The Challenge

Detecting anomalies in real time to serve customers.

A leading home services website updated the way they track how members contact service providers. With this new data they want to capture when irregular instances in this contact data appear so they can investigate issues in near real time.​ Credera was engaged to create the system to detect these anomalies as they appear and alert the correct teams to initiate action if needed.​

The Solution

Find the right machine learning and artificial intelligence technology for the long run.

Credera and the home services website company leveraged many technologies to find the right solution to accomplish the organization's goals.

• Implemented platform for artificial intelligence in current AWS environment by onboarding them to AWS SageMaker for anomaly detection​.

• Created multiple machine learning models and delivered to production the one that most maximized the criteria defined for a successful model in this instance​.

• Investigated and delivered anomaly detection for first use case of number of contacts to service providers throughout the day​.

• Provided analytics expertise in machine learning to guide the employees through the process of implementing artificial intelligence projects. Now that the platform and best practices are provided, the organization can accomplish other machine learning goals.

The Results

Easily accessible data that keeps the organization up-to-date on anomalies while continuously learning.

Through the partnership, the organization achieved near real-time status on anomalies that are continuously learning to provide the best customer experience.

CLARITY INTO WHAT IS NORMAL VS. ABNORMAL

No longer is there an executive checking Kibana dashboards over the last few days to make educated guesses. The system is aware within the hour if something truly anomalous happens​.

CONTINUALLY LEARNING SYSTEM​

Weekly retraining of machine learning models means the system is continually learning and making incremental improvements in accuracy​.

AUTOMATED DETECTION​

All outliers are discovered in near real time instead of over the course of a couple days.​

IMMEDIATE ALERTING TO CORRECT TEAMS​

Leverages current alerting platform to get notifications to the proper team as soon as the anomalous event is detected.​