Dynamic Pricing on a two-sided Marketplace for Travelers

Developing a dynamic pricing feature for a two-sided marketplace

In times of a booming travel and tourism industry, it is essential to offer outstanding products and services at competitive prices in order to differentiate your platform from competitors in a market where competition is fierce. One of our customers, a scale-up who is developing a two-sided online marketplace in the travel industry was struggling to gain a competitive advantage over its competitors. After they got inspired by how Airbnb is supporting their hosts to establish a dynamic pricing strategy reflecting the changing market conditions and how they help them to optimize their overall revenue. However, they had no data science skills in-house, and no clue where to begin. Therefore they reached out to Data Factory to partner up and launch this project. Our objective was not only to develop a dynamic pricing feature, but to prepare them for innovating further, by building a team and bringing them to the next maturity level.

Contrary to Airbnb, our customer did not have the knowledge and skills of hundreds of experienced data scientists, engineers and related profiles, and significantly less data. This meant that our team from Data Factory had to come up with a specific approach to get this job done. To tackle this challenging project with no available solutions on the market, we decided to apply for an Innoviris R&D ("shape") subsidy. This government funded subsidy enables our customer to invest considerably more time and resources into the project. Thereupon, we recruited a data engineer and data scientist, on-boarded them, and coached them while working on the dynamic pricing features.

To get optimal outcomes from the dynamic pricing feature for each day and activity, 3 different models are working together.

  1. The booking probability is predicted, based on a binary classification model.
  2. The model is trained on features about the activity itself, about supply and demand of visitors on each date, and on temporal effects like weather.
  3. The optimal price is predicted, which balances between the booking regret, and price decision recall.

Finally, the price is personalized based on specific requests from the customer.

Due to this dynamic pricing feature implemented on our customer's two-sided marketplace, travelers can be assured that they pay the optimal price. Moreover, it helped our customer to maximize revenues, the Annual Recurring Revenue already having doubled within the first 6 months after the launch of the dynamic pricing Minimum Viable Product. Also, our customer now has a team that is not only capable of improving this dynamic pricing feature further, but that can now also innovate further along other topics.