Pricing for Real Estate Auctions

Optimizing pricing recommendations to increase sales rates and revenue

Pricing assets with valuation models alone is insufficient

Traditionally, pricing in real estate is based on the notion of a fair market value as decided by either a human expert or an Automated Valuation Model (AVM). However, in a dynamic market setting such as in online marketplaces, this valuation-based approach proves insufficient, as it fails to account for changing market conditions, therefore leaving money on the table. Intuitively, a good pricing strategy should start high to test the waters and see if there is demand, but not too high that buyers are turned off altogether.

To address this problem of pricing in dynamic online marketplace, we partnered with one of the largest online real-estate platforms to build a market-driven dynamic pricing model. The model would inform sellers on optimal pricing decisions for their properties, with the goal of maximizing overall revenue.

Accurate and intuitive sellability model as the basis for pricing

To find the optimal price, we need to balance the price and the probability that the asset sells at this price. At the core, an accurate prediction and understanding of the probability of sale for any given price is instrumental to good pricing. We used Optimal Classification Trees to build the sales probability model.
  • High accuracy

    The model predicted the sale probability in unseen data with over 80% precision, better than human portfolio managers.

  • Meaningful reasons for sales outcomes

    The prediction logic was presented to and validated by experts, with their feedback incorporated to improve the model interpretability and performance.

  • Up-to-date market conditions

    The model is updated with data under current market conditions to ensure it is always applicable and relevant.

Illustrative Optimal Classification Tree to predict probability of sales

Dynamic multi-run pricing strategy

A unique aspect of real estate sales is the temporal structure, where potential buyers arrive over time, and the prices can be updated dynamically in response to the observed market conditions. This presents an analytic challenge: given these correlated sales opportunities, what is the best pricing strategy over time?

Using dynamic programming and outputs from the sellability model, we build a life-time value optimization model that recommends a sequence of prices, which is accompanied by the reasoning at each run.

Schema for the life-time value optimization model

Increased sales rate and revenue for porfolios

Our client is currently using the pricing model every day to identify mispriced assets and recommend better pricing. The porfolio managers find the recommendations intuitive and useful in supporting their conversations with sellers.

In a pilot study, the client identified that the new pricing tool could result in 18% additional sales and 24% overall higher revenue. A joint white paper is coming out soon that details the methodology and impact at the client site.

Unique Advantage

Why is the Interpretable AI solution unique?

  • Market-driven sales probability

    As opposed to just another valuation model, this model learns market appeal and sellability from historical transactions

  • Communication with evidence

    The tool provides portfolio managers with reasons that support price changes

  • Formal optimization

    Rigorous optimization to derive the optimal pricing strategy in a complex multi-period process

Want to try Interpretable AI software?
We provide free academic licenses and evaluation licenses for commercial use.
We also offer consulting services to develop interpretable solutions to your key problems.

© 2020 Interpretable AI, LLC. All rights reserved.