How do product analytics tools calculate retention differently, and why do two tools often give you retention curves that disagree?

“: Supports different retention behaviors and time intervals. Its retention methodology can distinguish rolling intervals from calendar intervals.”

“Amplitude also documents weighted retention calculations.”

“Amplitude explicitly documents that incomplete intervals are excluded from certain overall retention calculations, which can cause an apparent upward curve when data is still accumulating.”

“: Its documentation distinguishes between exact-day and on-or-after retention. It also supports rolling 24-hour and strict calendar-day calculations.”

“PostHog, for example, explicitly distinguishes between first-time, first-ever-occurrence, and recurring retention cohorts.”

“PostHog explicitly offers both simple and weighted mean calculations.”

“: Offers several cohort-entry methods, two retention reference options, and simple versus weighted averages.”