The pattern for derived signals computed from observed behavior rather than directly observed: lifetime value buckets, propensity scores, lifecycle stage, predicted segment membership. Distinct from raw data because inferences age, drift, and need refresh cycles. Distinct from real-time decisioning because the inferences are persistent attributes available to many downstream recipes.
The capability a stack needs to run this pattern, and the vendors that provide it, on Martech Stack Builder.
Show account-relevant content to known and reverse-IP-identified B2B visitors, with explicit handling of the wrong-match case
Push first-party signals (predicted value, churn flag, recent conversion) to ad platforms as bid modifiers so the platforms stop paying premium for impressions that will not convert
Lower acquisition cost by feeding the bidding algorithms better conversions, deduplicated, consent-valid, and weighted by value, rather than just more of them
Surface adjacent products in the rhythm of how the original item is actually used, instead of dumping every cross-sell into one email
Give the AI a memory of each customer that is curated and decaying, not a raw event firehose it cannot use and should not keep
Let a model propose the next action when it can, and serve a known-good rule when it cannot, so the experience never depends on the model being up
Turn search queries, chat, support tickets, and reviews into structured intent signals, without mistaking sentiment for intent
Combine realized revenue with a predicted future, and carry the model's uncertainty into the decisions LTV drives, so the bid that depends on it knows what it is standing on
Surface the items the customer is most likely to want next, with explicit handling of the cold-start visitor, the privacy-restricted slice, and the rule that prevents the recommendations from collapsing to whatever everyone clicks
Reorder category results to the visitor's inferred preferences, with a default sort for unknowns and an honest signal of when sorting is personalized
Catch the customer while they are quietly disengaging, not after they have already decided to leave
Time the reorder reminder to the customer's own consumption cadence, not a category-wide average that is wrong for most of them
Feed customer lifetime value back into search bidding so the algorithm chases high-value acquisition, not just cheap conversions
Vary the renewal sequence and the incentive depth by predicted churn risk, instead of running the same series for every renewing customer
Match the content sequencing to the long consideration cycle the traveller is actually on, with the price-sensitive decision moment at the end where the booking either happens or moves to a competitor