Guide
Day 1, Day 7 and Day 30 retention for Minecraft servers
Retention asks one plain question: do new players come back? Day 1, Day 7 and Day 30 are the three numbers most server owners use to answer it, but tools define them differently, and two servers quoting "Day 7" can be measuring different things.
This guide defines each number, shows how a Day 7 figure differs from a within 7 days rate, explains why a fair chart shows some players as pending, and ends with what to change when a number is low.
What Day 1, Day 7 and Day 30 mean
Take the players whose first ever join fell in one period, say one week. That group is a cohort. Day N return is the share of the cohort who started a session on day N after their own first join.
Count a day as a 24 hour step from each player's first join, not as a calendar day. A player who first joins on Monday at 20:00 has Day 1 from Tuesday 20:00 to Wednesday 20:00, and Day 7 from the next Monday 20:00 to Tuesday 20:00. Calendar days would give someone who joined at 23:50 a Day 1 ten minutes later, and the result would change with the time zone of whoever reads the chart.
Decide what counts as a session, too. A quick reconnect or a failed login is not a player coming back. A simple rule is a confirmed login on a server with a login plugin, or a few minutes of play on one without.
- Day 1: did the first session make them want a second one?
- Day 7: did the first week give them a habit or a goal?
- Day 30: is there still a reason to be here after a month?
Day 7 versus a within 7 days rate
"Were they here on day 7?" and "did they come back at all in the first week?" are different questions, and both get called 7-day retention. Five players show the gap:
| Player | Came back on | Day 1 | Day 7 | Day 30 | Within 7 days |
|---|---|---|---|---|---|
| A | Day 1 and Day 2 | Yes | No | No | Yes |
| B | Day 3 and Day 7 | No | Yes | No | Yes |
| C | Day 5 | No | No | No | Yes |
| D | Never | No | No | No | No |
| E | Day 12 and Day 30 | No | No | Yes | No |
| Rate | 20% | 20% | 20% | 60% |
Day 7 says one player in five was around a week later. The within 7 days rate says three in five came back at least once. Both are true, and they answer different questions.
Check which one a tool shows before you compare numbers. Analyse's retention docs, for example, define their day 7 figure as a return at any point within 7 days of the first session, as of September 2026. MCAnalytics shows both under different names: the 7-day return rate on Overview and Growth is the within 7 days number, and Day 7 on the Retention page is the seventh day only.
Use the within 7 days rate to judge sources, because it gives one steady number per source early on. Use Day 1, Day 7 and Day 30 to judge the experience, because they show where along the first month players drop off.
Pending windows
A player who joined three days ago cannot have a Day 7 yet. If you count them as not returned, your rate drops every time a wave of new players arrives, which is exactly when you watch it most closely, after a big video or a launch.
The fix is to leave open windows out. For Day 7, only players whose day 7 has ended belong in the denominator. Everyone else is pending: neither a return nor a loss.
A hypothetical example
300 players joined in the last 10 days. For 100 of them, day 7 is over, and 20 of those came back on day 7. The Day 7 rate is 20 out of 100, or 20%. Dividing by all 300 would report 7%, and the only thing that number measures is how recently people joined.
When no window has closed yet, for example in a new server's first week, a fair chart says so and shows when the first windows close, instead of printing 0%.
Reading a cohort table
A cohort table has one row per week of first joins and one column per day. It separates what you changed from who happened to arrive.
- Read down a column to see whether a change helped. If the weeks after an update have a better Day 1 than the weeks before, the update probably helped.
- Read along a row to see how one group fades. The biggest drop between two columns shows where to look first.
- Mind the size of each row. A week with 30 new players can swing by ten points on a handful of people, so do not react to one small row.
- Split before you decide. A cohort mixes sources and player types, and a strong group can hide a weak one. Split by join address, and by premium Java, cracked Java and Bedrock.
- Note what happened each week: updates, resets, events, outages, a creator video. A week that dips usually has a story, and it is easier to find while you still remember.
What to do with the numbers
Day 1 is low
The first session is the problem. Join on a fresh account and time how long it takes to understand what to do. Look at the spawn, a long login or register step for cracked players, a resource pack prompt with no explanation, lag at join, and a wall of chat messages nobody reads. Change one thing, then compare the next cohorts with the ones before.
Day 1 is fine, Day 7 falls away
Players liked the first visit but found no reason to return. Give the first week a shape: a goal that takes several sessions, daily rewards, scheduled events, and an easy way into your Discord so they hear about them. Party and friend features give them someone to come back for.
Day 30 is low
The first month runs out of things to do. Look at long-term progression, the economy and how often you reset. Players who went quiet and then came back are worth a look of their own: what brought them back may be worth offering earlier.
One source is much lower than the rest
The source sends the wrong players, or the right players to the wrong place. Check where its address lands, the lobby or the mode the creator played, and compare what each deal cost per returning player; see measuring server list and creator ROI.
What about benchmarks?
This guide gives no "good" retention numbers, on purpose. We have not found a public dataset that covers Minecraft servers broadly, and figures quoted online rarely say how they define a day, which players they count, or what counts as a session. A number nobody can check is not a target. Compare each cohort with your own earlier cohorts, and judge a change by whether it moved your own numbers.
Where MCAnalytics fits
The MCAnalytics Retention page shows Day 1, Day 7 and Day 30 as defined above: 24 hour steps from each player's own first join, a session that counts only with a confirmed login or, on a network with no login system, at least 180 seconds of play, and Pending instead of 0% while a window is open. The weekly cohort grid covers 4 to 26 weeks and splits by join address or identity type. The 7-day return rate for each source is on Overview and Growth.
Read the 7 day return rule for the exact rules, or see player retention for the page itself.