Your annual reminder that your monthly metrics will be down 10% in February because February is 10% shorter than January. Please do not contact your data team about it.
@xian@seldo No, no, trailing 28 days. Always use an even number of weeks, because lots of things have weekly patterns, so if you use something that isn't a multiple of 7 days, you end up introducing a bit of a "beat" to the time series depending on how many weekends your trailing days include.
@xian@abulsme@seldo Italian phone companies tried to apply this reasoning a few years ago, to lure customers into paying "monthly" fees 13 times in a year (instead of 12).
They got fined by the antitrust authority and the practice has been deemed illegal since then, luckily.
@lazza@xian@seldo Ha. Well that is a bit different than metrics tracking. Trailing 28 day metrics are a possible substitute for monthly metrics, but they aren’t actually “monthly” and anyone useing them should be clear on what they are. What you describe sounds intentionally deceptive.
@Almafeta@lazza@xian@seldo 23 is prime and unaligned with any normally used time frames. Is their any logic behind that timeframe other than to confuse people and make them more likely to accidentally miss a payment?