Ask why a machine was down for 45 minutes today and you'll get one of two answers. "It jams on this SKU about twice a shift and each jam runs 20 minutes to clear" is a problem you can fix. "It was down" is not. The gap between those two answers is the entire discipline of downtime tracking.
Planned vs unplanned
Planned downtime is scheduled before the shift starts: preventive maintenance windows, changeovers, tooling changes, no-order gaps. It's a cost, but it's a known and budgeted one, and it's typically excluded from the Availability side of an OEE score because that time was never meant to make parts.
Unplanned downtime is everything that shows up uninvited: breakdowns, jams, starves and blocks, material shortages, quality holds. This is where tracking effort belongs, because it's the category that's both controllable and where the real cost hides — a plant that has fully budgeted its changeovers can still be bleeding margin from unplanned stops nobody has bothered to categorize.
The reason code is the whole product
A downtime reason code is the short label an operator or a PLC attaches to a stoppage the moment it happens — "E12: infeed jam," "M03: blade change," "Q07: quality hold." Without it, downtime data is a duration log: the line was down for 45 minutes, full stop. With it, the same 45 minutes becomes something you can act on: infeed jams accounted for 12 stops and 40 of those 45 minutes this week, on one SKU, on one shift.
This is why the reason code list matters more than the software collecting it. A good list is short — 15 to 30 codes, not 200 — assignable by an operator in under five seconds, and mutually exclusive enough that two people watching the same stop pick the same code. A code list built by a committee that tries to capture every conceivable failure mode produces data nobody enters consistently, which produces reports nobody trusts, which is how downtime tracking software ends up abandoned six months after go-live. The tracking is the reason codes; everything else is a chart.
Costing an hour of downtime
Lost throughput is the visible part of the cost, and it's usually not the whole cost.
Take a line that runs 3,000 units an hour when it's up, selling at $40 a unit with $28 of variable cost — a contribution margin of $12 a unit.
- Lost throughput: 3,000 units × $12 margin = $36,000
- Expedite freight to still hit a customer ship date after the delay: $1,200
- Scrap on restart — the first few minutes after a jam clear typically produce reject material while the process re-settles: 150 units × $28 variable cost = $4,200
- Idle labor paid regardless of output, four operators at a fully loaded $35/hour for the hour: $140
Total cost of that one hour: $41,540 — 15% more than the lost-throughput figure most dashboards stop at. The expedite, scrap and labor lines don't show up unless someone goes looking for them, which is the same reason a reason code has to exist before any of this is fixable: you can't cost what you haven't categorized.
What tracking actually gets you
Hopak Machinery cut estimated downtime 80% after moving to a smart track-based packaging machine that surfaces stoppage data in real time. Turatti's food-washing line saw a similarly steep decrease, up to 80%, once machine downtime became visible and attributable rather than absorbed into a shift's general "slow day." On the predictive-maintenance side, Fiberon avoided 178 hours of downtime and saved $274,000 over eight months by catching failures before they became stops at all, a 2.5x return on the monitoring investment.
Be honest about the sample behind figures like these. Neither of our benchmark pages publishes a downtime median, on purpose. Process optimization holds four parseable downtime figures in the entire corpus (12%, 20%, 27%, 80%) and predictive maintenance holds five — of which three are the identical round number 80%. That last detail is the important one: it tells you more about how vendors round a number in a case study than about what a plant achieved.
The direction is consistent — every deployment that reports a figure cut stoppage time by double digits — but there is no honest average to quote, and the 80% you see above is the top of a very short list, not a typical result. Treat any vendor quoting a tidy industry-wide downtime percentage as quoting a marketing number until they tell you the sample size.
Where to go next
- Downtime is the largest lever on the Availability factor of OEE — see OEE calculation for how a downtime minute turns into a score, and what is OEE for how to read the result.
- Downtime eats into the production time a line needs to hit its pace — see takt time for what happens when a bottleneck can't keep up with demand.
- The process optimization software hub is where downtime tracking and reason-code platforms are evidenced against real deployments; the use-case hub and benchmarks hold the full dataset behind the figures above.
- Darigold avoided 105 hours of downtime and nearly $1M in cost over six months with the same machine-health approach as Fiberon, on 34 monitored assets.
AI Roadmap — Eight workflows in your plant where AI would pay back this quarter. We know which. 12 pages, in 72 hours.