventually walk from one end to the other. In reality, most users of mainstream software neither stay novices nor ever reach expert — they park at some intermediate level for the long haul, and that parked state isn't a transitional phase, it's a permanent steady state. This population of "perpetual intermediates" is typically the largest single group by headcount, often outnumbering novices and experts combined — a fact that directly caps the "optimize for novices or optimize for experts" framing itself, since that framing quietly writes off the group with the most people in it.

Why it happens

Staying at an intermediate level isn't a failure to learn — becoming an expert requires deliberate investment: repeated exposure and practice deliberate enough to compress an action into recall. That investment requires a user willing to trade current task progress for future speed. But most people open a piece of software to get a specific thing done right now, and their attention is naturally pulled toward "finish this with the method I already know" rather than "stop and learn a better way" — this task-completion-first tendency, which overrides the motivation to learn, is exactly what's repeatedly documented as the paradox of the active user: users would rather push through with an inefficient but known path than spend time learning a faster one, even when the faster path would pay off in the long run. The result is that the same person can look like an expert on the small handful of core functions they use every day (genuinely drilled to recall level) while staying permanently at a novice or recognition level on most of the rest of the software's functionality. This mixed, "locally expert, globally novice" profile is the norm — not a temporary waypoint on the road to full mastery.

Studying it

Confirming this claim for a specific product requires looking at the distribution of feature-usage frequency, not a vague "average proficiency" figure: log each user's calls to each feature, and plot the distribution of feature-coverage breadth against per-feature usage depth per user. If most users stay concentrated on a small subset of features for the long term, with usage outside that subset barely growing over time, that's direct evidence of a perpetual-intermediate population — not a theoretical inference. This kind of analysis needs to be repeated across a sufficiently long time window; a single snapshot can't distinguish "still learning" from "has stopped learning."

Where it stops holding

This doesn't deny that genuine experts exist. Users who treat the software as a daily work tool and have professional motivation to master all of it — administrators, dedicated operators — do keep investing and do reach genuine expert status, and this population makes up a much larger share in certain product categories (professional software, productivity tools). The claim mainly applies to mass-market consumer software: in that setting, expecting most users to eventually "graduate" into experts who have mastered the entire feature set is a design assumption that doesn't hold.

Applying it

• Don't design the "onboarding phase" as a temporary state most users will eventually grow out of — most users will stop advancing once they reach some intermediate level, and the interface's default information density and default visible operational support should be configured for how this group experiences long-term, parked use, not for an assumption that they'll eventually learn everything.
• Verify the actual proficiency distribution with your own product's feature-usage histogram: if usage concentrates on a small cluster of features and stays there over time, optimize the default interface to make that small cluster feel effortless while keeping the rest discoverable through recognition — rather than piling on more onboarding in the hope that "just a bit more learning" will close the gap.
• Verification: pull the same cohort's feature-usage logs across a window of six months or more and compare feature-coverage breadth early versus late. If coverage breadth stays roughly flat while usage depth on the same features keeps climbing, this cohort has settled into a permanent intermediate state, and guidance aimed at "helping them learn more features" has limited payoff — resources are better spent polishing the experience they're already using.

Related

• Same group: A11.04.1 Novices rely on recognition, experts rely on recall · A11.04.3 Onboarding guidance is a net cost for experts
• Adjacent: A6.06 The forgetting curve · A9.13 Choice overload
• Search: paradox of the active user · perpetual intermediate · production bias