CUSTOMER VALUE
Customer lifetime value prediction: what each customer will be worth
Customer lifetime value prediction with TIMi estimates the future value of each customer and shows what drives it, so marketing and sales budgets follow the nature of the value, not only its size.
Prepared in Anatella
Purchase history, contracts and interactions brought together in one analytical table.
Built on value states
Models estimate how likely each customer is to move between value states, churn included.
Explained per customer
See which factors drive the estimate, such as purchase frequency, basket size or tenure.
HOW IT WORKS
How customer lifetime value prediction works.
Three steps, in one engine, on a standard server with no cluster.
1. Prepare the data
Anatella gathers purchase history, contract data and customer interactions, cleans them and structures them into variables the model can use.
2. Model the transitions
Customers are placed in states, such as high-value, medium-value, low-value or churned. States come from business rules or from a segmentation built in Stardust. Modeler builds one model per transition, for example the probability that a medium-value customer becomes a low-value one within 3 months.
3. Project the value
Your business defines an average value for each state. Anatella combines the transition probabilities with these values to project the value of each customer over 3 months, 6 months or more.
GETTING IT RIGHT
Pick the method that fits your data.
Lifetime value can be built in more than one way. The right one depends on how you already segment customers and how long your history is.
State transitions
Best when you already work with value segments. It projects value step by step and shows how the customer base moves between states. The number of models grows with the number of states.
Direct estimate of future revenue
Modeler also handles continuous targets, such as profitability or share of wallet. A model can estimate future revenue directly when your history is long enough to observe it.
Keep value variables out of the predictors
Variables such as the tariff plan or the lifetime value itself work better to define segments than as direct inputs. One model is then built per segment.
EXPLAINABILITY
A number says how much. The drivers say why.
A white-box model shows which factors weigh most in each customer’s estimate. A customer who is valuable because they buy often and one who is valuable because their basket is large call for different commercial actions.
For marketing and sales
Acquisition, retention and dedicated service budgets can follow the nature of a customer’s value, not only its amount.
Combined with churn
Crossing value with churn risk isolates high-value customers likely to leave, a segment that calls for priority action.
For model owners
Modeler’s built-in Drift Report shows when a model loses accuracy, so you update it when needed instead of on a fixed calendar.
DATA
What data a lifetime value model uses.
Anatella connects to more than 80 sources, so this data can be gathered without prior custom development.
Purchase history
Frequency, amount, regularity and tenure.
Contract data
Offer type and commitment length.
Customer interactions
Support requests, complaints and the channels used.
Behavioral signals
How the use of the product or service evolves over time.
PROOF
Built on models already in production.
We do not yet have a published customer story dedicated to customer lifetime value. This use case rests on churn models, which TIMi customers run in production.
Attrition cut from 10% to 8% a year
Contacting the riskiest business clients identified by the model saves 7,000 more professional accounts every year.
Churn detection accuracy raised from 45% to 57%
MTN analyzed 6 months of data for 21.4 million subscribers in under 5 days, using Social Network Analysis.
1,013 variables flag churners across 37 categories
Delhaize detects disengagement early enough to act, and pairs detection with relevant offers.
FAQ
Common questions.
Is the predicted value an average or an individual estimate?
The transition probabilities are computed for each customer, so each customer gets a different projection. The value attached to each state is set by your business.
Can lifetime value be combined with a churn score?
Yes. Both models can be crossed to identify, for example, customers with a high future value and a high risk of leaving, a segment that calls for priority action.
Do we need a long history?
A longer history improves precision. With a shorter one, a model can rely on the behavior of similar customers already observed, usually at the cost of a wider margin of error.
How many models does it take?
In the state-transition method, one model per transition between states, so the number grows with the number of states. Modeler builds these models automatically with AutoML.
How often should the value be recalculated?
Customer behavior evolves, so regularly. Modeler includes a Drift Report that shows when a model has degraded, so you update it when needed instead of on an arbitrary schedule.
RELATED USE CASES
Keep exploring.
Churn Prediction
Score every customer on the risk of leaving and see the variables behind it.
Next Best Offer
Rank the offers each customer is most likely to accept, with the reasons why.
Cross-Sell Prediction
Find the customers most likely to buy a complementary product, and why.
SEE FOR YOURSELF
Test it on your own data.
Run TIMi on your own data with a free trial, or talk to us about your case.
