February has just 28 days. You probably have customers whose subscriptions were due at 28th, 29th, 30th and 31th. In February all those people got charged at 28th. That means you've got 4 days worth of recurring revenue booked to a single day.
Yay! Except that the peak at Feb 28th will trigger a problem in so called “rolling metrics” in March.
In March, the apps that are based on “rolling metrics” do this fun thing:
See that huge dip in MRR? That’s what happens when the day that carried 4 days worth of money drops out from the 30-day period.
No, your MRR didn’t crash overnight and yes, the numbers are right. This is just how this type of trend data behaves.
I don’t know why the drop in the chart happens at 27th and not 30th like it should, though. Maybe there's some fuzzy logic to make the bloat smaller and keep the worst of it out from the weekly reports.
Does FirstOfficer have rolling metrics?
When I was building FirstOfficer and the first SaaS metrics app appeared that used the 30-day period to calculate MRR, my accountant brain went:
“Abomination! You CANNOT calculate MRR like that. That’s not MRR. What is that anyway? Daily-Monthly? I’ll never have that in my app!”
But guess what… that way to calculate “MRR” became hugely popular.
And it taught people that they could have “MRR” any day that they wanted, not just once per month. And people kept asking me to add metrics like that too.
FirstOfficer has rolling metrics - but only in the weekly reports
I eventually added the 30-day rolling recurring revenue to my weekly reports.
I'm not against 30-day rolling averages. I just think this particular metric should have been divided by 30 and given a proper name, like "Daily Average Normalized Subscription Revenue".
Gritting my teeth, I named it “MRR” as people already knew it by that name:
So this means that…
My customers have been affected by this problem too
I should have let you know about this already before, but I only noticed this after a customers asked about the extra money in the weekly report.
The good thing is that if you’ve been following your MRR from the dashboard, you’ve seen the unbloated figures all along.
And these 30-day numbers are right too - this is just the way they behave.
I understand if you feel that this a lousy explanation when you expected your MRR to be higher.