Tracking Website Uptime

As website designers and managers we want to provide fast and reliable online experiences for our customers. But where do we start on what seems like such a complex and highly technical task?

The first thing you need is a means of tracking your website’s uptime and performance so lets start by looking at monitoring tools. Pingdom is our particular favourite, offering some useful control over when and how you get alerted about any downtime on. So if you want to receive an SMS as soon as a site goes down, or just an email after an hour’s sustained outage, you can set that up. A number of hosting companies offer something similar, Rackspace for example has its Rackwatch service which not only alerts you but also someone from their hosting support team, who will leap into action to diagnose the problem. Google’s Webmaster Tools can even help here, sending you an email alert if Google experiences any connection issues when trying to crawl your site.

With your monitoring in place the next question which usually arises next is what is acceptable when it comes to website availability? The short (but not particularly helpful) answer is ‘It depends’, because what is acceptable when it comes to downtime will vary based on the nature of your business and how mission-critical your website is to it. You would expect Amazon for example to be a bit more downtime-conscious then say, your local pub. People generally see any momentary downtime on the BBC homepage as heralding the downfall of civilisation as we know it, but in its early years people came to accept that Twitter would be down at some point pretty much daily. When it comes to making the judgement for your own website think about what that downtime might cost you, either in direct sales or reputation.

In 2013 website monitoring service Pingdom registered an average uptime of 97.4% across the 500,000+ sites they tracked over the year – which equates to about 19hrs downtime a month per site. At The Pixel Parlour we’ve established 99% as the minimum acceptable uptime for a website over a month. In any given month the average uptime across the client sites we monitor is 99.8%, so near perfect website availability is achievable without breaking the bank.

Some hosting and online service providers will offer clients a Service Level Agreement (SLA) which usually covers aspects like service availability and support response times. At The Pixel Parlour our hosting SLA guarantees 99% or better uptime on our managed solutions and details what office hours and out-of-hours support a client can expect from us. In practice the most important thing for a website owner to know is who to contact if your website goes offline, how quickly they can help and what it might cost you.

But uptime alone isn’t the full picture. If your website is online but painfully slow for 5 hours before actually going down for just 5 minutes, that downtime measure alone doesn’t provide the full picture. All sorts of things can cause your website to slow without actually going off-line and for real users that experience can be just as bad. To help achieve this fuller picture services like Pingdom also offer something called Real User Monitoring. With this in place (which just involves adding a snippet of code to a site) in addition to alerting you if your site goes offline the service also monitors how fast pages are loading for real visitors.

This provides a whole host of useful information from how your site is performing across different locations, browsers, device types. It even lets you look at how individual pages within your site compare on speed. Google Analytics provided something similar (albeit in a more limited form) in the Site Speed section of the Behaviour reports. From reporting like this we might identify a key page that is loading slower than others or perhaps performance issues around a particular browser. All of which are useful, actionable nuggets of info about your site’s performance.

As well as individual data points we also want to be on the lookout for patterns. If a site is consistently slow (or even down) at particular times then this is worth investigating – perhaps there is something happening on the server at that time that could be slowing things down (like backups or system updates) or perhaps it coincides with peaks in visitor traffic, suggesting that a beefier hosting solution is needed. Again, the key is to identify where there is room of improvement, act and then measure the result.

One other area that arguably falls into the reliability camp, but is much harder to monitor in an automated way, is broken content. Missing images, garbled text, video or third-party content not loading all have a detrimental affect on our website’s user experience but won’t be highlighted by traditional uptime monitoring. We’ve written before about making proper content checking part of a monthly website maintenance plan and using tools like SEOPowerSuite’s Website Auditor, Integrity or Xenu Link Sleuth to crawl your site looking for broken links or missing images. In addition to running on demand some of these applications can be set on a schedule and will email you the results, which is a great way to take some of the human element out of remembering to check.

But human input can play an important role too. Think of your website visitors as like a army of content checkers, crawling their way through the depth and breadth of your site every month. If you provide them with an obvious and easy means of feeding back on any issues they spot you’ll be surprised how much useful data you’ll get back to help improve your site.

In summary, website reliability can seem a pretty dry topic compared to design or content considerations, but as website owners or managers we should be looking at it as key competitive advantage. By arming yourself with the tools to start measuring your website’s speed and reliability you’ll be making an important first step in improving its performance. The next step is to engage the necessary technical expertise to realise the kind of change you would like to see.