Optimizing for long-term value through CLV Attribution
Ideally all marketing activities should aim at attracting and retaining customers with the highest (possible) customer lifetime value (CLV). CLV is the net profit a customer contributes to a company over his entire lifetime. In essence, it is the sum of what a customer paid for a company’s products and services minus the costs of production and expenses for acquiring and retaining that customer. Ultimately the goal of all business is to maximize their CLVs.
CLV orientated marketing does not focus on short-term goals, such as a single transaction or attracting a maximum of new customers. Instead it maximizes the long-term value along the whole customer lifetime.
A holistic marketing attribution approach would hence lead to allocating marketing resources based on CLV rather than on short-term gains. For example, if channel A has a better CPA (Cost per Acquisition) than channel B, but channel B has better CLV based ROAS (Return on ad spend), then a CLV orientated approach would allocate more budget toward channel B.
Sum of revenues as a pragmatic proxy for CLV
Calculating the exact margins and profits per customer, which is necessary for an accurate CLV measurement, is quite costly. Integrating data from backend systems and data warehouses is often necessary but time consuming and tedious. Therefore, a pragmatic first step is to start with the sum of all revenues a client has generated within a 12-24 months period as a proxy for CLV.
This approach has many benefits compared to a short-term, single transaction focused CPO based marketing attribution and works well in most cases. There are instances, where the sum of these revenues differs substantially from a real CLV, e.g. because customers ordering frequently or higher priced product also return their products more often and cause disproportionate costs. In these cases it is recommended to actually integrate data from the DWH and to use operative margins for calculating CLV.
Data-driven attribution modeling as the basis for CLV attribution
In marketing attribution, and also when working with data-driven attribution models, often times the focus is on attributing the value of a single conversion to the previous marketing touchpoints. A marketing touch point usually isn’t credited with the value of more than one conversion or order.
Let’s take the following example: A customer had his first conversion and ordered after a display ad view and a following paid search click, on a generic keyword ad. After this first order and maybe a week later, he clicked on a brand paid search ad and following that, clicked on a retargeting ad, after which he ordered a second time. The short-term, non-CLV perspective would only credit the brand paid search ad and retargeting click with the value of the second order:
The simple CPO or ROAS perspective leads to only including the first order for calculating the effectiveness of the generic paid search ad and the display ad view.
The CLV orientated approach is to apply data-driven attribution modeling to calculate how much of the value of the second order should be attributed to the display ad view and the first paid search click. The attribution model should answer how much of the generic paid search click and the display ad view contributed to the second order:
This allows giving fair credit to marketing touchpoints that happen early in the customer journey and helped in acquiring a new customer. These marketing campaigns might look unprofitable when applying a short-term, CPO orientated approach, which would only credit them with the value of the first conversion.
The aim of CLV based attribution is to account for the long-term value of a new customer and give credit to marketing campaign touches before the first conversion for part of this long-term value, based on a data-driven attribution model. This allows for smarter, more advanced and efficient marketing decisions.
A data-driven and machine learning based attribution model capable of this sort of CLV attribution has to learn these interdependencies. Machine learning based attribution modeling isn’t trivial to begin with, even for the simplistic, short-term CPO attribution approach. Specifically generating suitable trainings sets and adequate sampling of non-converting journeys can be quite a challenge. But to incorporate a holistic view on the customer journey across multiple conversions and marketing touchpoints and to enable the models to “learn” what dependencies between those events exist is even more demanding.
Proprietary customer data as a competitive advantage in biddable channels
Following a CLV based attribution model, Google Adwords ads, keywords and campaigns, that have the best CLV-Return on Ad Spend should be preferred. This ultimately allows moving away from a short-term CPO or ROAS perspective. Following the CLV based attribution allows for much smarter bidding and for potentially outbidding competitors on highly competitive keywords:
Taking the example from the above diagram, an advertiser would be able to increase the bid for the generic paid search ad significantly, since the attributed value is not only $90, but $170: In highly competitive environments, with many advertisers bidding on the same set of keywords, a CLV attribution approach is especially relevant. This allows for outbidding competitors on important keywords and thus attracting more new and valuable customers.
Generally speaking, exploiting proprietary customer data such as the sum of revenues or the CLV per customer, facilitates a more differentiated and more competitive marketing budget allocation. As in the above case, only the advertiser able to track CLVs and attribute them across all marketing campaigns, can adjust the Adwords bids respectively and outbid its competitors. This does not only apply for Adwords, but for all biddable channels, such as Facebook, Instagram, Bing, almost all DSPs, Twitter, Linkedin or retargeting services.
Adtriba’s CLV attribution and Adwords integration
Adtriba’s attribution modeling is based on a holistic view of the customer journey, as shown in diagram 3. The holistic ROAS displayed in the Adtriba dashboard allows for adjusting the budget allocation based on CLV attribution.
In addition to displaying these performance metrics in the Adtriba dashboard, the CLVs attributed to Adwords campaigns, based on Adtriba’s attribution model, can be exported into Adwords on a Google Click ID (Gclid) level. This allows for a direct bid-adjustment in Adwords on all possible levels, from keyword- to adgroup- and campaign-level, all based on CLV attribution.
Adwords offers machine learning based "target ROAS" bidding as part of the smart bidding strategies, which optimize for conversion value in each and every auction. This target ROAS bidding can work on the CLV attribution values imported into Adwords from Adtriba. This is a powerful combination as it leads to automated and AI based CLV attribution and bid optimization:
CLV attribution requires cross-device tracking
Adtriba integrates with mobile app tracking services (e.g. Adjust, AppsFlyer) and cross-device identity services (e.g. Tapad). This allows tracking users across devices and attributing campaign effectiveness while accounting for cross-device journeys. Accordingly, CLVs are being attributed across the whole journey considering device switches. Otherwise there would be a skewed view on campaign performance. For example, let’s assume the second order in the above example took place on another device then the first two marketing touchpoints. Without cross-device tracking there would be no way to attribute part of the value of that second conversion to these marketing touchpoints, and we are back to a simplistic, non-CLV attribution.