Content generated with AI — it may contain mistakes.

ExyliaAnalytics/GuidesUpdated in v1.2.0

Track growth and retention

Players › Growth, Activity › Hours & playtime, Retention, Player sources and Campaigns: what each chart shows, how every number is defined, and which agent events feed it.

Page 5 of 2810 min read

In a few minutes you can tell whether the server is growing, whether new players stay, where they come from, and whether a campaign paid for itself. These pages read the network sessions (from the proxy or a standalone server) and the presence on each game server. Hostname, country, language, client version and Java/Bedrock come from the session a player opened; playtime, AFK and activity from the presence.

Before you start#

You need a linked proxy or standalone server that has seen players join, see Connect your server. A new link needs days of data before comebacks and retention fill in; a chart saying "Collecting since ..." is normal at the start. To judge campaigns, point each place you advertise at its own hostname on your proxy.

Steps#

Check growth

Open Players › Growth. Is the server growing? reads the last 8 full weeks as Growing, Stable or Declining. Players gained per player lost above 1 means the base grows.

Check that new players stay

Open Retention. Read Played again on Day 1, Day 7 and Day 30, and the dashed target lines (40%, 20% and 10%).

See where players come from

Open Player sources. Sources and campaigns lists where new players first joined from, with how many played 10+ minutes.

Create a campaign

On Player sources › Campaigns, pick New campaign and choose a channel (Discord, YouTube, Ads, Server list or Something else). Fill in Name, Start of the address and Your main server address, then Create campaign. Create the address it shows under Link to hand out in your DNS and point it at your proxy, otherwise players cannot join with it.

Record what you paid

Use Record a cost: choose the Campaign, the Day, the Amount and a Note, then Add cost.

Check that it works#

Growth shows a line for new players and a verdict on the momentum. Retention shows the three Day numbers once enough days have passed (a cell that is not there yet shows —, never 0%). After you point a hostname at a campaign and a player joins through it, the campaign's New players counts one.

Going further#

Definitions#

The same words mean the same thing on every page.

TermDefinition
Active playerHas a network session that started that day.
DAU / WAU / MAUUnique active players in the trailing 1 / 7 / 30 days.
StickinessAverage DAU ÷ MAU over the range.
PlaytimeSession time minus AFK time. AFK time is shown on its own.
Players onlineFrom the regular check-ins: the proxy's count on a network, the sum of standalone servers otherwise.
IdentityBedrock when any of the player's sessions was reported by Floodgate, or their UUID is a Floodgate one (00000000-0000-0000-…). Cracked when the UUID is an offline-mode one made from the name. Premium otherwise. A Bedrock player linked to a Java account keeps the Java UUID and counts as Bedrock only where the agent could ask Floodgate.

Growth accounting#

GroupDefinition
NewFirst session ever in the bucket.
Kept playing (retained)Active in the bucket and in the one before.
Returned after a break (resurrected)Active, not new, and inactive for at least 30 days before.
Came back (returning)Active, and none of the above.
Gone (churned)Last session exactly 30 days before the bucket ends: a player churns the day they reach 30 days away. Drawn below the line.
Players gained per player lost (quick ratio)(New + Returned after a break) ÷ Gone. Above 1 the base grows.

Lifecycle segments#

Decided as of the end of the range, in this order, each player in one segment only:

SegmentDefinition
NewFirst seen less than 7 days ago.
Gone (churned)Last seen more than 30 days ago.
Fading away (at risk)Last seen 7 to 30 days ago.
Loyal (core)Active on 12 or more of the last 30 days.
RegularActive on 4 to 11 of them.
Occasional (casual)Active on 1 to 3 of them.

Growth#

Open Players › Growth.

ChartShows
Is the server growing?The weekly trend over the last 8 full weeks: Growing above +2% per week, Stable, or Declining below -2%.
Players gained per player lostPlayers gained for every player lost over the range, with the break-even line at 1.
New, regular and returning playersWho the active players of each period are.
The player base over timePlayers in each lifecycle segment at the end of every period; gone players are left out.
New players by identityPremium, Cracked and Bedrock among new players.
Where players are headingDaily active players projected ahead as a straight line through the range, with a 95% band. No seasonality: a weekend pattern shows up as noise. Needs enough days in the range.

Hours and playtime#

Open Activity › Hours & playtime.

