The pattern under most production personalization infrastructure: call the model or scoring service, and if it doesn't return within the latency budget, serve a deterministic fallback (rule-based, default segment, or static experience). The fallback isn't a failure case, it's the design that makes the recipe robust enough to deploy in production at all.
The capability a stack needs to run this pattern, and the vendors that provide it, on Martech Stack Builder.
Record why a particular customer saw a particular experience, in a form a regulator, an internal model reviewer, or the customer themselves can reconstruct months later
Warn the customer before the overage, the roaming charge, or the plan-mismatch cost lands on the bill, so the conversation is about what to do next rather than about why the bill is what it is
Pick the cheapest channel that achieves the outcome (owned email before owned push before paid social) rather than firing every channel by default, and reserve the expensive channels for the moments they earn
Enrich the submission with company size, industry, and role at the moment the form lands, so routing and personalization work on the enriched record rather than on what the prospect typed
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
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
Identify the company behind anonymous visitors via reverse IP lookup, with the honest handling of the false-positive rate that decides which use cases the match is good enough for