Every hosting plan advertises an uptime number, and nearly every web manager nods at it without asking the question the number hides. How long can your website be down before it starts to cost your organization money, or trust, or both? This post is built around that question, because the busiest weeks of the year are two weeks away, and the answer decides what preparation is worth doing now.

The purpose here is to help you put a real figure on downtime for your own organization, retail or nonprofit, and then to walk the short list that figure buys: a tested backup, a monitor that tells you before a customer does, a holding notice, a contact card, and the ten minutes of diagnosis that separate a short outage from a lost day. Most of it is free. All of it is easier in November than in the week it is for.

Two numbers that have names

Chapter 4 of the book frames hosting around a two-part question. If your website vanished this afternoon, how long could the organization afford to be down, and how much recent work could it afford to lose forever? The first number is the recovery time objective (RTO), the amount of time between failure and restoration your organization can tolerate. The second is the gap between your last good backup and the moment of failure. Both are business decisions before they are technical ones.

The marketing numbers deserve thirty seconds. A 99.9 percent uptime guarantee allows roughly eight and three-quarter hours of downtime a year. A 99 percent figure allows more than three and a half days. Hosts advertise the former, and you hold them to it with measurement you control.

The arithmetic, for a store and for a nonprofit

Take a small retailer whose online store does most of its December business. Divide the revenue you expect from the site in the busiest week by the hours in that week, and you have a rough hourly cost of being down. For many small stores that number is modest on a Tuesday in February and startling on Black Friday afternoon, when a single hour can carry a day's worth of ordinary sales. An outage that costs almost nothing in the off season can cost the whole promotion in season.

Now take a nonprofit. Year-end giving arrives on a calendar the organization does not control: the appeal letter lands, Giving Tuesday falls on December 1, and the final days of the year bring donors who give because the deadline is real. The donation page is the organization's register for those weeks. Run the same division with last December's gifts, and remember that a donor who meets an error page often does not try again the next day; the gift goes to the organization whose page loaded.

In both cases the honest answer to the cost question is seasonal. The same site can tolerate a four-hour outage in July and almost none on its peak days, which is why the preparation below is timed for now.

Which landlord gets the call

An outage has a clock on it, and the first ten minutes decide which of three companies you call; calling the wrong one adds a day. The registrar holds the name, the Domain Name System (DNS) provider publishes the directory, and the host serves the files. The symptoms tell you which has dropped it. If the site is down but email arrives normally, the host gets the call. If the browser says the domain cannot be found at all, check the domain's expiration first, then DNS. If the site is down and email is bouncing too, the domain or DNS is the problem, and nothing at the host will fix it. A server error on screen, a 500 or 502 or 503, points at the host or at something you changed.

Before any ticket, confirm the site is actually down for the world and not only for you: load it from a phone on cellular, and check a third-party tool such as downforeveryoneorjustme.com.

Two questions that solve most outages

Did the host break it? Check the host's status page before opening a ticket. A platform-wide incident means the fix is theirs, and your job becomes communication rather than repair; restoring backups or changing DNS during a host incident is repairing a building during an earthquake.

Did we break it? Note what changed in the last day: a plugin update, a new plugin, a DNS change, a certificate renewal, a developer deploy. A site that died minutes after an update did not die of coincidence.

Restore from backup when the site itself is broken: a bad update, a corrupted database, a failed migration. Do not restore when the server or network underneath is failing; the restore cannot fix that layer and may add new problems. When unsure, the host's error logs are the tiebreaker.

Being down without looking absent

The site being down is visible. The organization looking absent is optional. A short factual notice on the busiest social channel, down for maintenance and back shortly, with a phone number or email in the meantime, keeps visitors from concluding you have closed. Brief the front desk with the same one-liner so callers hear one story. Skip speculation and apology arithmetic; a calm two-line notice, updated when the facts change, is the whole job.

What the cost figure buys

Once you know what an hour costs on a peak day, the preparation list prices itself.

A backup you have tested. Host backups live inside the same account as the site and vanish with it, so the standard is three copies, on two kinds of storage, one of them somewhere the host does not control. An untested backup is a hope, not a plan; the quarterly restore test is what turns it into one.

A monitor you did not pay for. UptimeRobot's free tier checks the homepage every five minutes from the outside and alerts a role-based address when it stops responding. The web manager should learn about an outage from a notification, never from a customer; Saturday's quick win sets it up.

Staging, if the number justifies it. For a site with commerce, donations, or heavy customization, updates belong on a rehearsal copy first; many managed hosts include one in a click. For a brochure site, a fresh backup and a quiet hour do the same work.

A contact card and a holding notice, written now. The host's support line and account number, the helper who takes the call, the registrar's recovery page, and two sentences for the social notice, on one sheet where a shaken person will find it.

The log that ends the argument

Every outage ends with one line in the incident log: what failed, which layer, how long it was down, the cause, the fix. One bad afternoon is weather. A pattern of them, documented with dates, is a landlord problem, and Chapter 4's hosting scorecard turns that pattern into a decision rather than a mood. Reliability is a promise your organization makes without saying it out loud, to every customer and donor who arrives expecting the door to open. Knowing what an hour of downtime costs is how you decide, in advance and on purpose, how much that promise is worth keeping.

Read the chapter before the busiest week. Chapter 4 of Strategic Web Management: First Steps for the Accidental Web Manager covers hosting end to end: the 3-2-1 backup rule, the restore test, uptime monitoring, staging, and when a host has earned replacement. The chapter's Hosting Evaluation Scorecard and Backup Verification Log are free with a Web Managers account at https://webmanagers.com/register.

Categories

Governance  Management  Fundamentals  

Terms