Booking Personalization System

Requests identify the trip by flightId and routeId, while the personalization tables read customer context fields such as customer.deviceType, customer.loyaltyStatus, customer.visitSource, and customer.bookingHistory; anything missing simply falls through to the default rows.

Screen size drives the first branch: a customer.deviceType of 'mobile' returns 'compactSearch', 'simplifiedNavigation', and 'touchFriendlyControls' with a 'compact' layout that hides extras, 'tablet' and 'desktop' rows step up to 'medium' and 'full' layouts with 'detailedFilters' on desktop, and a blank catch-all serves a 'standard' layout under the first hit policy.

Suppressing ancillary content on phones and reserving detailed filters for desktop follows conversion practice on airline booking sites, where small screens punish dense fare displays. These are UX conventions rather than regulatory requirements, so a carrier would tune the element lists against its own funnel data.

Tier recognition is a straight ladder on customer.loyaltyStatus with first hit: 'platinum' gets priority 1, a discountPercentage of 15, and features including 'conciergeService' and 'exclusiveDeals'; 'gold' and 'silver' step down to 10 and 5 with smaller feature sets; and both 'none' and the blank fallback land at priority 4 with 0 discount and a 'joinLoyaltyBanner'.

A three-tier elite ladder with escalating soft benefits mirrors how most frequent-flyer programs are built, and showing enrollment prompts to non-members instead of discounts is standard acquisition practice. The 15/10/5 percentages are plausible loyalty economics but a commercial choice, not anything mandated.

Channel attribution decides which promotion surface appears: a customer.visitSource of 'partner_site' carries the largest addDiscount at 5, 'social_media' gets 3 with 'socialMediaDeals', 'email_campaign' gets 2, 'search_engine' shows 'searchSpecials' with 0 extra, 'direct' traffic sees no special offers at all, and a blank fallback serves 'standardOffers'.

Giving the deepest incentives to partner and social traffic while direct visitors get none reflects plain channel economics: direct bookers already show intent and need no inducement, while paid and referred channels are where airlines spend to convert. The exact 5/3/2 spread is marketing calibration, not an industry rule.

Past behavior is bucketed on two counters read together under first hit: totalBookings above 10 with recentBookings above 2 marks a 'frequentTraveler' with showRecentlyViewed switched on and a discount of 3, more than 5 bookings with recent activity is a 'regularCustomer' at 2, any history with recentBookings == 0 is a lapsed 'returningCustomer' at 1, and zero bookings makes a 'newCustomer' with nothing extra.

Pairing a recent-bookings window with lifetime volume is the classic recency-frequency split airlines and travel platforms use to tell active frequent flyers from lapsed ones, and reserving the recently-viewed module for active customers avoids surfacing stale content. The cutoffs of 10 and 5 bookings are sensible segmentation choices rather than fixed industry standards.

Score assignment cross-tabulates three signals: whether customer.bookingHistory.totalBookings is above 0, the loyalty-derived customerPriority, and whether customer.visitSource is 'social_media' or 'partner_site'. Under first hit, a priority-1 member with history arriving from a promoted channel scores the maximum 65, the same member from other channels scores 50, each lower priority tier sheds roughly 10 to 15 points, first-time visitors with top status still earn 30, and the bare catch-all bottoms out at 5.

Weighting booking history and elite tier ahead of arrival channel matches how personalization teams rank signal quality, since owned behavioral data is more predictive than referral context. The point values themselves are arbitrary rungs on an internal scale; what matters is that they line up with the 50/30/10 template thresholds applied in the next step.

Three closing values wrap the run: totalDiscountPercentage sums the base, additional, and history discount percentages, bookingFlowTemplate maps personalizationScore through breaks at 50, 30, and 10 into 'premium', 'enhanced', 'standard', or 'basic', and allFeatures flattens uiElements together with loyaltyFeatures into a single list. Summing the three discounts additively is what makes the earlier tables composable instead of mutually exclusive.