2022-06-05 | Article | Insights
Not all customers have the same value. The Pareto principle states that 20% of your customers account for 80% of your sales. Predicting Customer Lifetime Value (CLV) is a way to find out which of the customers make up that 20%, not only in the past, but also in the future, by answering these types of questions:
How many purchases will the customer make in a given period of time in the future? How much time will pass before the customer becomes permanently inactive? How much monetary value will the customer generate in a given future period?
Predicting CLV helps you decide how much to invest in advertising, build value-based buckets of customers to target with advertising, and set up experiments to find the trade-off between accuracy and scale and increase marketing efficiency. Clusters of high value customers serve as a foundation to extend your target audience based on similarity to users with high customer lifetime value.
You can build and use predictive CLV audiences deploying Google Marketing- and Google Cloud. The first step to start with the process is to ingest your training data from your CDP/ CRM/ e-commerce solution and your GMP accounts in Google Cloud BigQuery and process it using Dataproc so you can explore it. ID matching helps you connect the user data from the various sources based on mutual identifiers. You can group all available user identifiers in Google Analytics 4 into one BigQuery table to create an identity graph - that is a collection of user IDs such as device IDs that can be associated with a user based on their interactions with your website or app.
With SQL skills, BigQuery ML can be used to create, evaluate, and predict audience segmentation models as in this use case needed using SQL constructs. With data scientist expertise, the Vertex AI can be deployed to build and use optimized models at scale. After building, training, and deploying your machine learning model, you can obtain the CLV predictions.
All tasks are executed and orchestrated through Google Cloud Composer.
Using Query-time import you can enrich your data in Google Analytics by storing your predictions in a custom dimension and adding the values to the hits that have already been processed. Audiences are recreated after each data import, which makes Query-time import the best method for data enrichment. Consider building various scenarios as clusters such as top 10%, top 20%, etc. customers to be able to experiment with. As soon as your audiences are built-in Google Analytics, they are natively transferred to your linked DV360 account and are ready to use.
The built audiences can be then used to expand the outreach of your campaign towards new users who share similar interests with your high CLV audiences using DV360 similar audience targeting which is now part of the overarching feature of targeting expansion. While doing so, you can exclude the first party audiences that you used as foundation to find similarity to ensure a clean testing scenario for your experiments.
You can also use the predicted CLV audiences to sync them to your CDP solution for activation via newsletter or for instance with Meta ads.
The main pillars of creating tailored audiences based on CLV predictions and their activation across channels involve selecting relevant online and offline data, setting up an automated cloud infrastructure, and establishing an user identity graph.
With a robust infrastructure you are able to automate the process of utilizing data.
Building an user id graph for a matching of on- and offline first party data lets you use and combine your data from sources such as Google Analytics, DV360, CM360, SA360 and your CRM/ CDP and your e-commerce solution.
With an individual approach to the customers, the conversion rate and the efficiency of the marketing measures can be significantly increased, factoring in the uniqueness of customers and their purchasing behavior.