🚀 Training ServiceNow Machine Learning Models with Fewer Than 10,000 Records

ServiceNow’s Predictive Intelligence empowers organizations to automate tasks, classify records, and route issues intelligently using machine learning. However, when attempting to train a new ML solution, especially in development or lower environments, you might encounter the following error:

Error: “Recommended setting for solution training requires minimum 10000 records. Please use the system property glide.platform_ml.api.csv_min_line to train with a smaller number."

This article walks you through what this message means and how to configure your ServiceNow instance to allow model training with smaller datasets.

🧠 Why This Error Occurs

Out of the box, ServiceNow enforces a minimum of 10,000 records for ML training. This ensures the algorithm has enough data to learn meaningful patterns and produce accurate predictions.

However, during early testing, sandbox development, or small-scope use cases, it’s common to have fewer records. Fortunately, there’s a configurable workaround.

🛠️ Solution: Lower the Record Threshold with a System Property

To override the default 10,000 record requirement, you can create a system property called:

glide.platform_ml.api.csv_min_line