What to Do When Your Website Goes Down

First: is it down for everyone, or just you?

Before anything else, establish that the problem is real. A surprising share of the site is down turns out to be local and self-clearing: a cached DNS answer on your own computer, a browser extension, an office network, or your internet provider having a bad afternoon.

Three checks, in about ninety seconds:

1. Load the site on your phone with wifi turned off. That puts you on a different network with different DNS resolvers. If it loads there, the site is up and the problem is on your end.
2. Try a third-party checker that loads the site from somewhere else. If it reports the site as up, believe it.
3. Ask one other person, ideally someone not in your building and not on your network.

If it is only you, the cures are flushing your computer's DNS cache, restarting the router, or waiting. Do not call anyone yet, and do not change anything.

If it is down for everyone, keep going. The next step takes ten seconds and is worth more than any other on this page.

Write down the exact error, word for word

The precise wording tells you which of four separate companies owns the problem, and calling the wrong one costs a day. Copy the message exactly, including any code in capital letters, before you touch anything.

• This site cannot be reached, or the address not being found. The domain is not resolving - a domain or DNS problem, belonging to your registrar. The web server is probably fine.
• Your connection is not private, or a certificate date warning. The certificate expired or does not match. Your host. The site and its files are untouched.
• A 500 error, or a critical error on this website. The application ran and failed. Your host or whoever maintains the site, and the first question is what changed recently.
• A 503 error, or a resource limit message. You hit the ceiling on your hosting plan. Your host.
• Account Suspended. Your host, and the first question is whether it is billing or abuse. Very different conversations.
• A red full-screen warning that the site is deceptive. A browser safety warning, meaning the site has been compromised. Security problem, not a hosting problem.
• Somebody else's website, or advertising links. The domain expired or moved. Go straight to checking the domain.

Take a screenshot too. Support staff ask, and what an owner remembers is often not what the screen said.

Is it the site, the DNS, the domain, or the certificate?

Four independent layers have to work in sequence for a visitor to see your homepage. Knowing which one failed is most of the diagnosis.

The domain registration is the lease on the name. It renews annually, and if it lapses everything above it stops. Its status is machine-readable: since ICANN retired the old public lookup system on 28 January 2025, the definitive source is RDAP, the Registration Data Access Protocol, and every registrar offers a lookup. Check the expiry date and the status codes. A status of clientHold, serverHold, redemptionPeriod, pendingDelete, or no nameservers listed at all, ends the investigation immediately - no amount of talking to your host will fix it. ICANN publishes what each status code means.

The DNS is the phone book turning your name into a server address. It can be run by the registrar, the host or a third party, and it is separate from the registration. If the domain is current but the site still does not resolve, the records are the suspect - which usually means somebody changed one.

The hosting is the server holding the files. Errors in the 500 and 503 range come from here, and so does a suspension. The certificate makes the padlock appear, expires far faster than it used to, and is renewed separately from everything else.

A small business can easily have those four at four different companies without knowing it. That is normal, and it is why the first thing to write down is which company holds which.

The three most common causes at this size

1. The domain expired, usually because the card on file lapsed. The most common cause and the most avoidable. Auto-renew is on, the owner is certain the domain cannot have expired, and the stored card expired eight months ago - or the renewal notices go to an address nobody reads, very often a former designer's inbox. After expiry the domain moves through a grace period and then a redemption period, typically thirty days, during which only the original registrant can restore it and the fee is far above a renewal. After that the name is released. The worst part is that the email dies at the same moment, so warnings about the problem arrive at a mailbox that no longer exists.

2. The certificate expired. Visitors get a full-page browser warning most will not click past, while the site itself is fine underneath. This has become sharply more common for a documented reason: the maximum lifetime of a public certificate dropped to 200 days on 15 March 2026 and falls to 100 days on 15 March 2027, while Let's Encrypt stopped sending expiry reminders in 2025. Any certificate not renewed automatically will take the site down without notice.

3. The host had an outage. Real, not your fault, and further down the list than most people assume. The tell is that other sites on the same host are down too and the status page says so. Confirm it, get a ticket number, ask for an estimate, and tell your customers.

Everything else - a broken update, a forced version change, a resource limit, a careless DNS edit, a hack that led to a suspension - is real but less frequent.

