CUSTOMER GROWTH
Upsell prediction: find who is ready for a higher offer
Upsell prediction with TIMi estimates, for each customer, the probability of moving to a higher or more complete version of what they already have, and shows the usage signals behind it.
Upgrade propensity
Modeler compares customers who already moved up with those still on the base offer.
Usage signals
Usage approaching the limits of the current offer is a typical signal to include.
Full base, one server
Every customer scored on a standard server, for each tier you want to evaluate.
HOW IT WORKS
How upsell prediction works.
Three steps, in one engine, on a standard server with no cluster.
1. Prepare the data
Anatella gathers usage history, the offer each customer holds today and the signals that approach the limits of that offer. The result is one row per customer, with a target column that says whether the customer moved up in the following period.
2. Build the model
Modeler compares customers who moved to a higher offer with those who stayed on the base offer, and estimates a probability of upgrade. With several tiers, one model per higher tier can be built.
3. Score and explain
Each customer gets an upgrade probability with the factors behind it, such as usage near the limits of the current offer or a profile close to that of customers who already moved up.
GETTING IT RIGHT
What makes an upsell model accurate.
The signal is in how customers use what they already have.
Capture usage near the limits
Volume, frequency and the features used show whether a customer is outgrowing the current offer. If these signals are not collected today, putting them in place comes before the modeling.
Define the prediction window
The target says whether the customer moved to a higher offer within a given period after the observation date, based on the profile the customer had at that date.
Use relationships between customers when you have them
Where a relationship graph exists, such as calls between subscribers, Social Network Analysis variables help upsell models too, not only churn and cross-sell models.
EXPLAINABILITY
A probability says who. The usage signal says why.
A probability alone tells you that a customer may move up, not why. A white-box model shows the factors for this customer, so the pitch rests on a real need, such as a limit being approached, instead of a generic upgrade offer.
For sales teams
Sales effort concentrates on customers with a high probability, with a pitch built on the factor the model identified.
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 an upsell model uses.
Anatella connects to more than 80 sources, so this data can be gathered without prior custom development.
Usage history
Volume, frequency and features used over time.
Current offer
The tier held today and the gaps with higher tiers.
Limit signals
Usage approaching the limits of the current offer.
Past migrations
Customers who already moved up, as a learning base.
PROOF
Built on the same propensity models.
We do not yet have a published customer story dedicated to upsell. Upsell follows the same propensity method as cross-sell, which these deployments show in production.
Upsell and cross-sell opportunities across 300,000 stores
The daily model also delivers upsell and cross-sell opportunities. Scoring runs at about 2.4 million scores per second.
Cross-sell revenue doubled in 15 days
30 product models, a campaign to 2 million customers and an offer allocation engine, built by 2 data miners on one standard desktop PC.
Conversion up from 4% to 34%
Models built per product rather than per product category are now the standard at BNP Paribas.
FAQ
Common questions.
What is the difference between upsell and cross-sell?
Upsell offers a higher version of the same product. Cross-sell offers a different, complementary product. Both follow the same modeling logic, with different variables.
Does it work for offers with several tiers?
Yes. A model can estimate the probability of moving to each higher tier, not only the one immediately above.
How do we know a customer is approaching the limits of an offer?
This kind of signal (volume, frequency, features used) must be present in the usage data sent to Anatella. If it is not collected yet, it is a prerequisite to put in place before modeling.
Does it apply in B2B too?
Yes. The principle is the same: the model looks for customers whose usage profile is close to that of customers who already moved up, in B2C or B2B.
RELATED USE CASES
Keep exploring.
Cross-Sell Prediction
Find the customers most likely to buy a complementary product, and why.
Next Best Offer
Rank the offers each customer is most likely to accept, with the reasons why.
Customer Lifetime Value Prediction
Estimate what each customer is worth over time, and what drives that value.
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.
