---
language: en
title: "Machine learning"
description: "Machine learning trains models on data to predict, classify, or generate content. Learn how to evaluate examples, performance, and limitations in marketing."
canonical: "https://tropk.ai/en/resources/glossary/machine-learning"
last_modified: "2026-10-12T16:13:41.302Z"
---

# Machine learning

Reviewed on 2026-10-06.

## What it means

Machine learning is an approach in which models are trained on data to produce predictions, classifications, or content. During training, the model adjusts mathematical relationships based on the examples or signals it receives. These relationships are then used to process new inputs. The goal is to achieve useful performance beyond the examples it has seen.

## Why it matters for marketing

In marketing, the concept appears in demand forecasting, audience segmentation, and identifying contacts who are more likely to buy. Quality depends on the data and how the task is evaluated. A model that works on older historical data may fail with a different offer or audience, making it risky to treat its prediction as an individual certainty.

## Example in practice

Consider a team training a classifier to separate sales inquiries from support messages. In the historical examples, almost all sales inquiries contain the word “quote.” When testing new messages, “I need a quote for fifty units” is classified as support. The team decides to expand the examples and review errors by category before automating the routing.

## How to apply or verify

Define the expected outcome and check where the data and the responses used as references come from. Set aside examples that were not used in training to evaluate the model. Examine error types, segments, and time periods, as well as an overall average. After adoption, monitor changes in inputs and maintain a way to correct inappropriate classifications.

## Important distinctions

Supervised learning uses examples with known answers; unsupervised learning looks for patterns without those answers; reinforcement learning uses reward signals. Training differs from inference, when the model processes an input. Not every application learns continuously from each use. Adjusting a [prompt](/en/resources/glossary/prompt) or publishing a page is not the same as retraining the assistant model.

## Sources and references

- [What is Machine Learning?](<https://developers.google.com/machine-learning/intro-to-ml/what-is-ml>) — Google for Developers
- [Datasets: Dividing the original dataset](<https://developers.google.com/machine-learning/crash-course/overfitting/dividing-datasets>) — Google for Developers
- [Machine Learning Glossary: inference and training](<https://developers.google.com/machine-learning/glossary#inference>) — Google for Developers

## Continue a leitura

- [Training data](</en/resources/glossary/training-data>)
- [Inference](</en/resources/glossary/inference>)
- [Artificial intelligence](</en/resources/glossary/artificial-intelligence>)

## About the author

[Vinicius Pacheco](</en/team/vinicius-pacheco>)

Responsible for Tropk’s technical content, focusing on crawling, indexing, structured data, and agent access to content.

---

[All pages in Markdown](<https://tropk.ai/sitemap.md>)