I have spent a long time working with clients who do most of their customer acquisition through Google and Facebook, and I even spent two years at Facebook as its head of marketing operations and analytics. Time and time again, I have noticed that marketers who let Facebook or Google auto optimize their ads to maximize for return on ad spend (ROAS) end up with less-than-ideal results.
Facebook and Google often don’t have enough data to provide an accurate measurement, meaning that marketers end up with ads that are optimized based on just the first conversion. As a marketer, you understand that cost per acquisition (CPA) is only the “cost” side of the equation; the customer’s repeat buying behavior is the “value” side. Unfortunately, Google and Facebook have little visibility into your customers’ history, repeat purchase behavior, or the rich, versatile data you have collected based on your interactions with your customers. Knowing that, would you pay $10 more in CPA for a different customer that brings $100 more value? Of course! But to do that, you’d have to be able to measure and predict the value at the point of conversion. This is where expected lifetime value (ELV) comes into play, a metric I developed to pinpoint the most probable customer lifetime value (CLV) you can expect from a given customer who has followed a certain path to purchase.
In today’s world of artificial intelligence and advanced analytics, it is not only possible, but also very convenient, to compute these numbers — and view your campaigns through the lens of ELV. ELV differs from predictive CLV in one simple way: To compute predicted CLV, you must collect data from a few transactions – and possibly wait months – to determine the value of your campaign, whereas ELV relies on the signals available to you at or before conversion.
ELV enables us to combine signals to estimate a customer’s value in the marketing funnel pre- and at-the-moment of conversion. To estimate the value of a customer pre-conversion or at conversion, you must bring in signals such as demographic data, click-stream/web-flow, marketing touch points, and types of products browsed and purchased.
A simple example can be seen in the illustration below. Consider three customers: John, Amanda, and Kelly.
Gender | Product Price | CPA | Type | Churn Probability | ELV | FB/Google ROAS | ELV ROAS | |
John | M | $45 | $25 | One-time | .87 | $53 | 1.8 | 2.1 |
Amanda | F | $10 | $25 | Subscription | .10 | $100 | 0.4 | 4.0 |
Kelly | F | $15 | $25 | Subscription | .22 | $68 | 0.6 | 2.7 |
In the example above, both Google and Facebook will compute the return on ad spend (ROAS) by dividing product pricing by CPA, which would suggest that John has the highest ROAS — and that the business should acquire more customers like John. However, when you look at the ELV-based ROAS for Amanda, who spent the least on her first transaction, it turns out she is going to give us the highest return on our ad spend.
As a marketer, you shouldn’t believe that all of your customers are of equal value. If you have a good internal data science team, work with them to figure out how to compute ELV for your customers using your rich customer data. You can also come up with your own simple model by estimating the probability of churn and average order value for different customer segments and computing ELV based on the signals available to you at the time of conversion. A more sophisticated model may be able to estimate a more accurate lifetime value for each customer, and thus more accurately compute ELV. At Retina, we live and breathe customer value and can also answer questions about how your business should think about this value.
We recommend you take the first step by estimating your conversion ELVs for all of your customers, then take a look at the ROAS of your Google and Facebook campaigns. My guess is you will be surprised.