L28 Engagement

L28 engagement is the number of distinct days a user was active in the trailing 28 days — a value from 0 to 28 that is read as a distribution rather than an average.

L28 (from last 28 days) standardizes active days onto a fixed 28-day window so users are comparable; why 28 and not 30? Because 28 is exactly four weeks and cancels day-of-week effects (busy Saturdays, quiet Fridays). The key point is that L28's value is not in its average but in the shape of its distribution: a histogram of users by active days usually has two peaks — a mass near 0–3 (casual) and a bump near 28 (daily users). That distribution is the basis of the power-user curve and tells a team growth should push users from the left peak toward the right. The common trap is reporting an average L28, which hides the bimodality and produces a number no real group matches. L28 is meaningful for products with an expectation of frequent use, not inherently low-frequency ones.