Your hosting provider promises 99.9% uptime. Sounds solid. Three nines. Enterprise-grade. You sign the contract and assume your site stays online.
Then you do the maths. 99.9% uptime allows 8 hours and 45 minutes of downtime per year. That's a full working day when your site doesn't load, your API returns errors, your customers can't check out. For most businesses, that's thousands of pounds in lost revenue. For some, it's tens of thousands.
The real problem isn't the SLA percentage - it's that most businesses have no idea what their actual uptime is. They trust their provider's dashboard, which measures the provider's infrastructure, not the customer's experience. Your host might be up while your database is down. Your CDN might be responding while your origin server is dead. The SLA says 99.9%. Your real uptime could be 98%.
The Maths Nobody Explains
SLA percentages compress annual downtime into a single number that hides the impact:
99.9% uptime = 8.75 hours of downtime per year = 43 minutes per month. A single bad deployment during business hours burns through your monthly budget. A database migration that takes longer than expected? You're over.
99.95% uptime = 4.38 hours per year = 22 minutes per month. Better, but still fragile. Two incidents and you've breached your SLA.
99.99% uptime = 52 minutes per year = 4.3 minutes per month. This is where it gets expensive. Achieving four nines requires redundancy at every layer, automatic failover, and constant monitoring. Most businesses don't need this. Some absolutely do.
The percentages make downtime sound acceptable because they're cumulative. 99.9% feels like 'almost always up'. But 8.75 hours is a working day. If your business runs online, that's a day with no revenue, angry customers, and support tickets flooding in. The SLA doesn't compensate you for lost sales. It refunds a fraction of your hosting bill.
What Your Provider Actually Measures
Hosting providers measure uptime at their infrastructure layer. Are the servers responding? Is the network reachable? Can they ping the machine? If yes, they're counting it as uptime. What they're not measuring:
Can customers actually load your site? Is the database accepting connections? Are third-party APIs responding? Is the CDN serving stale content? Did a deployment break checkout but leave the homepage intact?
This gap is where most businesses get burned. The provider's status page shows green. The SLA is technically met. But customers are seeing errors, transactions are failing, and nobody's measuring it. You find out when revenue drops or support calls spike. By then, you've lost the downtime window and the customers who gave up.
Why Most Businesses Don't Measure
Monitoring costs money and takes effort. Synthetic monitoring - automated tests that mimic user behaviour - requires setup and ongoing maintenance. Real user monitoring - tracking actual customer experiences - requires instrumentation in your codebase. Most small businesses assume their hosting provider is handling this. They're not.
The result? Nobody knows their real uptime. Not the percentage on the SLA, but the actual availability customers experience. A site can have 99.9% infrastructure uptime and 97% customer-facing uptime because of untracked issues in application code, slow database queries, or third-party service failures.
For e-commerce, this is brutal. A checkout flow that fails 2% of the time doesn't trigger alarms. It just quietly loses sales. You optimise conversion rates, improve product pages, run ads - all while bleeding revenue from an uptime problem you don't know exists.
The Real Cost of Downtime
Calculating the cost of downtime is uncomfortable because it makes the SLA percentages feel inadequate:
A business doing £500,000 annual revenue loses £57 per hour of downtime. At 99.9% uptime (8.75 hours down), that's £498 in lost revenue per year. Sounds manageable.
But revenue isn't distributed evenly. Most businesses have peak hours - weekday mornings, Black Friday, product launches. Downtime during peak hours costs far more than the average. An hour down during a product launch could mean losing £5,000, not £57. The SLA doesn't account for timing. Downtime at 3am costs you nothing. Downtime at noon costs you everything.
Customer trust is harder to quantify but often more expensive. One bad experience and they're gone. Some come back. Most don't. The lifetime value of a lost customer dwarfs the immediate revenue hit. The SLA refunds your hosting costs for that hour. It doesn't refund the customer who switched to a competitor and never returned.
What Good Monitoring Actually Looks Like
Real uptime monitoring tracks the customer experience, not just server availability. That means:
Synthetic checks from multiple locations testing critical user paths every minute. Can users log in? Can they add items to cart? Does checkout complete? If any of these fail, you know immediately - before customers start complaining.
Real user monitoring embedded in your application measuring actual performance. How fast do pages load for real customers? Where do users encounter errors? Which browsers or regions see degraded performance? This catches issues synthetic tests miss.
Dependency monitoring for every service you rely on. Your payment gateway, your email provider, your CDN - if it goes down, you need to know before customers notice. Most outages involve third-party services, not your own infrastructure.
The cost of proper monitoring is a rounding error compared to the cost of downtime you don't detect. A few hundred pounds per year buys you visibility into what customers actually experience. That's cheap insurance against the 8.75 hours your SLA allows.
The Question You Should Be Asking
Not 'what's my hosting provider's uptime?' but 'what's my actual uptime as experienced by customers?' The gap between those two numbers is where revenue disappears.
99.9% isn't a promise your site stays online. It's permission for your site to be down for 8.75 hours per year. If you're not measuring those hours - when they happen, what breaks, how many customers are affected - you're flying blind. The SLA protects your provider, not your business. Monitoring protects your business.
Start measuring. You'll be surprised how much downtime you didn't know you had.