Glossary · 28 Sep 2026

What is machine learning?

← Glossary

Machine learning is software that learns patterns from past data and uses them to make predictions about new data. IBM defines it as "the subset of artificial intelligence (AI) focused on algorithms that can 'learn' the patterns of training data and, subsequently, make accurate inferences about new data". Instead of a person writing the rules, the algorithm finds them in examples.

How it works

A model is trained on examples where the answer is known: past loads with their weights and volumes, past logs with their grades, past weeks with their harvest totals. It learns the relationship, then applies it to cases where the answer isn't known yet. Its accuracy is measured on examples it didn't see in training.

In a timber business

Machine learning is used to predict harvester productivity, grade logs from images, forecast harvest volumes and prices, detect defects on boards, and flag unusual transactions. In each case, the quality of the prediction depends on the quality and quantity of past examples.

In our view, the number of usable examples is often the tighter limit. A 2022 study of harvester productivity started with 144,973 machine records collected over 28 months and tested 24 algorithms. The authors write that "After cleaning the initial database, we used only 1.12% to build the model". The best model explained about 71% of the variation. The abstract doesn't say what the cleaning removed. Some records may have been faulty, and some simply outside the study's scope.

What it isn't

Machine learning is one part of AI, not all of it. A large language model is one kind of machine learning, trained on text. Any model learns only what its examples contain, including their errors.

What to check

Before choosing a model, count your usable examples: records that are complete, consistent and linked to the outcome you want to predict. Ask how accuracy was measured, and on data the model hadn't seen.

How Quarri works explains the platform as a layer over existing systems, not a migration.

Sources

  1. IBM, "What is machine learning?", updated 2 April 2026: ibm.com
  2. Munis and others, "Machine Learning Methods to Estimate Productivity of Harvesters" (title shortened), Forests 13(7): 1068, 2022, abstract via Crossref: doi.org

Quarri is an AI-native data platform for the timber supply chain. It connects buying, production, sales and inventory for forest management, sawmill, wood products and pulp, paper and packaging operators.

See it on your own data.

Live in two weeks, on the systems you already run.