Was ist Untitled

---
title: "Underfitting"
slug: "underfitting"
type: "glossary"
---
Definition
Underfitting occurs when a machine learning model is too simplistic to adequately capture the underlying structure and patterns within the data. Consequently, the model delivers insufficient predictive performance on both the training data and new, unseen data.
Explanation
In the B2B and SaaS landscape, underfitting causes business-critical AI applications to overlook crucial relationships, such as complex interaction patterns within Customer Relationship Management (CRM) systems. For instance, if a model aims to predict the success of sales activities but only accounts for one-dimensional, linear relationships, it will severely oversimplify reality. The root cause typically lies in insufficient model complexity, inadequate training data, or overly restrictive regularization. To address underfitting, organizations can implement more complex algorithms (e.g., deep neural networks instead of linear models), enhance feature engineering, or relax regularization constraints. Only by overcoming underfitting can businesses ensure that AI-powered analyses deliver valuable and nuanced insights.