When a SCADA vendor advertises three, four, or five nines of availability, they are making a precise claim about how much downtime the system is allowed per year, and the difference between those tiers is dramatic once you convert the percentages into minutes. This is uptime in the reliability and service-level sense, which is a different measurement from the availability rate in an OEE calculation, where availability means the share of scheduled production time a machine actually ran. This guide translates the nines into real downtime figures, shows what architecture each tier demands, and gives you a table so you can judge competing uptime claims on the same footing.
Availability and the Nines in one line: Availability, in the uptime sense, is the percentage of time a system is operational and reachable, and the nines are shorthand for how many nines appear in that percentage. Three nines is 99.9 percent, four nines is 99.99 percent, and five nines is 99.999 percent, corresponding to roughly 8.8 hours, 53 minutes, and 5 minutes of downtime per year respectively. Each additional nine roughly cuts allowable downtime by a factor of ten and demands a step up in redundancy, so the nines are both a measure of reliability and a rough proxy for how much the system costs to build.
The nines feel abstract until you convert them into time, because a percentage close to one hundred hides how much slack it really allows. A year contains about 525,600 minutes, so an availability figure is just that total minus the permitted downtime. Ninety percent, a single nine, sounds respectable but permits roughly 36.5 days of downtime a year, which no control system could tolerate. Adding nines shrinks that budget fast: 99 percent still allows about 3.65 days, 99.9 percent allows about 8.8 hours, 99.99 percent allows about 53 minutes, and 99.999 percent allows only about 5 minutes across the entire year.
The table below lays out the common tiers so the jumps are visible. Each extra nine multiplies the reliability requirement by about ten and correspondingly divides the downtime budget by about ten, which is why the gap between four nines and five nines is not a minor upgrade but a different class of engineering. Availability tier: 99 percent (two nines) allows about 3.65 days of downtime per year. 99.9 percent (three nines) allows about 8.8 hours per year, or roughly 43 minutes per month. 99.99 percent (four nines) allows about 53 minutes per year, or roughly 4.4 minutes per month. 99.999 percent (five nines) allows about 5.3 minutes per year, or roughly 26 seconds per month. 99.9999 percent (six nines) allows about 32 seconds per year. Reading the table, a buyer can see that a five-nines claim leaves almost no room for even a single unplanned reboot, which is a strong hint about the architecture underneath it.
Downtime budgets translate directly into design because each tier of nines is only reachable with a matching level of redundancy and operational discipline. Three nines, at nearly nine hours a year, is broadly achievable with a single well-maintained server, good backups, and prompt manual recovery, since a couple of hours of outage for a reboot or a restore still fits inside the budget. It does not require automatic failover, which is why it is the tier a modest on-premise deployment tends to land at without special effort.
Four and five nines change the picture entirely because manual recovery no longer fits the budget. Four nines, at under an hour a year, effectively rules out any outage that a human has to notice and fix by hand, so it needs redundant components and automated failover that swaps to a standby in well under a minute. Five nines, at roughly five minutes a year, demands that recovery be near-instantaneous and that even planned maintenance happen without taking the service down, which means redundant everything, no single point of failure anywhere in the path, and often geographic separation so a whole site can be lost without breaching the budget. Beyond five nines the cost climbs steeply for gains most operations cannot justify, which is why five nines is the usual ceiling for serious commercial claims.
For a buyer comparing cloud SCADA platforms, the nines are useful only when you read the fine print around them, because the same headline figure can mean very different things. The first question is what the number covers: an availability claim for the central platform says nothing about the link between a field site and the cloud, which crosses cellular, satellite, or radio networks the vendor does not control. A meaningful comparison separates the availability of the hosted supervisory layer from the availability of the field connection, since a customer experiences an outage the same way regardless of which side failed.
The second question is how availability is measured and what the vendor owes if they miss it. A credible service-level agreement defines the measurement window, states what counts as downtime, excludes scheduled maintenance explicitly or folds it in, and specifies remedies. A platform such as Merobix that runs across multiple availability zones can pursue a high availability tier for the central service, while the buffering built into edge gateways protects data continuity across the field link even when the connection itself dips. When you evaluate claims, insist on the converted downtime figure, confirm what is and is not included, and treat any five-nines claim without a matching architecture description as marketing rather than engineering.
Five nines means 99.999 percent availability, which permits only about 5.3 minutes of downtime across an entire year, or roughly 26 seconds per month. That budget is so small that recovery must be effectively automatic and even planned maintenance cannot take the service offline. Reaching it requires fully redundant components with no single point of failure, often spread across separate locations.
No. Here availability means uptime, the percentage of time a system is operational and reachable, and is tied to reliability and service-level agreements. The availability rate in Overall Equipment Effectiveness is a production metric measuring the share of scheduled run time a machine was actually producing, after subtracting breakdowns and changeovers. They share a word but measure different things, one about system reliability and one about manufacturing utilisation.
Each nine roughly divides the allowable downtime by ten, so moving from three nines to four cuts the budget from about 8.8 hours a year to about 53 minutes, and four to five cuts it again to about 5 minutes. Below an hour a year, any outage a person has to notice and fix by hand blows the budget, forcing automated failover, redundant everything, and eventually geographic redundancy. The engineering and cost climb steeply because each tier removes another category of tolerable failure.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.