Dynamic Packaging: An Implementation Approach with Machine Learning and Personalisation

For long, travel has been a connected experience wherein different service providers like air, road, rail, and sea as well as hospitality players come into the picture at different stages of a journey. However, from a traveller's perspective, he or she would love to have a single booking experience with all desired modes of travel and accommodation availed as a single package. This would save them the trouble of having to search for deals in each case and then match up timings and dates to accommodate schedules of each item. But for tour operators and travel companies, offering such a flexibility for customers for a hassle-free packaged booking experience is no easy deal. But technology has made this process very simple.

Think about a vacation booking solution that allows for endless flexibly in mixing or changing your tour content. Such solutions are powered by dynamic packaging systems. Dynamic packages have been around for a while now, and even though some might have thought it was just a fad, it looks like it is going to be around a little longer and beyond 2020. In fact, it is a market that keeps on growing. Customer experience personalisation also heats up the interest for packaged tour solutions.

So what exactly is dynamic packaging software? The ability to combine multiple products on the fly, depending on what suits you, and offering that as a product to you, thereby -- because of the opaqueness it can create as a product -- bringing in a price arbitrage or a price advantage to you as a customer. The main difference between tours as they existed historically and dynamic packaging is that users can configure their trips via a single packaging interface and receive a full price estimate based on individual pricing and availability of each component in dynamically generated tours. It allows OTAs and airlines offering their own directly contracted products from their inventory and/or products from external suppliers. The overall price of the package is derived by combining the prices of the underlying components, subject to personalisation, price adjustment and then presented opaquely to the traveller.

Here we focus on the multi-step model of search, selection, matching and packaging of different components using personalisation and dynamic pricing. Take the combination search of flights and hotels/other components that are selected in various steps.

Step 1 – Search

1. Search for flights for origin/destination for a given date.
2. Search for hotel or other product from the provider (this can be done in parallel).
3. There is a flight selector which looks for personalising flights based on user context (the flights could be shortest, cheapest, non-stop, no red-eye, etc
4. Based on the user-context, the flight is selected and hotels /other products are combined and it is ensured that the hotels are package-able with the chosen flight—we call this the winner flight.
5. If not, a new search will be initiated to meet the need.
6. Chosen flight and hotels are then processed through a rule engine which looks at various restrictions based on check-in–check-out requirements and pricing, where extra adjustments are also done, if needed (price mark-up or mark-down).
7. Hotel options are now presented to the customer for selection.

Step 2 – Search Selection with Lead Product

1. Once the hotel is selected, this hotel is matched to combine with all possible fight solutions again, considering the same rules used in point 6 of step 1.
2. All flight options are now presented to the customer.

Step 3 – Availability Check

1. Based on the user-selected flight and hotel, another price check is performed to make sure the inventory is still available.

Step 4 – Purchase

1. Based on the property type like car or hotel that gets attached to the flight, the payment processor could be external or from the airline itself, and so we need to consider that option as well.
2. Booking happens from the third-party or airline provider as individual products.

Dynamic packaging architecture is composed of three main layers, namely supplier integrations, dynamic packaging, business logic and UI layer. These layers are articulated below.

1. Supplier Integrations

Supplier integration is a foundation layer for any packaging engine. It defines the inventory availability for booking. The selection of suppliers depends on various factors including target markets (different suppliers provide access to inventories in different destinations), inventory availability, API support etc.

The main engineering problem here is inventory search consolidation and integration. The information is distributed via various APIs that differ both in terms of the inventory they offer and the technology. The common source of inventories for dynamic pricing systems are:

• Global Distribution System (GDS) GDS is a network system providing travel retailers with information covering airlines, hotels, and ground transfers. It returns basic data including availability rates and also allows for booking. Currently, GDSs cover a significant share of the market. A retailer that plans to integrate its service with the GDS must meet a number of criteria.
• Central Reservation System and Computer Reservation System (CRS) of hotels and airlines These systems consolidate the information about inventories on the supplier side. An OTA or tour operator can source inventories directly from a supplier. The main drawbacks of this approach are (a) legacy systems, (b) absence of their own APIs, and (c) the need to connect with many suppliers to achieve decent coverage.
• Wholesalers and bed banks These providers mostly cover the accommodation sphere. Wholesalers and bed banks suggest detailed information on travel inventories and provide modern API support. While they consolidate fewer suppliers than GDSs, the integration doesn't require additional certification. The systems here are usually more innovative than those of GDSs and suppliers.
• Events and attraction providers This category includes tour and activity OTAs, local providers, and B2B wholesalers that supply information and booking capacities for attractions such as museums, landmarks, and events.
• Car rental suppliers Besides integrating directly with local car rental services, a smart move is to consider the main car rental players as they offer both information and booking capacities via their APIs.

2. Dynamic Packaging Engine

Dynamic packaging engine combines different travel products to construct the package. The travel products are fetched from the local and external supplier interfaces. The overall price of the package is derived by combining the prices of the underlying components, subject to certain price adjustment and rounding rules, and presented opaquely to the traveller. The core components of the dynamic packaging engine are listed below.

