User guide

Instructions for each chart

The book beside a chart opens its metric-specific explanation: meaning, calculation, limitations and testable actions. Conversion codes appear in chart titles and summary cards. Product charts start with the top ten enabled; manual choices persist.

If all lines are off, enable them in the legend. Per-player mode requires DAU for the same date.

Six primary benchmarks

The Benchmarks tab starts with six comparable cards: average playtime, D1 retention, D7 retention, payer conversion, average revenue per paying user and Home recommendation play-through rate. By default, Latest selects the available date per card and shows it in the hover tooltip: D1 and D7 may lag revenue. Choosing a calendar date leaves Latest mode.

One Daily / 7-day average switch updates the window for every card. A calendar with arrows sits beside it: Daily shows one selected date, while the seven-day mode shows the exact rolling seven-day range. Each arrow shifts the selected end date by one day.

Each benchmark uses its last confirmed day no later than the selection, with the actual day or seven-day range available in its hover tooltip. D1 and D7 are dated by the cohort's first join and require 1 or 7 days of observation plus Roblox processing. The tooltip marks older periods as latest available.

A connected helper prioritizes the selected history. Value, exact percentile, P50, P90 and change come from one scorecard for the same date and window; daily percentiles are never averaged into seven-day percentiles. If no confirmed data exist, the value or gauge is not invented.

Analytics Query API ratios are normalized to percentages only for percentage metrics. Each card opens the corresponding chart in the correct analytics section. ARPPU uses the hexagonal Robux symbol.

Retention: chart order and benchmarks

The section starts with the new-user first-session retention curve styled after Creator Dashboard: the large value is the share still playing after five minutes, with an elapsed-time axis from launch to minute 30. The chart has its own request and refresh button; an open tab checks stored results every minute without waiting for other breakdowns. MARPLA requests the official TotalSessionsEndedInBucket metric broken down by SessionTimeBucket and stores all daily observations without a top-30 limit.

Numeric values 0, 30, 60, …, 1800 mean seconds; each count already represents surviving players. Over the selected dates, MARPLA divides the summed count at each second by the summed count at launch, never adds different thresholds or averages daily percentages. Duplicate days are not added.

The period ends on the latest complete observation day, not the sync date. If the selected day is not ready, the previous complete day is shown; actual dates are always visible below the heading. Ranges aggregate daily cohorts through the latest available day.

Incomplete range history falls back to an explicitly dated latest complete day without inventing missing values. A partial refresh does not erase the saved chart. The exact five-minute value appears only when its observations are present.

Smoothing changes only the line between confirmed points. Average playtime and Average session time follow. Both daily charts render a gray band only from stored exact Roblox P50–P90 boundaries for the comparable population; MARPLA neither substitutes an estimated range nor bridges missing benchmark points.

D1, D7 and D30 retention and the remaining return metrics follow. If Roblox has not returned the observations or benchmark yet, the interface shows missing data instead of a fabricated line.