Cox Automotive - Providing a Transparent and Reliable Estimate of Car Repair Costs with ML

Team members:

Eric Wityk, Sr. Data Scientist, with:

  • Sepehr Piri
  • Ashkan Heydari
  • Megan Williamson

Country: United States

Organization: Cox Automotive

Cox Automotive makes buying, selling, owning and using cars easier for everyone. With our technology, market intelligence, and products and services, Cox Automotive simplifies the trusted exchange and mobility of vehicles and maximizes value for dealers, manufacturers and car shoppers. The global company’s team members and family of brands, including Autotrader®, Clutch Technologies,®, Dealertrack®, Kelley Blue Book®, Manheim®, NextGear Capital®, VinSolutions®, vAuto® and Xtime®, are passionate about helping millions of car shoppers, 45,000 auto dealer clients across five continents and many others throughout the automotive industry thrive for generations to come.

Awards Categories:

  • Best Acceleration Use Case
  • Best MLOps Use Case
  • Best Approach for Building Trust in AI


Business Challenge:

The automotive repair and maintenance industry is a complex one, with a wide range of prices for different services. This can make it difficult for consumers to know what they should expect to pay for a specific repair or maintenance service.

Kelley Blue Book (KBB) Fair Repair Range is designed to address this challenge by providing consumers with a transparent and reliable estimate of the cost of repairs. The Fair Repair Range is based on data from actual transactions, and it excludes data from warranted repairs. This means the Fair Repair Range is a reliable estimate of the cost of repairs.

Here are some of the business challenges that KBB Fair Repair Range addresses:

1. Building trust and engagement between consumers and repair shops

Because the Fair Repair Range is based on real-world data, consumers can be confident that the estimates are accurate. Additionally, the Fair Repair Range can help consumers to make informed decisions about where to take their vehicles for repairs. It also helps consumers to negotiate with repair shops for a better price. The Fair Repair Range gives consumers a starting point for negotiations so that they feel confident when they are negotiating with repair shops.

2. Customer attraction and retention

The lack of transparency can damage the reputation of dealers and service centers, making it difficult for them to attract new customers and retain the existing ones.

3. Vehicle ownership frustrations

The lack of transparency can also lead to friction in customer journeys. For example, customers may shop around for quotes or negotiate with dealers or service centers on prices. This can be a frustrating experience for customers and discourage them from returning to the same dealer or service center in the future.

Business Solution:

We created and automated multiple machine-learning models using Dataiku’s built-in tools. Dataiku’s drag-and-drop interface simplified the creation of data pipelines and integrated seamlessly with Snowflake data sources. Then, Data Science used the Visual Analysis functionality to prototype and evaluate different modeling approaches and automatically tune hyperparameters.

Using scenarios, we were able to automate the entire machine-learning lifecycle and refresh the model monthly. Dataiku also provided many useful tools for monitoring and fine-tuning these models.

Using dashboards, we could identify and remove suspect data from the source data. We also monitor the changes in predicted prices month-over-month and investigate any large price swings to identify any potential issues with our estimates. Finally, we were able to share high-level insights with stakeholders.


Day-to-day Change:

Before moving to Dataiku, the process of refreshing FRR was much more manual and time-consuming. Dataiku allows us to easily automate this process and monitor the results in a unified ecosystem.

Dataiku’s modular flow environment allows us to add new service and repair categories with ease. Finally, Dataiku facilitates collaboration among Data Scientists who create the models, as well as the engineering and model ops teams that productionize the process.

Business Area Enhanced: Product & Service Development

Use Case Stage: In Production


Value Generated:

The Fair Repair Range can help consumers and repair shops in several ways, including:

1. Building trust

The Fair Repair Range can help to build trust between consumers and repair shops. This is because the Fair Repair Range is based on real-world data, so consumers can be confident that the estimates are accurate. This can help consumers feel more comfortable taking their vehicles to repair shops, knowing that they are not likely to be overcharged.

2. Making informed decisions

The Fair Repair Range can help consumers to make informed decisions about where to take their vehicles for repairs. It can help them to choose a repair shop that offers fair prices and quality service. The Fair Repair Range also provides information about the factors that affect the cost of repairs, so consumers can understand why a repair might be more or less expensive than they expected.

The impact of this product has been recognized in the media and across the automotive industry. An article by Auto Remarketing on KBB’s Fair Repair Range quotes Tully Williams, the fixed operations director of The Niello Company, as saying,

“The complexities of how to maintain and repair a vehicle, as well as understanding the associated costs, are a major pain point for most consumers, […] Not only has our dealership increased service cost transparency and communication with our customers, but they also have an increased sense of trust from knowing they are being fairly charged for every service visit.”


Value Brought by Dataiku:

With Dataiku’s Visual Analysis tools, Data Science identified improvements to the machine learning model and automated hyperparameter tuning. Dataiku allowed Data Science to automatically refresh the model, saving time each month. This allowed Data Science to devote more resources to improving the Fair Repair models.

Dataiku improved the transparency of the model and input data, so suspect data points or model estimates can easily be identified and investigated. Finally, Dataiku’s modular user interface made the lift and shift process of moving SQL queries and Python notebooks from the prior platform to the Dataiku environment simple and painless.

Value Type:

  • Save time
  • Increase trust


Version history
Publication date:
01-08-2023 04:26 PM
Version history
Last update:
‎08-02-2023 09:34 AM
Updated by: