A small organization faced a challenge in identifying its customers accurately and efficiently when they interacted with different offices and branches. They had taken photos of their customers during their interactions, but these photos were not always grouped together accurately. In some cases, they may have had multiple photos of the same customer, but they were not linked together, which made it difficult for them to get a complete view of the customer's interactions and history.
This challenge led to various issues, such as inaccurate identification of customers, duplication of customer profiles, and difficulty in tracking customer interactions across different channels. This, in turn, resulted in poor customer experiences, missed sales opportunities, and a reduced ability to spot fraudulent activity or other profile-linked alerts.
Commercial
ML and AI, Risk and Fraud
Databricks, Python, Tensorflow, AWS
To solve the challenge, we helped the small organization by using a solution based on Python, AWS services, TensorFlow, and the Databricks platform. We loaded their existing collection of customer profile images into a dedicated AWS S3 bucket to serve as the input data for a new image recognition system. We created a gallery of customer profile images and trained a deep learning model to compare and match images from the gallery. The deep learning model was trained to identify subtle differences and similarities between different images accurately.
The image recognition system was then used to compare new images uploaded by the organization's branches internal users with the existing gallery of customer images. When the system identified two separate profiles that had images that were identified to be sufficiently similar, it flagged it for manual review by the organization's customer service team. The customer service team then reviewed the flagged profiles and manually merged them together into a single customer profile.
This approach allowed the organization to accurately identify their customers and link their interactions across different channels, resulting in a more personalized and seamless customer experience. The use of the Databricks platform and AWS S3 for storage also enabled the organization to scale their image recognition system to handle a large volume of customer profile images and improve the accuracy and speed of their customer identification process.
Improved accuracy and efficiency in identifying customers by removing over 10% of total number of profiles that were duplicates.
99% reduction in occurrences of duplicate customer profiles after introduction of duplication detection.
Improved customer service by providing employees with a more comprehensive view of each customer's interactions, leading to increased customer satisfaction.
Improved fraud detection capabilities by enabling analysis of data across previously disjointed profiles.
We’ll work with you to deliver a product built on technological experience, the research know-how, with built-in measurable results.