Weekly/Monthly Retention
IntermediateWeekly or monthly retention measures whether users come back within a week or a month rather than on a specific day, for products that are not expected to be used daily.
The cohort's timeline is divided into weekly or monthly buckets, and a user counts as retained in a bucket if they were active on any day inside it. The reason is straightforward: a product whose natural rhythm is weekly looks catastrophic on a daily curve while nothing is actually wrong. That same bucketing inflates the figure, since each week offers seven chances to be counted, which makes a weekly number incomparable to a daily one. The subtler trap is the bucket definition: calendar weeks and weeks counted from each user's own start date give different answers, because under calendar weeks someone who joined on Thursday has a three-day first week.
How it is calculated
Week-N retention = cohort users active at least once during week N ÷ cohort size
The bucket definition must be published with the number, stating whether weeks are calendar weeks or weeks counted from each user's own first day. The unit is a percentage, the denominator is the cohort size at bucket zero, and bucket zero is one hundred percent by definition and is usually dropped from the chart. The same formula applies to monthly retention with months in place of weeks.