What Is Uptime (And Why 99.9% Isn’t Always 99.9%)

Every hosting provider promises 99.9% uptime. They all use the same number. They all mean different things. And none of them tell you the whole story.

Understanding uptime is one of the most useful skills for anyone running a website. It tells you what your host actually delivers, what your users actually experience, and what to look for in a new host.

This is the plain-English explanation, with the math you need to evaluate providers honestly.

What “Uptime” Means

Uptime is the percentage of time a service is available and responding correctly. 100% means never down. 99.9% means down for some small amount of time per month or year.

The basic formula:

Uptime % = (Total time - Downtime) / Total time × 100

So 99.9% uptime over a month (30 days = 43,200 minutes) means:

Allowed downtime = 43,200 × (1 - 0.999) = 43.2 minutes

That is roughly 43 minutes per month. Sounds small. Adds up to 8.7 hours per year.

The math is straightforward. The promises get complicated.

Why All Providers Claim 99.9%

The number 99.9% is the marketing sweet spot for shared hosting. It sounds impressive (“only 43 minutes a month!”) without promising enterprise-grade reliability.

In practice:

• Cheap shared hosting: “99.9% uptime” with broad exclusions, vague measurement, and small service credits. Often actually delivers 99.5-99.9%.
• Better shared hosting: “99.9% uptime” with status page, specific measurement, and clearer remedies. Often actually delivers 99.9%.
• Premium hosting: “99.99% uptime” with financial backing, independent monitoring, and 100% service credits for major outages.
• Cloud platforms (AWS, GCP, Azure): Region-specific SLAs with detailed measurement and real service credits (10-100% of monthly fee).

The number alone tells you nothing. The SLA behind it tells you everything.

What Counts as Downtime

This is where the trickery happens. Different providers measure differently.

Strict definition: any moment a request to your site returns an error or times out. Includes DNS failures, network issues, application errors.

Loose definition: only time when the server itself is unreachable. Application bugs that return 500 errors do not count. Network issues between your host and your users do not count. DNS issues do not count.

Provider-favorable definition: downtime only counts when the provider’s own monitoring shows an issue. If their monitoring missed it, it did not happen.

The honest definition: time during which a real user trying to access your site cannot do so. This is what matters. This is rarely what is measured.

When comparing providers, ask specifically: “What counts as downtime in your SLA?” The answer tells you how serious they are.

How Uptime Is Measured

Three common methods, all with different biases.

1. Provider’s own monitoring. The provider watches their own service. They see it as up. Convenient for them. Useless to you unless you trust them completely.

2. Internal synthetic checks. Scheduled requests from inside the provider’s network. Catches server downtime but misses user-perceived issues like routing problems.

3. External third-party monitoring. Services like UptimeRobot, Pingdom, Better Uptime, Datadog Synthetics. Real users in real locations check your site. Most accurate.

Most SLAs use method 1 or 2. Your experience should be measured by method 3. The gap between them is where marketing becomes fiction.

If your provider only counts uptime by their own monitoring, you are getting a worse deal than the SLA claims. Set up your own external monitoring. UptimeRobot has a free tier.

Translating Percentages to Real Time

The percentages sound similar. The reality is dramatically different.

99% uptime:

• Per day: 14.4 minutes
• Per week: 1.7 hours
• Per month: 7.2 hours
• Per year: 87.6 hours (3.6 days)

99.9% uptime:

• Per day: 1.4 minutes
• Per week: 10.1 minutes
• Per month: 43.2 minutes
• Per year: 8.7 hours

99.95% uptime:

• Per day: 43 seconds
• Per week: 5 minutes
• Per month: 21.6 minutes
• Per year: 4.4 hours

99.99% uptime (“four nines”):

• Per day: 8.6 seconds
• Per week: 1.2 minutes
• Per month: 4.3 minutes
• Per year: 52.6 minutes

99.999% uptime (“five nines”):

• Per day: 0.86 seconds
• Per week: 6 seconds
• Per month: 26 seconds
• Per year: 5.3 minutes

The difference between 99% and 99.99% is roughly 80x less downtime. The price difference is usually 5-10x. For mission-critical workloads, the math justifies the cost. For everything else, it does not.

Why 99.9% Is Often Not Really 99.9%

Three reasons the marketing number and the reality diverge.

1. Measurement gaming. If downtime only counts when the provider’s monitoring says so, and their monitoring has blind spots, real downtime goes uncounted.

2. Exclusion abuse. “Scheduled maintenance, DDoS attacks, third-party outages, and force majeure” can cover almost anything. A 4-hour outage from a DDoS attack does not count. A 6-hour maintenance window does not count. None of it shows up in the SLA calculation.

3. Monthly averaging. Some providers calculate uptime per month, others annually. A bad week spread across an annual calculation hides behind a good year. Monthly numbers are more honest.

The actual delivered uptime for many “99.9% SLA” providers is somewhere between 99.5% and 99.9%. Still good. Just not what the marketing suggests.

How to Verify Real Uptime

Do not trust the SLA. Verify.

