Once upon a time, eHost.com advertised the same thing every shared host advertises: rock-solid uptime, backed by a guarantee. Then the brand was wound down, its customers were shuffled off to sister companies, and every uptime promise eHost ever made evaporated along with the brand itself. Nobody is honoring an uptime guarantee from a host that no longer exists. We tell the full story in our EHost.com review, but the short version is the founding lesson of this article: an uptime promise is only as durable as the company making it, and even then it is only a promise.
The good news is that you do not have to take any host's word for it. Uptime is one of the few hosting qualities you can measure yourself, independently, for little or no money. The webmasters who came out of the eHost era unscathed were the ones who monitored their own sites, kept their own backups and treated vendor promises as marketing rather than data.
This guide covers what uptime percentages actually mean in hours and minutes, how to read the fine print of an uptime guarantee, how to set up independent monitoring in an afternoon, and the difference between a site that is technically up and one that is actually usable.
What uptime percentages really mean
Hosts quote uptime as a percentage because percentages sound reassuring. Ninety-nine percent sounds nearly perfect. It is not. One percent of a year is more than three and a half days of your site being dark. The gap between the marketing number and the lived experience becomes obvious the moment you convert percentages into clock time:
| Uptime | Downtime per month | Downtime per year |
|---|---|---|
| 99% | About 7 hours 18 minutes | About 3 days 15 hours |
| 99.5% | About 3 hours 39 minutes | About 1 day 20 hours |
| 99.9% | About 44 minutes | About 8 hours 46 minutes |
| 99.99% | About 4 minutes | About 53 minutes |
Two things jump out of that table. First, every added nine is roughly a tenfold improvement, which is why the jump from 99% to 99.9% matters enormously while sounding trivial. Second, on shared hosting, 99.9% is a reasonable expectation and anything promising materially more should be read skeptically — the infrastructure required for true four-nines reliability is not what powers bargain plans.
Also ask when the downtime happens. Forty-four minutes a month spread across 3 a.m. maintenance windows is invisible; the same forty-four minutes during your busiest sales hour is a small crisis. Averages hide this completely.
Uptime guarantees and SLA fine print
An uptime guarantee is a service level agreement, and an SLA is a contract — which means the interesting part is the fine print. Before you let a guarantee influence a buying decision, check how it handles four questions:
- What counts as downtime? Scheduled maintenance is almost always excluded, and hosts control how much maintenance they schedule. Some SLAs only count downtime the host's own monitoring confirms.
- What do you get? The remedy is nearly always account credit — a partial refund of your tiny monthly fee, not compensation for lost sales or damaged reputation. A credit worth a dollar or two does not motivate anyone.
- How do you claim it? Most SLAs require you to detect the outage, document it and file a claim within a deadline. The burden of proof sits with you, which is itself an argument for independent monitoring.
- Who is standing behind it? This is the eHost clause. A guarantee from a disposable brand in a large portfolio can be discontinued along with the brand. Ownership and track record matter more than the number in the badge.
None of this makes SLAs worthless — a host willing to write 99.9% into a contract is at least staking something on it. But an SLA is an insurance policy with a low payout, not a reason to skip your own measurement.
How to monitor uptime independently
Here is the part that actually protects you. External uptime monitoring services ping your site from their own infrastructure at short intervals and alert you when it stops responding. Free tiers typically check every few minutes, which is plenty for a small site; paid tiers add faster checks, more locations and better reporting.
Setting it up properly
- Monitor a real page over HTTP, not just the server. A ping can succeed while the website is serving errors. Check the homepage or another meaningful URL and confirm a successful response.
- Add a keyword check if your tool supports it. Verifying that the response contains an expected phrase catches the nastiest failure mode: a page that loads but shows a database error or a blank template.
- Check from multiple regions. Outages are sometimes regional — a network issue between one part of the world and your data center. Multi-location checks separate "my host is down" from "one route is broken."
- Route alerts somewhere you will actually see them. Email plus a push notification or messaging app. An alert you read tomorrow is a report, not an alert.
- Let it run before you judge. A week of data means little. Two or three months of measurement gives you a real availability figure to compare against the marketing number.
Do this during your money-back window at a new host and you convert the refund period into a genuine evaluation. It is the single highest-value habit we recommend in our guide to choosing a good web hosting service.
Up is not the same as usable
Uptime monitors answer one question: did the server respond? They say nothing about how fast it responded, and a site that takes ten seconds to produce a page is functionally down for most visitors even while every uptime check passes.
