One of my favorite churn reduction experts, Lincoln Murphy, recently published a fine article about customizations in your churn rate: Your SaaS Metrics Are Wrong if You Include These Customers.
Lincoln’s article is an excellent introduction to the things that you’ll face when you calculate the SaaS metrics. He’s one of the few experts who understands the whole SaaS model lifecycle and offers advice that’s useful for everyone, no matter if your SaaS is small or big, bootstrapped or funded.
This post adds a couple of important details that you should know about.
What to do with customers who can't churn?
In the beginning of the article, Lincoln warns you about the biggest mistake people do when calculating churn rate - including customers who can't churn.
Lincoln is right, the "real" Churn Rate that you use to calculate LTV should always use the “realistic user definition” and include only customers who can actually churn.
But what he calls “standard user definition” is what I call the user definition for Net Churn Rate.
Different customer definitions produce two different churn rates.
Both Net Churn Rate and regular Churn Rate are needed because they are used for different purposes.
What to do with customers who aren't really customers yet?
Lincoln also suggests that you experiment with different active customer definitions by e.g. leaving out newcomers when you suspect that they won't become long-term customers.
However, when you do anything else than exclude customers who can’t churn, you are stepping into a dangerous area.
There’s a pitfall:
The new churn rate is valid only for the filtered group of your customers, the ‘active customers’.
And not only the churn rate - all the other metrics calculated using the churn rate too, like CLTV. ￼
This is important because lots of people I know like to associate the Life-Time Value to new signups. They say: “Yay! I just got a new customer with LTV of $500.”
When you modify the active customer definition, you must remember to stop saying that.
The CLTV is no longer the value of ANY customer, it’s a value for the customers who go through the drills and fill your active customer definition.
You’ve just made yourself a new metric to follow - active customers - calculated based on your definition. And all the metrics you calculate, they are valid for that group and for that group only.
What customer definition does FirstOfficer use?
FirstOfficer’s active customer definition for churn calculations is:
A customer who has a paid subscription up for renewal in this month.
This means that I cannot sport those fancy boxes that you see in some apps:
“You’ve just got a new customer with life-time value X”.
Because if those customers get a refund and leave before the month is up, they aren’t active customers - they are a failure to convert a new customer.
I can’t just snatch the LTV from the other group of customers and say it’s the same for these guys.
The new customer count that FirstOfficer shows at the end of each month is the active customer count, cleaned from refunds and immediately lost customers, ready to use for CAC calculations.
That’s the thing - these metrics are meant to be used together with the numbers from bookkeeping, so I can't just freely pick and choose between formulas and definitions.
This way I can make sure that the conversion failures don’t affect the churn rate and only include ‘churnable customers’. Yet, I can still show you just one ‘new customer count’ that pretty much matches what everyone thinks it to be.
I think this is the best compromise, sidestepping most of the problems Lincoln mentions in his article.
How do you calculate CAC with different customer definition?
If you go on and use special active customer definition, you'll need to update your CAC calculations too. Here's how to do that.
Instead of dividing your marketing costs among all new customers, you need to calculate the CAC to the new active customers. If you have chose an active customer definition that e.g. forces you to wait two months until you consider a customer truly active, it takes two months until you get the CAC numbers.
Don’t mix and match months. The active customers should carry the marketing cost from the time when they signed up and the actual marketing work was done for them.
For example, if a customer signed up in January and becomes active in March, you’d use count him in to calculate January CAC using January’s marketing costs.
Now you have the right CAC against your modified churn which caused your modified LTV.
Will FirstOfficer support modified Churn Rate formulas in the future?
I continue to offer just one formula and one active customer definition, for a reason I stated above:
All and any changes to active customer definitions will propagate throughout the system to all metrics.
If I’d offer a choice in churn rate formula, I’d have to double-implement the calculation of ALL the metrics and somehow fit in the new ‘active customer count’ alongside with the current customer counts and then make it clear which one was used to calculate the metrics.
My knowledge of Activity Based Costing (ABC, now long dead except in LTV calculations) gives me a clear advantage here. I don’t have to actually implement this to see what kind of a mess it would get me into.
Use the cohort charts to see if you need custom formulas
However, Lincoln is right: in some cases redefining the active customer definition may give you great insights. Even if you’d end up calculating the metrics with that new definition just once per 3 months.
The good thing is that you can actually see if you should do that - just use the cohort charts in the FirstOfficer's Retention view.
When customer behavior is consistent, the cohort lines are smooth:
￼ When customer behavior isn’t consistent, you’ll get kinks in the line, like you see here after 3 months:
￼ If you cohort retention lines look more like in the lower chart, you might want to experiment with the customer definitions.
Just check that you have enough customers for this type of analysis. If you have too few, the cohort lines will be all around the place and tell you pretty much nothing.
Also check that the changes are similar in several cohorts and not caused e.g. upgrades due to a marketing campaign. It’s also normal for many businesses to see a dip after 12 months when annual subscribers get to leave, this does not require custom calculations either.