Service level agreement

What we commit to #

These are the numbers we are prepared to be held to, not the numbers that look best in an advertisement. Nobody here is going to tell you we do five nines on shared hosting at $8.99 a month. What we will tell you is what happens when the target is missed, in writing, with the arithmetic shown.

Everything on this page starts the day your website goes live. Getting it built and live in the first place is a different commitment with a different remedy, and it is set out in the delivery and fulfillment policy: the seven day build is a target rather than a guarantee, and no credit on this page attaches to it.

Who this applies to #

Every customer on a paid hosting plan, Basic or Complete, whose account is up to date. It forms part of the terms of service.

It does not apply during a period when your service is suspended for non-payment or under the acceptable use policy, and it does not apply to a site that is not live yet.

Some of the response times differ by plan, and those rows say so. The uptime target is the same for everybody, because everybody is on the same machines.

The web hosting uptime guarantee #

We guarantee 99.9% availability in each calendar month, for your website and for your mailbox accepting mail. What turns that from a number into a guarantee is the service credit in section 11, which we pay when we miss it. It is measured per month rather than per year, which matters: a yearly figure lets a supplier hide a whole bad day inside a good twelve months.

Percentages are hard to feel, so here is what each one is in minutes:

A 31-day month allows a little more and February a little less. The calculation always uses the actual number of minutes in the month.

How we measure it #

Monitoring runs from outside our own network, because a supplier grading its own homework from inside its own data center will always give itself an A.

• Your home page is requested every 60 seconds from several locations.
• Downtime starts at the first failed check that a second location confirms. One location failing on its own is usually the checker, not your site.
• Downtime ends at the first successful check.
• It is measured to the minute and added up across the whole month.
• Mail is checked the same way, by attempting a delivery to a test address on the system.

Our figures are what we use for a credit. If your own monitoring shows something different, send us what it recorded and we will look at both. Where they disagree and we cannot tell which is right, we take yours.

What counts as downtime #

Your site being unreachable, or answering with a server error, to an ordinary request from the public internet. Also:

• the site loading so slowly that it times out, which for this purpose is the same as being down;
• an expired or broken certificate stopping browsers reaching you;
• your mailbox refusing to accept mail for more than 15 consecutive minutes;
• a failure in our DNS, our network or our content delivery network that stops visitors arriving;
• a maintenance window we ran without giving you the notice section 7 requires.

What does not count #

• Planned maintenance announced in advance under section 7.
• Your domain expiring, or being suspended by the registry, where the name is not with us or the renewal was yours to make.
• DNS you changed yourself, or a domain you pointed somewhere else.
• A third party you added: a booking widget, a chat, a map, a review feed, a payment provider. When one of those breaks, your pages usually still load, and we will help you take it off.
• Something published from your own login that broke the site. We will put it back from the pre-change backup, at no charge, but the minutes do not count.
• Your own connection, your device, your office network or your internet provider. If the monitoring can reach the site and you cannot, that is what has happened.
• Suspension for non-payment or under the acceptable use policy.
• Things outside anyone’s control: a failure of a major part of the public internet, an act of government, a natural disaster.

Emergency security patching is the one gray area, and this is how we handle it. If a serious vulnerability has to be closed at once, we do it and tell you as soon as we can. The first 30 minutes of that work does not count as downtime. Anything beyond 30 minutes does count, exactly like any other outage, because at that point the length of it is on us.

Planned maintenance #

Servers need patching, and the right course is to do it on a schedule at a quiet hour rather than pretend it never happens.

• The window is 1 a.m. to 5 a.m. Eastern time, on a Tuesday or a Wednesday.
• We give you at least five days’ notice by email, and it goes on the status page.
• We aim for no interruption at all, and where there has to be one we aim to keep it under 30 minutes.
• Most months there is nothing to announce, because most patching happens with the site still up.

If we run maintenance without giving you that notice, and your site is down for it, those minutes count as downtime. That rule exists to make sure we give the notice.

How fast we answer #

A total outage is watched by monitoring around the clock, so in most cases we have started before anyone writes to us. The row above is the outside limit on hearing from us, not on us starting work.

These are first response times. How long a fix takes depends on what broke, and we will tell you what we know and keep telling you until it is done.