• Dynamic package build and pricing rules Dynamic package pricing rules dictate how the opaque package price is to be derived from the prices of the underlying components. Price adjustment is applied on the package potentially to increase the package price to take advantage of high demand, for any other strategic or tactical marketing purposes, or to encourage users to book the package by offering it at a reduced price.
• Dynamic personalisation and recommendation This is a critical component of the solution providing a personalised end-to-end travel solution. It supports scoring and classification, and content-based and collaborative filtering-based personalisation.
• Revenue management This makes the dynamic packaging solution more competitive and uses different techniques like customer segmentation, supply and demand insights, yield and pricing data to achieve optimal sales.

Dynamic package build rules allow the tour operator to define, for lead products matching certain characteristics, what ancillary products are to be suggested to construct a dynamic package.

Allowed dynamic package types: User must be able to define type(s) of dynamic packages that the distributor/airline/OTA may sell at the destination.

• Dynamic package schedule rules
Define the payment schedule rules related to when the payments must be received from travellers (or their agents) for dynamic package bookings.
• Dynamic package agency commission payment rules Agency commission payment rules relate to the commission, if any, paid to a travel agency that makes a dynamic package booking on behalf of the traveller.
• Dynamic package revenue recognition rules Revenue recognition rules relate to the posting of revenue for a dynamic package booking to the company's general ledger system.

Dynamic package pricing rules dictate how the opaque package price is to be derived from the prices of the underlying components.

• Define dynamic package price adjustment rules They allow the tour operator to apply a subsequent adjustment to the dynamic package price that initially starts out as a simple sum of the best fit prices of its components.
• Define dynamic package fancy pricing rules When the net price of two or more travel products is marked up by some amounts (potentially of different percentages), combined, and then the combined price adjusted by perhaps another percentage, the result is likely to be an odd-looking number that doesn't "look" like a package price. The system allows flexibility to round prices up or down to a given price point and will refer to this as "fancy pricing".
• Define dynamic package price adjustment allocation rule Between the price adjustment rule applicable to a package and the fancy pricing rule, there is likely to be a non-zero amount that is to be added to or subtracted from the package price as calculated from the sum of the individual component prices. The package price is opaque to the end-user, and hence this net price adjustment can simply be applied to the overall package price for presentation in the shopping layer.
• Define dynamic package deposit calculation rule The dynamic package deposit calculation rules relate to how deposit requirements (if any) are calculated for dynamic package bookings. There can be a different rule for each supported dynamic package type, and/or a master deposit calculation rule applicable to all dynamic package types without overriding rules. The package's deposit requirement is the sum of the deposit requirements of its individual components or there is a dedicated deposit calculation rule specifically applicable to dynamic packages.

• Dynamic personalisation and recommendation engine consumes the data from the customer experience management system data or from various external and internal systems.
• It uses the data to get a 360° view of the customer and builds a persona using machine learning (ML) models by grouping and transforming the relevant attributes received from the data sources.
• Using the ML models, the engine can recommend dynamic packaging products by combining BRMS rules such as scoping, pricing, and prioritisation.
• The engine supports different personalisation recommendation techniques/types like classification, and content-based and collaborative filtering-based personalisation.

There are three basic personalisation recommendation types that can be used in the travel industry:

• Collaborative filtering This method suggests tailored travel products based on options picked by other users with similar characteristics.
• Scoring and classification This method divides users into broad segments by type of travel, location, budget, and destinations.
• Content-based personalisation This finds similarities between a traveller's previous shopping patterns and predicts the buying options.

Revenue management makes the dynamic packaging solution more competitive and uses different techniques like customer segmentation, supply and demand insights, yield and pricing data to achieve optimal sales. ML-based revenue management can be configured to make use of different types of data collected from different sources which are then analysed to build ML-driven revenue management systems. The system can analyse behavioural data, social media data, and market data, assuming the right prices for the right people at the right place.

The typical dynamic packaging system includes the following functional modules:

• Inventory management
• Booking system
• Payment system integration
• Multilanguage and multicurrency support systems
• Yield management or revenue management systems
• Workflow management
• Data import/export and reporting (business intelligence)
• Package validator – price and consistency of package

To enable dynamic packaging, all these components have to be cross-integrated with efficient data flow among them. So, the key barriers to integrating dynamic packaging in existing booking systems are the lack of consistency in the data collected and the slow exchange of information.

Creating a user interface with inspirational landing pages, personalised promotional campaigns and open search capabilities will inspire and engage travel shoppers. The user interface design can follow module architecture to logically break down the parts of the package and provide better visibility of price and schedule at each step. An explicit collection of user input and their interest along with a personalisation and simple rule can result in better booking rate. Chat bot or voice bot-based user interface can help and make booking process simple and personalised.

With a myriad of options available at their disposal, travellers look for differential or variety experiences that help them make a choice on their tour partner. Dynamic packaging with real-time availability, real-time pricing, and ML-based personalisation can better address the needs of online shoppers looking for different types of breaks during the year instantly. With personalisation aided by AI and ML, intelligent revenue management, and highly configurable business rules, dynamic packaging helps travel companies build a seamless transitional experience for travellers across different modes of transport in one place, broaden product offerings, and set them on course for improved profits from better conversion rates.