ChartShows
Time playedPer player per day (AFK left out), visits per player per day, and total playtime.
When players are onlineAverage players online per weekday and hour: the best time for events and restarts.
How long visits lastEach visit bucketed into under 5 min, 5–15, 15–30, 30–60 min, 1–2 h, 2–4 h and over 4 h, with the median.
Where players spend their timeHours on each game server per period, AFK included. Needs game servers that report joins and leaves.
What players do per hour playedChat messages, commands, deaths and player kills per hour of playtime, from the counters each server reports when a player leaves it.
Who does most of the playingThe share of playtime from the top 10% and top 1% of players, the typical player (median) and the top 10% player (90th percentile).

Retention#

Cohorts are the players who arrived on the same day or week. Retention can be filtered by Campaign, Country, Language and Edition (All Java, Premium, Cracked, Bedrock).

MetricDefinition
Played again on Day 1 / 7 / 30Share of a cohort that came back on day N or later.
How many new players stayShare of a cohort active exactly on day N (or week N) after arriving.
Return by join dateOne row per arrival day or week, each cell the share that played again after. More green means more came back.
Are new players staying longer?The Day 1 / 7 / 30 numbers per cohort, over time: is retention getting better?

A cohort only counts for day N once day N has happened.

Comebacks, win-back and what brings players back

The lower half of the Full analysis tab follows every player, not just new ones, and ignores the filters above.

MetricDefinition
Returned after a week awayAn active day whose previous active day was 7 to 29 days earlier.
Returned after a month awayAn active day whose previous active day was 30 or more days earlier.
Share who returnedPlayers with a comeback ÷ active players.
Won back after a week / a monthOf the players who went quiet for 7 (or 30) days in a period, the share that came back within 30 (or 90) days.
Returners who stay (comeback stickiness)Share of comebacks followed by another active day within 7 days.
Time between visitsEvery gap between visits, bucketed from 1 day to more than 90.
What brings players backComebacks by the hostname and campaign they came back through.

Other sections on the same tab: Players at risk, Regulars slipping away, Why new players don't come back, What the players who stay did on day one and Where they quit.

Player sources#

Open Player sources, titled Where players come from.

SectionShows
KPIsNew players, Played 10+ minutes (activation: share whose first day holds 10 minutes of playtime or more), Bedrock share of new players, and Votes with the number of voters.
Sources and campaignsNew players, cost, revenue and payers of each source or campaign.
Join addresses, Countries, Languages, Client versions, Java and BedrockNew players by the address they typed, where they connect from, their game language, their Minecraft version and their edition, each with its activation and Day 7 retention.
Tracked sources, Unknown or direct IPSources you named with a campaign, against joins from an address that no campaign claims.
First visitWhat new players did on their first visit, by source.
VotesVotes per period and which voting sites bring them. Empty without NuVotifier.

Countries come from the player's IP, looked up on the ingest ("IP Geolocation by DB-IP"). Languages come from the client's language setting. Client versions come from the protocol number, which a standalone Paper server reads from ViaVersion when it is installed.

Campaigns#

A campaign is a promotion with its own address: give each place you advertise a hostname, such as tiktok.yourserver.net, point it at your proxy, and record what you paid. A player belongs to the campaign of the hostname of their first session, for ever.

  • New campaign takes a name and one or more hostnames, separated by commas or new lines. Templates prefill the name: Discord, YouTube, Ads, Server list, Something else.
  • Record a cost: an amount on a day, with an optional note. Costs on more than one day make the campaign a shaded period on the charts.
  • The Hostnames table lists every hostname new players used in the range, and maps each to a campaign. Hostnames are compared normalised: lowercase, no port, no trailing dot, and the extras that Forge, TCPShield and BungeeCord forwarding append are stripped.

Each campaign's own page shows New players, Revenue, Players who paid, Played 10+ minutes, Spend, Cost per new player (CAC: spend ÷ new players), Revenue per player and Revenue per 1 spent (ROAS: store revenue after refunds from the campaign's players ÷ spend, so 2.00× brings back two for each one spent). Their countries, editions, versions, retention cohorts and top spenders follow; revenue and spenders appear only for members who can see revenue. The dashboard also defines Return (ROI) as (revenue − cost) ÷ cost.

Creating campaigns, mapping hostnames and editing journeys needs Create campaigns, map hostnames and edit journeys (manage_campaigns).

Something missing on this page? Tell us on Discord