Uptime percentages, converted into minutes

While you are unhappy with your host, it is worth knowing what its availability promise allows. A thirty-day month contains 43,200 minutes, so:

• 99 percent - 7 hours 12 minutes a month, about three and a half days a year
• 99.5 percent - 3 hours 36 minutes a month
• 99.9 percent - about 43 minutes a month, roughly 8 hours 45 minutes a year
• 99.95 percent - about 22 minutes a month
• 99.99 percent - about 4 minutes 19 seconds a month, under an hour a year

Two footnotes. Hosts compute against the actual month, so a 31-day month at 99.9 percent allows nearer 45 minutes than 43. And the percentage conceals what you care about: one four-hour outage on a Tuesday afternoon and forty-three one-minute blips at three in the morning both score 99.9 percent. The first cost you customers; the second cost you nothing.

Understand also what a service level agreement pays. Almost universally it pays in service credits rather than money, capped at a fraction of your monthly fee, only if you notice and file inside a short window, with scheduled maintenance, network attacks and your own plugins excluded. Nobody has ever been made whole this way, so the promise is not worth a premium.

The questions worth asking are different: what was your measured uptime last year, how many incidents ran longer than an hour, and can I speak to a person at eight on a Saturday morning.

Who to call, and what to tell them

You know which layer failed, so you know who owns it. Now make the call efficient.

Have ready: the domain name, the exact error text and your screenshot, when you first noticed, whether anything changed in the last few days - an update, a plugin, a DNS edit, a payment - and the account email on file.

Open a ticket and phone. The ticket creates a timestamped record; the call gets attention. Write down the ticket number.

Ask two questions and write the answers verbatim: what happened, and when will it be back. A host that will not answer either has told you something worth remembering later.

If it is the registrar, ask whether the domain can still be renewed normally, whether it has entered redemption, and what the restore process is. Three different situations, three different costs and deadlines.

Then tell your customers. Almost nobody does this, and it recovers most of the lost business. Post on your Google Business Profile and your Facebook page: our website is down today, please call us. In a market this size most of your traffic is people who already know you, and they will simply call.

What not to do while it is down

Outages get made worse by panic, and the specific ways are predictable.

• Do not let anybody re-point the domain somewhere new until you know what happened. Resolvers cache the old answer for the length of its time-to-live, so a hasty change turns a one-hour problem into a three-day one while you wait for caches to expire on top of the original fault.
• Do not change nameservers as a diagnostic step. If the previous provider was also serving your DNS, the zone can disappear when you move, taking the email with the site.
• Do not pay an unexpected invoice, an unlock fee or a ransom before establishing who holds what. Unsolicited renewal notices for domains, tech support and search services are a documented scam pattern, and the Federal Trade Commission has warned small businesses about invoices made to look overdue to create urgency.
• Do not start deleting things. If the site was compromised, evidence of how they got in is the only thing that stops it happening again in three weeks.
• Do not restore a backup before you know the cause. Restoring over a compromise reinstates the same entry point, and restoring over a certificate problem does nothing.

Once it is back

Spend fifteen minutes closing the loop, because the same outage twice is a pattern and a pattern is a decision.

Write down what actually caused it, not what you assumed. If the cause was a card, a renewal date or an inbox nobody reads, fix that rather than the symptom: update the card at both the registrar and the host, put the renewal dates in a calendar you look at, and change the account email to an address the business controls rather than a person.

Then write the index card. One page kept with the insurance policy: which company holds the domain and the email on that account, which company hosts the site, whose nameservers answer, where the backups live, and who has admin access. An hour of a four-hour outage goes to finding out who to call, and this is the only thing that prevents it.

Set up free monitoring if you have none, alerting to a phone number and an off-domain email address. Five-minute checks are adequate; paying for one-minute resolution buys four minutes of warning on a site nobody was visiting at three in the morning.

Finally, take stock honestly. If this was a domain or certificate problem, the host did nothing wrong and moving fixes nothing. If it was the third real host outage this year, or nobody answered the phone, that is a legitimate reason to move - planned, with the DNS time-to-live lowered in advance and email handled separately, not carried out in a panic on the day.