Predicting next years website traffic – and Load testing realism

Date: 29th November 2013
Author: Deri Jones

If you want to ensure trouble-free growth in your web traffic next year: there’s a helpful article at  eConsultancy : How can you predict next year’s website traffic?  And I’ve pulled out some key tips, to help turn that into a plan for a realistic website load testing – helping to construct that plan is something every marketer should be involved in.

Planning next years business growth is a regular task for online retailers – but in many cases, the marketer does not take that analysis and extend it into the realm of Load Test planning:  This can leave a site vulnerable to performance problems later on due to inadequate testing.

With just a little extra effort, the marketer can reduce significantly the risk of traffic peaks (more visitors is a good thing!) from causing website slow downs in user experience and lost sales ( a bad thing).

It’s especially vital to recognise if maybe your organisation is one that could benefit from  a better understanding between business and technical teams – especially if in the past there has been a mis-match of language, when the whole issue of load testing was being discussed.

Size of the Store online

‘Size of the Store’ –  is the approach, of setting what traffic we expect to see, to get the business objectives for the site next year.

But whereas the normal business analysis will plan for a Size of Store online in terms of ‘£ sales per year’ orders per month’ etc:    for website load testing it’s all about the traffic peaks – the maximum few hours of the year, not the average daily or weekly etc numbers.

Tip 1 – use your web analytics and dial in time-lines of just a few hours, round your past  traffic peaks:   this is the raw traffic data, that in next years plan will need to be scaled up to meet goals.

Concurrent Users:  the worst metric for load test planning!

It’s a metric that tech teams often use, because it is so easy to pull from the web servers: available 24/7 from the site.

But whilst it has a value as a rough indicator of ‘how busy are we right now’ – it’s actually a misleading number.

Because 10,000 visitors who in the last 30 minutes hit your homepage and then left:  will count as equivalent to 10,000 visitors who hit 25 pages and placed an order!

Tip 2:  ignore any thought of  Concurrent user numbers:  and instead take the Web analytics data from above, for your peak traffic hours, and work out the break down % ratios between the core User Journeys.

You should end up with a table, something like this:

User Journey % of traffic
Find product and Add to Basket: by Search 15
Find product and Add to Basket: by Navigation 18
Checkout 7
Browse using Search 31
Browse using navigation 22
Leave a Review 3
Update an order / or etc etc

For each of the Journeys, you can work out how many pages long the Journey typically is:  and in addition set if you wish a % value for dropOffs – ie visitors who don’t make it to the end of the journey.

PC vs mobile vs tablet

Tip 3: If you don’t feel you have any data detailed enough for your next year’s plan – just provide a simple ratio %: of what you feel the ratio of traffic between those user types will be next year.  It’s enough to get your load test partners started: and if you come back later with a mobile breakdown % per Journey too all the better, it can then be dropped into the plan.

Taking the whole approach above  – will ensure the least risk of lost sales next year due to website capacity limitations. It will provide enough detail to allow your load test partner to spec up a meaningful, realistic load test.

 Also see

Keeping It Real: Avoiding Pitfalls When Load Testing Websites Using A CDN
Why Do Test Teams Want To Avoid Load Testing The Live Website?

Top