Step 1: Set up external monitoring. UptimeRobot (free), Better Uptime, Pingdom, or Datadog Synthetics. Monitor from at least 3 different locations.

Step 2: Configure alerts. Email, SMS, Slack. Get notified the moment your site is unreachable. Do not wait for users to tell you.

Step 3: Track results over time. Most monitoring services give you a dashboard showing uptime percentage. Compare to what your provider claims.

Step 4: Check the provider’s status page history. Most providers publish incident history. Look at the last 12 months. If they have had multiple 4-hour incidents, the “99.9% uptime” is fiction.

Step 5: Read third-party reports. Sites like CloudHarmony, Upptime, and individual blog posts aggregate uptime data across providers. Useful reality check.

The best test: ask your provider for their last 12 months of actual uptime data with incident timestamps. Most cannot or will not provide it. That tells you something.

What Different Uptime Means for Your Site

Match the SLA to your workload.

99% is fine for:

• Personal blogs
• Hobby projects
• Internal tools no one depends on
• Documentation sites with redundant mirrors

99.9% is fine for:

• Small business sites
• Marketing pages
• Low-traffic applications
• Most WordPress sites

99.95% is fine for:

• E-commerce sites
• SaaS applications
• Business-critical internal tools
• Anything with paying customers

99.99% is needed for:

• High-traffic e-commerce
• Financial services
• Healthcare applications
• Mission-critical APIs

Most workloads are in the 99.9-99.95% range. Anything more is overkill for non-critical workloads. Anything less is not enough for production business sites.

Pick the SLA that matches your tolerance for downtime. Do not overpay for 99.99% if your site can tolerate 43 minutes per month.

What You Can Actually Do About Downtime

Even with the best host, downtime happens. You can mitigate it.

Multi-region deployment. Run in two regions. If one goes down, fail over. More expensive, but real high availability.

CDN with always-on. Cloudflare, Fastly, and similar CDNs cache your content at the edge. Even if your origin is down, cached content keeps serving.

Database replication. Read replicas in multiple regions. Write goes to primary, reads from anywhere. Survives single-region outages.

Health checks and auto-failover. Tools like AWS Route 53 health checks, Cloudflare Load Balancer, or similar can detect failures and route traffic away.

External monitoring. Catch downtime the moment it starts. Alert users if it lasts more than a few minutes.

Status page. Communicate proactively. Users forgive outages when you tell them in advance.

For most sites, a CDN with always-on caching is the cheapest meaningful improvement to perceived uptime. Your origin can be down for 30 minutes while users still get cached content.

What to Look For in a Real SLA

A few signals that an uptime promise is serious.

Specific measurement methodology. “We measure uptime with synthetic checks from 5 external locations every 60 seconds.” Not “we measure uptime however we want.”

Service credits at meaningful levels. At least 25% credit for missing 99.9%. 100% credit for missing 99%. Anything less is not serious.

Independent uptime reporting. Some providers publish their own uptime history from third-party monitors. AWS, Google Cloud, Cloudflare all do this.

Short claim windows. Generous providers give 7-30 days to claim credits. Tight providers give 24 hours. Generous is better.

Public status page with multi-year history. Real uptime data going back years. If they cannot show this, they are hiding something.

Financial backing. For providers like AWS, Google, Microsoft, the SLA is backed by a billion-dollar company. The promise is real. For a $5/month host, the promise is mostly aspirational.

What You Can Do Today

You do not need to switch hosts or build a complex infrastructure. Start here.

Step 1: Set up external monitoring. UptimeRobot free tier. 5 minutes to set up. Alerts via email or SMS.

Step 2: Track your actual uptime. Let the monitoring run for a month. See what your real number is.

Step 3: Compare to your provider’s SLA. Are they hitting their own target? If not, you have evidence to demand credit or switch.

Step 4: Add a CDN. Cloudflare free tier. Caches your content. Even if your origin is down, cached pages still serve.

Step 5: Decide what uptime you actually need. Personal blog? 99% is fine. E-commerce? 99.95%. Match the SLA to your workload.

If you host with InterServer, their SLA terms and public status page let you check the last 12 months of uptime history. Cross-reference with your own monitoring for the ground truth. That combination tells you whether they are meeting their commitments in your region, for your workload.

Conclusion

Uptime numbers are marketing until you verify them. Every host claims 99.9%. Most deliver somewhere between 99.5% and 99.99%. The variation is in how they measure, what they exclude, and how they handle claims.

The right uptime target for your site depends on the workload. Personal blog? 99% is fine. E-commerce? 99.95%. Mission-critical SaaS? 99.99%. Match the SLA to your tolerance, not the marketing.

The cheapest meaningful improvement to perceived uptime is a CDN. The cheapest meaningful improvement to real uptime is external monitoring. The cheapest meaningful improvement to high availability is multi-region deployment. Pick the level your workload actually needs.

Do not pay for 99.99% if 99.9% is fine. Do not accept 99% if your business needs 99.95%. Match the SLA to the workload. Verify with your own monitoring. That is the entire job.