Compare metrics with benchmarks
Benchmarks put results in context by comparing the game with other projects.
How to use it
- Choose a metric and comparison date.
- Check where the game sits on the scale relative to P50 and P90.
- Open the chart to compare both the latest value and its change over time.
Advice and practical context
Reference depends on peer set
A benchmark describes the distribution of comparable games or a broader Roblox group. The set can update, so record its date and type. P90 is a high percentile, not a guarantee of future distribution or deal quality.
Read with history
When a game crosses P50, check whether its metric rose or the band changed, then inspect segments and guardrails. Do not select an investment from one percentile or compare metrics with different units.
Label units and direction: higher is favorable for many KPIs, while errors or bounce metrics run the other way. The benchmark must match metric and population. A P50 from one percentage cannot be applied to another metric.
Technical detailsCalculations, permissions and behavior
Exact percentile boundaries
The manual entry table has been removed. Only percentiles with verified helper provenance remain visible. Unverified history is retained but not plotted.
The flat previous-period range of the game itself is also removed from all charts. While Chrome with the paired helper is open and the Roblox session remains valid, latest cards and the last 30 days of P50–P90 are checked hourly; Roblox backoff and publication delays can postpone updates. No extension update is needed.
A separate persistent queue attempts to fetch official scorecards and P50-P90 bands, stores every accepted official date as its own revision, and never expires history automatically. An incomplete pass resumes on the next minute; temporary failures retry after 30 minutes with increasing backoff, 429 after one hour or later per Retry-After, and 401/403 after six hours. However, the cross-experience benchmark used by Creator Dashboard is not currently published by Roblox as an Open Cloud API: both the official API key and OAuth return 401/403 to MARPLA even though the owner's browser can see the cards through its session.
The queue therefore cannot reliably backfill exact percentiles in the background, and MARPLA neither copies browser cookies, invents values, nor presents a stale manual import as a fresh scorecard. The supported Analytics Query API continues to refresh and retain the game's own metrics.
Which benchmarks are available
Roblox displays cross-experience benchmarks for retention, playtime and sessions, ARPDAU, ARPPU, payer conversion, and Home recommendation play-through. The documented Analytics Query API returns the game's own line. P50, P90, and the exact rank come from a different session-bound Creator Dashboard method that does not currently accept MARPLA's server-side Open Cloud credentials.
Stored official responses are shown only on their exact date and missing dates are never bridged. An empty band means no verified official response, not zero. MARPLA never derives cross-experience boundaries from one game's history or substitutes manual or estimated values.
P50, P90, dates & gauges
The six primary benchmarks share one compact row on wide screens. The Daily / 7-day average switch selects the window, while the shared calendar selects its end date. Daily shows the selected day; 7-day average shows the range from six days before the end date through the selected date.
The arrows shift either window by one calendar day. By default, Latest shows the latest available date separately for each card. For a calendar selection, each card shows the latest confirmed period no later than the chosen day.
Its actual date or seven-day range is always printed below the title. D1 and D7 refer to the cohort's first-join date and require 1 or 7 days of observation plus Roblox processing; the latest mature day is included. A connected helper prioritizes the selected history without resetting Roblox backoffs.
The metric value comes from Analytics Query API, while the exact percentile, P50, P90, and trend come from the stored Creator Dashboard response for that same date and window. Scorecard history is retained by date. The coloured gauge ends exactly at currentPercentile; without an exact scorecard for the selected date, the gauge stays neutral with no invented position or colouring.