The movie theatre industry in Portugal has been characterized by a small number of operators and very infrequent changes in price over the past decade. The impact of pandemic and the shutdown of nonessential businesses for a prolonged period had a devastating effect on this industry. In addition to the decline in revenue, these circumstances created more room for external competition, like streaming services, to cement their position as a very popular alternative for entertainment. These factors led to a need for a revaluation of pricing policies, now accelerated by this new environment of rising operational costs.
A large industry operator partnered with LTPlabs to develop a pricing strategy that would maximize overall revenue for both box office and bar products (e.g., popcorn, beverages, snacks). This exercise is particularly complex due to the following factors:
Prior to the implementation of any pricing strategy, it is crucial to assess the magnitude of the impact that changes in price have on customer behavior and overall demand.
Due to the infrequent price revisions on both the box office and bar, a simple statistical analysis of historical data could not provide robust enough results that could simply be extrapolated to the future. Nonetheless, a simple regression model controlling for a few fixed effects (e.g., time, inflation) could still be used as a baseline to contrast with real price elasticity.
The approach agreed to provide the best results was to test multiple price combinations on different rooms of every complex in Portugal. Taking into consideration that the analysis of historical data revealed a strong monthly seasonality effect in demand, we agreed on a 6-week testing period to get at least a full month of data.
The goal of this pilot was to not only measure the overall impact of changes in price, in an interval from 10% below the original price to 10% above that price, but also to understand how factors like location, blockbusters, and room size could smooth or accentuate the magnitude of those elasticities.
We identified 7 key demand drivers and used them to characterize each room on all complexes. The rooms were then segmented into different clusters based on their profile using first a density-based clustering algorithm (to remove outliers) and then a distance-based algorithm to identify the profiles.
The desired homogeneity of each cluster was crucial to ensure that all testing groups were as balanced as possible, meaning that each group targeted for a relative price change contained as many rooms of each cluster as any of the other groups.
It became immediately clear following the pilot that a superficial analysis of the data would not output any immediate conclusion. The impact of confounding effects such as movie premieres during this period meant that in some cases an increase in price led to higher demand. This highlighted the importance of continuously iterating with the client and developing a deep understanding of the operations to allow the careful removal of a few outliers without introducing bias in the data.
An analysis of the variation in both box office and bar treated data from the various rooms led to the conclusion that demand is relatively insensitive to small price variations. This means that price increases entail an increment in unitary revenue that makes up for the reduction in demand.
Additionally, we identified further key factors (e.g., day of the week, location) which could be used to adjust the price setting for a given period (day, week) to maximize total revenue, opening the door for the implementation of a Dynamic Pricing strategy similar to what is being used in other industries (e.g., hotels, air transportation).
By: Paulo Sousa , Eduardo Ribeiro
