WatchA founder once proudly told a prospective customer his service had “99% uptime.” It sounded impressive, and the deal nearly closed on it — until the customer's engineer did the arithmetic out loud: 99% is over seven hours of downtime every single month, the better part of a working day, gone, every month of the year. Suddenly the impressive number looked like a liability. That's the trap with uptime percentages: they're designed to sound reassuring, and the only way to know what one really means is to translate it back into hours and minutes.
What the percentages mean in real downtime
The percentage itself is almost useless until you convert it, because all the meaning lives in the tiny gap between it and 100%. So here is that gap, translated into the only units that matter — actual time your service is down each month.
99% sounds great and is genuinely poor: about 7 hours and 18 minutes of downtime a month. 99.9% — “three nines” — allows about 43 minutes. 99.99% — “four nines” — allows about 4 minutes and 23 seconds. 99.999% — “five nines” — allows roughly 26 seconds a month. Notice the shape of it: each extra nine cuts the allowed downtime by a factor of ten, and is roughly ten times harder and more expensive to achieve. The distance from 99% to 99.9% is cheap; the distance from 99.99% to 99.999% can cost more than the rest combined.
What's a realistic target
With those numbers in hand, the right target becomes a business decision rather than a bragging contest. For most websites and SaaS products, 99.9% a month is the sweet spot: a solid, genuinely achievable target that gives you under an hour to absorb deploys, restarts and the occasional hiccup without breaking your promise.
Going further is a real commitment, not a setting you switch on. 99.99% means engineering for it — redundancy, automatic failover, no single points of failure — across your whole stack. And chasing five nines, with its 26-second monthly budget, rarely pays off unless every minute of downtime directly costs you large sums. For most teams, promising 99.9% and reliably hitting it beats promising 99.99% and missing it.
How to measure it honestly
Here's the catch that undoes a lot of uptime claims: a percentage is only as honest as the way you measure it. The same service can report 99.99% or 99.9% depending entirely on where you check from, how often, and what you decide counts as “down.” Before you quote a number, pin those down.
Measure from outside your own network, because uptime as seen from the same datacentre is uptime your customers never experience. Check at least every minute, so a short outage can't hide between probes. Confirm failures before counting them, so a single network blip doesn't unfairly dent the figure. And decide explicitly whether planned maintenance counts against the number — there's no single right answer, but you have to say which one you chose.
Turning a target into an SLA with WatchControl
Once you can measure real uptime honestly, you can do something powerful with it: promise it. An SLA puts a specific number in front of customers and holds you to it, and an uptime report is what proves you actually met it rather than asking them to take your word. The order matters, though — measure first, promise second.
The honest path is to start monitoring, watch your real number for a month or two, and only then commit to a target you've already shown you can hit. WatchControl is built for exactly this loop: it measures real uptime from outside your network, builds the report automatically, and powers both your public status page and your SLA from the same data — free to start. You end up promising a number you can stand behind, backed by evidence instead of optimism.