case study

Car IQ

Refuelling Revolution: ML-based Predictive Analytics for Car IQ

Business Impacts

Achieved 70-80% Average F1 score in predicting transactional vehicles across multiple timeframes

Implemented an efficient end-to-end MLOps pipeline, automating ingestion, data processing, model tuning, and result generation

Developed three comprehensive Looker Studio dashboards featuring 17 KPIs, offering detailed insights into model performance and results

Customer Key Facts

  • Country : USA
  • Industry : Fintech
  • Website : www.cariqpay.com

Problem Context

Car IQ, a fintech startup, specializes in streamlining cardless transactions at fuel stations for drivers associated with fleet management companies.

Car IQ’s primary objective is to delve into and decipher the behavioral patterns exhibited by drivers through vehicle-level data analysis. This initiative aimed to predict the likelihood of a vehicle engaging in a transaction and determine the anticipated timeframe for such transactions. Additionally, Car IQ sought to implement a recommendation logic capable of suggesting the nearest available fuel stations to vehicles expected to refuel.

Challenges

  • Predicting the likelihood of vehicle transactions (refuelling), and motivating drivers to transact by providing them with concurrent recommendations
  • Implementing a comprehensive end-to-end MLOps pipeline to automate functions such as ingestion, hyperparameter tuning, model training, and versioning

Technologies Used

Cloud BigQuery

Cloud BigQuery

Cloud IAM

Cloud IAM

Cloud Logging

Cloud Logging

Cloud Composer

Cloud Composer

Cloud Storage

Cloud Storage

Looker Studio

Looker Studio

Google Source Repository

Google Source Repository

Vertex AI

Vertex AI

Solution

  • Quantiphi identified pivotal factors impacting transaction events by leveraging the client's data. Through meticulous data transformations and engineering, Quantiphi crafted a machine learning model using Vertex AI, predicting the probability of drivers refueling across various timeframes
  • A recommendation model was developed by Quantiphi to propose the closest viable fuel station for vehicles anticipated to engage in a transaction
  • A robust MLOps pipeline was swiftly developed by Quantiphi using Vertex AI, meeting accelerated timelines
  • Additionally, insights into data and model performance, along with output visualization, were provided through Looker Studio

Results

  • Quantiphi developed an ML model with integrated MLOps pipelines, achieving a F1 Score of 70-80% in predicting transaction events
  • The Car IQ team was equipped with Looker Studio Dashboards by Quantiphi, facilitating a comprehensive assessment of model results

"Car IQ benefited a lot from Quantiphi's expertise in delivering Gen AI and Predictive Analytics solutions. They helped accelerate our timelines to market and understand industry best practices. The Quantiphi team was highly proficient, flexible and professional which made working with them a seamless experience for us."

Khalid El-Awady, Chief Product Officer, Car IQ

Thank you for reaching out to us!

Our experts will be in touch with you shortly.

In the meantime, explore our insightful blogs and case studies.

Something went wrong!

Please try it again.

Share