How reliable are Internet Speed Test and why do speed results differ

Website Speed Testing

How reliable are Internet Speed Test and why do speed results differ

This is written in response to all those calls agencies get telling them their clients sites are slow according to another websites point of view.

Scenario Clients will run a lighthouse test or GT Metrix test see an “E” and send an irate email to their agency. They will run the same test on a competitors site and then use their “D” rating to request ( usually demand ) that their site is brought up to this standard. “I am paying x and my site is slow according to this testing site”.

So what actually happens in a speed test.

The website is fetched by the speed testing application – while it records every action and downloaded element recording it saves timestamps along the way. A web page will have at minimum 200 data items and some times over 1000 items to be downloaded when loading. The tool then gets all this data and renders it into a report – usually color coded for ease of reading. The test is loaded into a queue and then fired off.

The result is an indication – a barometric reading as it were. Its telling you the site is leaning towards slow or fast. It is not definitive.

Why are the results not definitive – because every connection to a web server is dynamic depending on server load , line speed, and multiple other factors that fluctuate all the time.

The variables affecting the speed report that are not in your control and you have no view of how they are set :

  • The  testing tool method of measuring,
  • The algorithm they apply,
  • How they calculate the result,
  • Physical location of where the application is hosted,
  • Their line speed of their machines data connection, it will only be as fast as the slowest network joint.
  • The network connection of this machine to the internet – I.E the route out of the machine to your website.

Factors that make a speed test result misleading.

  • Testing from the wrong part of the world – customers in UK and testing a site from a US monitor. Pick a monitor closer to home
  • The same site tested with GTmetrix Pingdom Tools and Google Lighthouse will produce varying and different results.
  • Every Page load is a complete new load – no caching – not the same as the customer experience. Customers are 90% on cached elements after page 1 is loaded.
  • DOM is flagged as bad – breaking news – even googles search result page have a bad DOM
  • Signing up to a speed monitor site that pings your site every 10 minutes with a speed test. You are paying for traffic to the site that is non revenue earning and actually slowing the site for paying clients !
  • SEO tools like screaming frog, ahrefs etc etc, all when running cause a heavy load on a site –  run these after hours and certainly dont let them index during business hours
  • First Contentful paint ( FCP ) all speed testers measure this differently.
  • The speed of the speed testers internet connection
  • The ip address of the speed tester maybe flagged on a firewall as an attacker !

What the different parts of a speed test mean

TTFB –  Time to first byte. This is usually the time to navigate all the network to your webserver.

DOM – Document Object model . It acts as a set of instructions for web browsers on how to render a page IE what do we show when it reads <h1> and so on. BTW google’s search results page fails many DOM tests on some speed testers – who’s right ?

CACHING – Caching is where dynamic content is pre-generated and served as a static asset. It saves lots of time expensive database lookups.

MINIFYING – Removing the Line feeds and tabs from a script. Its like removing all the paragraphs in a book. Little or no effect on speed – and renders the code unreadable !

COMPRESSION – compression of images and data usually has to be decompressed on the clients machine –  this does not always deliver a faster site render to the client.

CDN – Content Delivery Network – this is where the assets of the site are stored in a network of web servers – the effect is that the content is delivered from the webserver nearest to your location.  Good for international sites.

Breaking news  –  Your customer is the one paying for the product on your site – not lighthouse or GT Metrix or any robot ever.

Focus on the customer experience.

Useless content that slows a site and adds nothing to converting a visitor into a customer.

  1. Lots of carousels of products on the front page – news – no one is scrolling on your front page – they search there.
  2. Auto playing videos – keep to a bare minimum
  3. Embedding google maps on every page – run a static image – when clicked goes to 1 page with the interactive map.
  4. Social Media Feeds –  – run a static image – when clicked goes to 1 page with the interactive feed.
  5. Lots of scrolling banners – news – when was the last time you waited 2 mins to see all the scrolling banners – no one does.
  6. Lots of tracking pixels from services you no longer use or their insights or services you can run locally. Hotjar, Trust Pilot etc etc.
  7. Live help running when no one is attending it.
  8. javascripts that load on every page but are only needed on 1 page – Culprits Hubspot and Salesforce forms implemented incorrectly because of bad implementation instructions by the vendors
  9. Massive images – bigger in native size and shrunk to size using HTML.
  10. Hotjar running and not being used – more remote calling scripts sending information out that does not need to run all the time.
  11. Websites writing logs for every visitor interaction – DB inserts cost time – is the data required – does anyone ever look at it. DB tables with 1 million rows = slowness.
  12. run your admin and data processing on another server – save the CPU –  better again just run at night.
  13. run your DB on another server – only do if the site is really busy – like event ticketing – it seldom makes a huge difference and if it does then there is a deeper issue on the server you are avoiding.
  14. use CDN if you have to have huge images
  15. store fonts locally if possible.
  16. get a decent spec dedicated server – its as cheap as €10 a day and makes the biggest difference.

All of the above can be implemented and will speed up the site.

Best test is to run this on your mobile phone across a 4g connection. This is the customer experience not some robots view of the world. If its fast this way and the speed testers say your site is a grade F, then look at the slowness flags and clear these and then test again on your phone to make sure its still fast. If it is then this is good.

Fast sites deliver increased revenues.  Green ticks on reports do not.

Beware of agencies using the following sales tactic. The first meeting in a joint jeering and slagging match of the old agency usually using speed test as a stick. Beware if you are doing this – its a common mistake. You get caught up in the hubris of a new supplier and are guaranteed to loose all the knowledge built up as the new supplier regards the old one as a “fool” not to be listened to. Lots of newbie agencies use as a sales tactic to woo you away from an experienced agency. Experienced agencies do not hard sell – their reputation sells.  They know what matters and what does not and more importantly know how to spend your money wisely. It’s your money and your lively hood.

Note: Speed testing sites mentioned all serve a purpose and they are not reporting incorrectly – but from different angles. They may point to deeper issues on the site. The recommendations they give have to be examined and actions prioritized.


By | 2023-06-01T23:19:44+00:00 June 1st, 2023|e commerce, ecommerce|