The predictive models LTV.ai scores every customer with
LTV.ai runs a suite of per-customer predictive models, calibrated so a 0.72 score means a real 72% chance, and degrades safely on small brands.
Quick answer: LTV.ai runs a suite of per-customer predictive models that score each customer on likelihood to act, then bucket them into named, targetable segments. Predictions are calibrated so a 0.72 score behaves like a real 72% chance, and models report insufficient data rather than guessing on brands with too little history.
What the models predict
The suite covers email click likelihood, purchase probability, churn risk, lifetime value, best send-time window, next purchase date, a composite engagement score, discount sensitivity, and follow-up propensity. Each emits a per-customer score plus a segment label (for example champion through dormant, or critical through safe on churn), and those segments feed audience building and smart inclusion and exclusion lists.
Why calibration is the point
A model that ranks customers is useful; a model whose scores are calibrated probabilities is far more useful, because downstream tools treat a score as a real likelihood, not just a ranking. LTV.ai calibrates its predictions so a 0.72 means roughly a 72% chance, which is what makes composing audiences with them trustworthy.
Honest validation and safe failure
Models are validated out of sample and report standard quality metrics, so their accuracy is auditable rather than assumed. On small brands, minimum-data floors cause a model to report insufficient data instead of training on noise. The system uses only first-party behavioral data, email and order history, not demographic or third-party data.
Frequently asked questions
What data does it use? Each customer's own email and order history. No third-party or demographic data.
What if a brand is too small? The model reports insufficient data rather than producing an unreliable score.
Part of the machine learning behind LTV.ai, a series on how the platform works under the hood.
See it on your store: book a demo.