The metric to watch here is time to first byte — how long the server takes to begin answering a request. On overloaded shared servers, TTFB is where the rot shows first: pages still load, checks still pass, but every response drags. Many monitoring services record response time alongside availability, and that chart is often more revealing than the uptime figure. A slow upward creep in response times over weeks is the classic signature of an oversold server, and it usually precedes the outages rather than following them. If that pattern sounds familiar, our article on whether to upgrade your plan or change hosts walks through the decision.
Status pages and transparency
A host's public status page tells you more about its culture than its uptime. The good ones post incidents promptly, describe the actual cause, and publish honest post-mortems after significant outages. The bad ones show a wall of green checkmarks that never change — including, memorably, during outages you are experiencing at that very moment.
Before buying, find the status page and scroll its history. A page with occasional, well-documented incidents is a better sign than a spotless one; every host has incidents, and the only variable is whether they admit it. No public status page at all is its own answer.
Red flags that predict reliability problems
After enough years on shared hosting, you learn that some warning signs are reliable predictors of downtime trouble ahead:
- Perfect claims. Promises of 100% uptime, or guarantees suspiciously stronger than what premium infrastructure providers offer, signal marketing-first thinking.
- Prices too low to sustain. Someone pays for the hardware and the staff. When customers barely do, corners get cut where you cannot see them — until you can. Free hosts are the extreme case, as we explain in how free web hosting works and why to avoid it.
- Support that goes quiet during incidents. If chat queues explode and tickets stall every time something breaks, the operation is understaffed for its customer base.
- A recent acquisition. Platform migrations follow ownership changes, and migrations are when reliability historically dips. Watch acquired brands for a year before trusting them with anything important.
- Vague infrastructure answers. A host that cannot say where its servers are or who runs its network is reselling someone else's capacity — another layer that can fail, and another company that can disappear.
The eHost lesson, generalized
Whatever eHost's real uptime numbers were in its lifetime, they are unknowable and irrelevant now — the monitoring graphs are gone, the SLA is void, and the brand's reputation was inherited by nobody. Every retrospective claim about how reliable it was is just memory and marketing residue. That is the general truth hiding in one defunct brand's story: historic uptime belongs to the company that measured it, and you only ever truly own the measurements you took yourself.
So flip the burden of proof. Let hosts advertise whatever they like, then verify with your own monitors during the refund window and act on what your data says. Reviews — including the ones in our hosting reviews hub — can shortlist candidates for you, but your monitor has no affiliate relationship with anyone.
FAQ
What is a good uptime percentage for shared hosting?
Treat 99.9% as the standard a competent shared host should achieve — that allows roughly three-quarters of an hour of downtime in a month. Anything persistently below 99.5% is a problem worth escalating or leaving over. Claims meaningfully above 99.9% on cheap shared plans deserve skepticism: the redundant infrastructure that four nines requires is generally not what powers a bargain plan, whatever the badge on the pricing page says.
Are free uptime monitors good enough?
For most small sites, yes. A free monitor checking a real page every few minutes from an external network catches every outage that matters and gives you response-time history as a bonus. Paid tiers earn their keep when you need one-minute checks, many monitored sites, multi-region confirmation or on-call style alerting. Start free, upgrade when your site's revenue justifies it — the important step is having any independent monitor at all.
Can I get compensation when my host misses its uptime guarantee?
Usually only in the form of account credit, and usually only if you file a claim with evidence inside the SLA's deadline. The credit is calculated against your hosting fee, not your losses, so it tends to be symbolic. The realistic value of an uptime SLA is as a signal of confidence and a lever in support conversations — not as insurance. Your actual insurance is monitoring, backups and the ability to migrate on short notice.
My site is up but slow — does that count as downtime?
Not to your host's SLA, but effectively yes to your visitors and to search engines. A page that responds in ten seconds fails the only test that matters: whether people wait for it. Track response times alongside availability, and treat sustained degradation as seriously as an outage. Persistent slowness on shared hosting usually means an oversold server, and the fix is a better plan or a better host rather than more patience.
The bottom line
eHost's uptime record died with eHost, which is exactly why it makes such a useful case study. Promises, guarantees and even genuine historic performance all evaporate when a brand does — the only uptime data you can rely on is the data you collect yourself, from outside, continuously.
The playbook is short: know what the percentages mean, read the SLA for what it excludes, run an independent monitor with real-page and response-time checks from day one, and watch for the red flags that precede reliability problems. Do that, and no host — living or defunct — gets to tell you how reliable it was. You will already know.