Everything is dealt with on live chat and by email, and that is deliberate. Written support gives you a record of what was said and a link you can go back to, and it means an answer can arrive while you are on a job rather than only when you are free to sit and wait for one. Answering people quickly is worth more than most owners think: answering inquiries in ten minutes instead of two days is the same argument from your customers’ side of it.

Backups #

A full backup of your site, your database and your files runs every night, and another is taken immediately before we publish any change. Backups are held separately from the live system, so a failure of one does not take the other with it.

The most work you can ever lose to a server failure is one day, and in practice a good deal less, because of the backup taken before every change. We test restores on a sample of accounts every month, since a backup nobody has ever restored is a rumor rather than a backup.

Please still keep your own copy of anything irreplaceable, particularly original photographs and your customer list. Both export from the dashboard in a couple of clicks. This is not us hedging: it is what anyone sensible does with anything that only exists in one place.

Restores #

• Say which day you want back, or just describe what went wrong and we will work out which backup you need.
• We start within the time in the table in section 8, and a normal small business site is usually back inside 30 minutes once we begin.
• We take a backup of the current state first, so the restore itself can be undone if it turns out you wanted a different day.
• You can restore the whole site, or ask for a single page or file back from a particular night.
• You can also ask for a copy of a backup to keep, rather than a restore.

There is no incident charge, no hourly rate and no fee for a restore, ever. That is the whole point of paying for hosting.

The service credit #

The credit is a percentage of the hosting fee for the month in which the target was missed, and it comes off your next invoice. It is not paid out in cash, and the most it can ever be in one month is the whole of that month’s hosting fee.

As section 18 of the terms says, the credit is your remedy for us missing the uptime target. That is what a service credit is, everywhere. It does not affect any other right you have, it is not counted against the liability cap in that section, and it does not affect your ability to cancel, which you can do in any month for nothing. Nothing on this page limits a right your state’s law gives you, and section 23 of the terms leaves your own state’s courts open to you with no arbitration clause in the way.

A worked example #

A customer on Basic hosting at $8.99 a month. In a 30-day month, a fault on the 14th takes the site down from 9:20 a.m. until 2:40 p.m. Nothing else goes wrong that month.

$2.25 is not a lot of money, and we are not going to pretend otherwise. That is the arithmetic of a percentage credit on an $8.99 bill, and any supplier whose worked example stops before the final line is hoping you will not do the sum yourself.

The credit is a penalty on us, not compensation for you, and it is deliberately small because the real remedy for a bad month is in section 14: we fix it for nothing, we explain in writing what broke, and you can leave at the end of any month with no fee and no argument. A hosting company that cannot be walked away from is the only one that needs a big penalty clause.

Claiming a credit #

Usually you do not have to. Our monitoring produces the monthly figure for every site, and where an account has fallen below the target we apply the credit ourselves and tell you we have done it.

If we miss one, claim it. Email [email protected] within 30 days of the end of the month in question, with the dates and rough times you saw the site down. You do not need evidence: we will check it against our own records, and if we cannot separate the two accounts of it, we take yours.

We answer a claim within five business days, and the credit appears on the next invoice after that.

What we do besides the credit #

• We start without being asked. Monitoring wakes someone, at any hour. In most outages we are already working on it when the first customer writes in.
• We say so while it is happening. Within 30 minutes of confirming an incident, the status page says what is affected, and we email the customers it touches. It is updated at least hourly until it is over.
• We write it up. For any outage over an hour, you get a plain English note within five business days: what broke, why, how long it lasted, and what we have changed so that particular thing cannot happen again.
• There is no bill. No incident charge, no restore fee, no emergency rate.
• If it keeps happening, we move your site to different hardware rather than wait for you to lose patience.
• And you can leave. Any month, no notice, no fee, with your domain and a full export, on the timings in the refunds and cancellation policy. That is the strongest guarantee on this page, and it is the reason the rest of it has to be true.

Changes to this service level agreement #

We may improve these commitments at any time and will simply publish the better version. If we ever needed to reduce one, we would email you at least 30 days before it took effect, and you could cancel before that date without paying anything further.

The date at the top of this page is the date it last changed.

Common questions about the uptime guarantee #

The six things people ask while comparing one hosting company against another. The numbers in them are the numbers in the sections above, not rounder ones.