The concept of product bundling has taken off in the global airline industry. In the past 18 months, airlines including Delta, Air Canada, British Airways, United, Qatar and Virgin Atlantic have rolled out pilot programs offering their customers packages of extra features and add-ons – commonly referred to as ancillaries.
The idea: Take the customer experience to the next level by, for example, combining a seat, wireless Internet, and a beverage of choice for a single price. Customers are delighted, as these bundles offer greater convenience and in most cases discounts, while the airlines are celebrating the increased revenue.
An opportunity exists to take this concept to even greater heights, by providing dynamic bundling of products based on predicting the ancillaries most relevant to each customer.
Micro-Segmentation and Predictive Analytics Produce Relevant Package Offerings
To achieve dynamic bundling, airlines and other companies must go to a granular level of analysis, focusing their attention on micro-segmentation.
Statistically-driven micro-segmentation identifies critical drivers of demand and price sensitivity at the lowest levels. Infuse some predictive analytics powered by sophisticated mathematical modeling, and companies can focus their attention on package offerings to individual customers, ensuring they receive something that is both relevant and at a price point they will consider.
By combining micro-segmentation and predictive analytics, companies can unlock the potential of ancillary revenue streams by determining the probability of initial and follow-on ancillary purchases.
Three Models Generate Successful Product Bundles
Dynamic bundling requires a balance of art and science, but utilizes three core analytical models:
- Micro-Segmentation – involves dividing the customer target market into subsets based on previous customer purchase behavior measured at the most granular transaction level
- Discrete choice modeling – determines the probability of a customer selecting an offer from a list of finite options, and the probability of additional spend based on this selection
- Market Response Models – facilitates price sensitivity measurement to identify customer willingness to pay for individual ancillaries and product bundles
If available, loyalty program data should be leveraged to further improve offers based on individual preferences and past behavior. Additionally, performing a market basket analysis will determine which purchases correlate with, and increase the probability of, additional spend.
Sophisticated Mathematical Models Ensure Optimal Pricing for Bundled Ancillaries
For the company intent on instituting dynamic bundling, you first need to determine which ancillary product a customer is most likely to purchase, followed by the most likely additional items.
Measuring the probability of specific ancillary purchases is achievable using historical individual customer behavior and the behavior of similar customer segments. A hotel, for example, would want to know a customer’s loyalty member history, including the channel(s) through which they purchased rooms, the original room type, typical length of stay, number of guests, etc., and the ancillary purchases that coincided with these stays.
Based on these predictive analytics, potential offers can be established and prioritized based on a combination of the probability of purchase and profitability to the hotel. Profitability depends on the revenue that can be realistically expected based on a certain level of probability.
Leveraging demand forecasts, competitive data, and price sensitivity models helps to guarantee, in real-time, that the right price is set for dynamically changing packages.
To meet this objective, you must first start with Market Response Modeling (MRM) to measure the impact of price on volume and product trade-offs. In addition, companies should combine this information with forecasted demand and available inventory to calculate optimal prices – then use real-time information and business rules to adjust the prices of these products and bundles up and down as needed.
Vast Revenue Uplift Potential with Dynamic Bundling
Naturally, with any takeoff, there are bound to be challenges. For example, if not properly structured and without proper business rules, market response and discrete choice models could lead to counter-intuitive results. Ongoing tuning and monitoring of model changes based on the latest data is necessary. Otherwise recommendations could become increasingly limited – resulting in a self-fulfilling prophecy.