SITUATION​
Our client, a leading global beauty retailer invested in expanding its machine learning capabilities with a focus on employee performance optimization. Historically, beauty advisor (BA) performance data - including critical KPIs such as sales per labor hour (SPLH) and transactions per labor hour (TPLH) - was siloed across multiple systems. Leaders had to manually aggregate metrics to evaluate team performance, generate feedback, and identify appropriate training resources. To scale and streamline this process, our client envisioned a machine learning model to analyze current BA performance metrics, recommending individual training content based on those metrics to drive improvement.​

SOLUTION​
An Everforth Apex team of data engineers, machine learning specialists, and solution architects helped to design and deploy an AI-powered training recommendation engine to deliver personalized, outcome-driven content suggestions for BA’s.​

• Conducting a deep dive into historical BA performance data and existing training resources.​

• Implementing data filtering and transformation layers to ensure only high-quality, actionable data was used for model training. This data clean-up had the added benefit of making metrics readily available to sales leaders.​

• Applied feature engineering techniques to identify variables most strongly correlated with improved BA metrics.​

With foundation in place, our team proceeded to actually build the machine learning model, which included the following steps:​

• Built a fully automated ML pipeline using Dataiku, using its built-in collaboration features for model development and deployment.​

• Used MLFlow, Databricks, and PySpark for versioning, experimentation, and large-scale data processing.​

• Developed multiple models, selecting the one with the highest predictive accuracy for training content effectiveness.​

• The final model was deployed as a Flask-based microservice API, hosted securely and integrated with the internal systems via Azure.​

The service dynamically recommends tailored learning content based on real-time performance metrics, enhancing individual coaching and team development.​

RESULT​
The results of the AI-powered training recommendation engine implementation have been significant and include:​

• BA’s now receive personalized training content aligned to individual performance needs, increasing content relevance and engagement.​

• Early results from pilot stores showed a 9–15% increase in SPLH and TPLH as a result of targeted training.​

• The model continuously refines its recommendations based on updated performance outcomes, ensuring long-term adaptability.​

• Leaders no longer need to compile fragmented data from multiple systems to evaluate BA performance, allowing for more direct coaching in addition to the recommended training.