Estimating ad-acquired subscriber value
I recently ran an ad campaign for Candy Japan that was a pretty close call, so I had to get a little bit more sophisticated to figure out whether it is worth continuing the campaign or not. Here's what I learned.
Suppose you are running an ad campaign for your subscription business.
You know the average value of your subscribers so far, so a simple way to determine if your ad campaign is successful would be to see how much it costs to acquire a customer through an ad, and then comparing that cost against the average customer value. This could be fine, if the numbers show that the campaign is clearly profitable.
But if it's not so clear, you might end up making a mistake with this simple approach.
Simple average ignores retention differences
Suppose you've had 1000 customers so far, and they have been worth on average $50 each to you in profit. Now when running an ad you would assume that those future customers would also be worth $50 each. But that could be wrong, if the quality of those visitors is different. For instance if all your existing customers came from word-of-mouth, but your ad is running on Reddit, you can imagine that those customers might behave very differently.
They may not subscribe for the same amount of time, and if your subscription business has tiers, then they might not be choosing the same tiers as often. If you just take the average value of all of your customers so far, that might differ significantly (positive or negative, likely positive) from the value of a customer gained through ads.
Luckily by running your ad for a while and looking at how customers initially behave, you can extrapolate to how they would likely behave in the future.
Taking retention differences into account
To give you a real example, here is a chart comparing Candy Japan customer retention for new subscribers gained from a YouTube campaign.
You can see from the chart the percentage of customers who remain subscribers over time.
The green line is data from all subscribers, going down over time as more and more customers eventually cancel their subscriptions. For instance after four months, a little over half of subscribers are still left. Luckily this trend flattens out — a fourth of customers are still subscribing after 11 months, and while the chart does not show this, many continue to subscribe for years.
The red line is for customers acquired through a YouTube ad campaign that ran for a short period four months ago. As you can see, although there is limited data, you can make a fair guess for how the red line would continue.
As those customers were more likely to cancel during the first four months, then one could assume that in the future they will continue to be more likely to cancel. Comparing the churn at the first three points where customers had an opportunity to cancel I can see that their likelihood to cancel is 1.23 higher than baseline. Continuing to apply this to the months that follow results in the following extrapolation.
Now with this result you can calculate the value of a customer acquired from the ad campaign vs. your other customers.
In my case the conclusion was that customers originating from YouTube ads are 20% less valuable over the first year than an average subscriber. Here's part 2 where I figured out whether the overall campaign was profitable or not.