CUSTOMER EXPERIENCE
Next best action: decide what to do next for each customer
Next best action with TIMi weighs commercial, retention and service actions for each customer and recommends one, with the reasons behind the choice.
One model per action
Offer acceptance, churn risk, service need: each question gets its own model.
Operational limits
Contact capacity and channel limits can be part of the decision.
Explained choice
Teams see why one action outranks another for a given customer.
HOW IT WORKS
How next best action works.
Three steps, in one engine, on a standard server with no cluster.
1. Prepare the data
Anatella gathers purchase history, customer interactions, risk signals such as churn indicators or complaints, and the list of possible actions.
2. Model each action
For each candidate action, Modeler builds a model that estimates its likelihood of success for a given customer, such as an offer being accepted or a customer being retained.
3. Arbitrate
The actions are ranked for each customer. The top one is recommended and the full ranking stays available for human review. Constraints such as contact capacity and available channels are handled with the optimization solver in Anatella.
HOW IT DIFFERS
Next best action versus next best offer.
The two are related, but they do not answer the same question.
A wider scope
Next best offer recommends a product. Next best action also covers non-commercial moves, such as a retention contact or a service call.
Models that compete for the same customer
The value is in comparing the models for one customer. A retention action may outrank a sales offer for a customer at high risk of leaving.
Execution is part of the decision
A recommendation only helps if it can be carried out. Contact capacity, channels and campaign calendars are inputs of the decision, not an afterthought.
EXPLAINABILITY
An action is easier to take when you know why.
A recommended action alone says what to do, not why this one rather than another. A white-box model shows the factors behind the priority for this customer, such as a high risk of leaving, a recent sign of dissatisfaction or an identified commercial opportunity.
For sales and service teams
An action can be launched with the argument that justifies it, instead of being followed without context.
For transparency requirements
A model that justifies each result variable by variable gives you the documentation to answer transparency expectations such as GDPR and the AI Act. The compliance assessment itself stays with your legal team.
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 next best action model uses.
Anatella connects to more than 80 sources, so this data can be gathered without prior custom development.
Purchase and interaction history
What the customer bought and how they have been in contact.
Risk signals
Churn indicators, complaints and signs of dissatisfaction.
Catalog of possible actions
Commercial offer, retention contact or service action.
Operational constraints
Contact capacity and the channels available.
PROOF
Built on models already in production.
We do not yet have a published customer story dedicated to next best action. These deployments combine the models it relies on: detecting needs, retaining customers and making offers.
Credit offers aimed at the 15% of clients most likely to need funding
ING contacted these business clients at two key moments of the year and raised conversion by 20%.
Attrition cut from 10% to 8% a year
A second model identifies the riskiest business clients, so the bank can contact them before they leave.
Retention offers triggered by churn risk
Delhaize flags churners across 37 product categories and uses its coupon engine to send them relevant offers.
FAQ
Common questions.
How does next best action differ from next best offer?
Next best offer recommends a product or a commercial offer. Next best action is broader: it can arbitrate between a commercial offer, a retention action and a service action, depending on what is most relevant for the customer at that moment.
How does the model choose between several actions?
Each candidate action is scored by its own model. The scores are compared and the action with the best expected result is recommended. The full ranking stays available for human review. The value attached to each outcome, for example a sale versus a retained customer, is set by the business.
Do we need one model per type of action?
Generally yes: one for churn risk, one for commercial opportunity, and so on. Next best action consists of making these models compete for the same customer instead of reading them separately.
Does it require integration with our contact channels?
The recommendation itself is produced in Modeler. Triggering it (email, call, notification) depends on the systems you already use, and Anatella can interface with them.
Can teams override the recommendation?
Yes. The full ranking and the factors behind it are available, so a team member can pick another action when the context calls for it.
RELATED USE CASES
Keep exploring.
Next Best Offer
Rank the offers each customer is most likely to accept, with the reasons why.
Churn Prediction
Score every customer on the risk of leaving and see the variables behind it.
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.
