# What the Downtime Number Actually Means ## The number is the cost of the problem. It is not the total market. A headline figure of approximately **$1.5 trillion** a year is in use in this field. It comes from an institute study on the cost of unplanned downtime. The figure shows the **losses of the 500 largest companies in the world**. It does not show the revenue that is available to a software vendor. > [!info] What the research of the institute shows > - The 2022 edition gives almost **$1.5T** a year in downtime losses. This is approximately **11% of revenue** for the Fortune Global 500 (Siemens / Senseye report on the cost of downtime). > - The 2024 edition gives approximately **$1.4T** a year. An average large plant loses approximately **$253M** a year. The average recovery time increased from approximately **49 minutes to 81 minutes**. The report compares approximately five years before with 2023. > - The 2024 report says that many businesses did not keep skilled maintenance labor after the post-pandemic resignation wave. The report says that this caused missing skills and missing knowledge, and that this led to longer recovery times. > > We verified these numbers against the published 2024 report in October 2026. The 2022 figures come from secondary sources. This document gives them as these sources reported them. ## There are three layers between the headline and a revenue line 1. **Scope**: The figure includes only large enterprises. A vendor that sells to mid-market manufacturers is not in the group that the study measured. 2. **Attribution**: Missing know-how causes only a part of downtime. Missing know-how also makes only a part of downtime longer. Problems with equipment, supply and planning cause all other downtime. 3. **Capture**: For the share that know-how causes, software removes only a fraction of the downtime. The buyer pays only a fraction of the benefit. > [!danger] A mistake with the figure > Some materials show the figure as the cost of the problem of knowledge that the plants do not keep. These materials mix two different things: the cost of downtime and the cost that missing knowledge causes. Documents show the first. The second has no number in the sources that we reviewed. The more recent edition also gives a lower headline number. Thus the materials use the older and larger number. ## Why it matters The number shows that the pain exists. The number does not show the total market. The only correct estimate of the market is a bottom-up estimate. The estimate is (plants that the company gets) × (realistic annual contract value) × (adoption). This estimate relates directly to [[Industrial AI Unit Economics]] and [[Deployment Velocity]]. ## Related - [[Foreman MOC]] - [[The Tribal Knowledge Thesis]] - [[Open Questions and Numbers to Reconcile]] - [[Industrial AI Unit Economics]]