Server Retention (Game)

Server Retention (ServerRetention) computes retention grouped by game server (server_id), comparing user stickiness across servers. It corresponds to the backend AnalyticsService.ServerRetention.

Game-specific model. It depends on the serverId (server ID) field reported by the C# SDK. See the C# SDK Integration Guide.

1. What Questions It Answers

  • How do servers (e.g. server 1, 2, 3) compare in next-day / 7-day retention?
  • Which server has the worst retention? (possibly server merges, operations, or version issues)
  • For a newly opened server, is early retention up to standard?

2. Difference from Generic Retention

ModelGroupingUse Case
RetentionBy cohort (registration day)Generic Web/APP
Server Retention (ServerRetention)By game server server_idGames, comparing servers

3. Backend API

gRPC: AnalyticsService.ServerRetention

FieldTypeDescription
time_rangeTimeRangeTime range (filtered by first active day)
app_iduint32 (optional)Filter by app
server_idstring (optional)Server ID filter (empty = compare all servers)
max_offset_daysuint32 (optional)Maximum offset days, default 7

Response ServerRetentionResponse: rows[] (ServerRetentionRow contains server_id / cohort_size first-day new users / retention_rates map of retention rate per offset day), offset_days (offset day list).

retention_rates is a map<string, double>, keyed by day number (e.g. "1"/"3"/"7") — the only map field among the 25 models.

HTTP (admin proxy)

POST /admin/v1/analytics/server-retention
Content-Type: application/json

{
  "timeRange": { "startMs": 1718169600000, "endMs": 1718774399000 },
  "maxOffsetDays": 7
}

4. Typical Scenarios

Server quality comparison

Omit server_id to see the retention comparison table for all servers and spot servers with abnormal retention (clearly below average).

New server launch evaluation

For a newly opened server_id, look at its first-7-day retention curve to assess how well the new server onboards players.

Server merge / operations effectiveness

Before and after a server merge or major version update, compare retention changes on the affected servers to evaluate the operations action.

5. Notes and Caveats

  • Depends on serverId reporting: the C# SDK must report serverId in events (landing in the events_fact.server_id column), otherwise grouping by server is impossible.
  • cohort_size basis: each server's cohort size is that server's first-day new user count.
  • Map field: retention_rates is a map; the frontend aligns it into a table by the offset_days columns.
  • Complements generic retention: games look at both overall retention and per-server drill-downs; combining them tells you whether a problem is global or server-specific.

6. Related Documents