We always keep an eye on retail websites to see how they cope with key events in their calendar. One such event is Valentine’s Day. So we monitored around 30 pages to see how their web page performance was affected on the run up to the big day– a mixture of florists’ home pages and Valentine’s Day gift landing pages from a selection of major retailers.
Broadly consistent performance
Load times varied quite a bit between the sites, with average daily load times ranging from just over 3 seconds to 30 seconds. However, what was really striking was the consistent performance from most of the pages over the period. There were just a few exceptions – a couple of sites slowed down just before Valentine’s Day and a couple got faster.
Two big slowdowns – but what caused them?
The two pages that suffered a serious drop in performance are highlighted on the graph in Figure 1. Surprisingly, they were both among the fastest overall.
Figure 1 – Two slowdowns (circled in red)
On closer inspection, it turns out that neither of these slowdowns could be blamed on a sudden surge of forgetful partners, frantically getting their orders in at the last minute. Instead, by overlaying page size (shown in grey below) on load time (shown in green), it’s clear that bigger landing pages caused load times to increase.
Figure 2 – Increased page size caused slower load times
In the first case, page size went up mainly because a new banner image was added – the 515KB JPEG put quite a dent in load time. This might just have been understandable if it had been a stunning photograph of a popular bouquet. But it wasn’t. Instead, it was simply a few lines of text on a coloured background. As such, JPEG was a poor choice of format, and a PNG would have been a staggering 57 times smaller, at just 9KB. Better still, the banner could have been created using text styled with CSS, reducing both the overall file size and the number of round-trips.
The reason for the big jump in page size on the second site is even more perplexing.
Bizarrely, one hero image was swapped out for another virtually identical image, whose file size was more than five times larger. The first image weighed in at 110KB, while its replacement was 650KB.
The irony is that both sites had previously been performing very well. It was only through some tiny, but ultimately quite careless, changes that both ended up around 50% slower during one of their busiest trading periods.
Last year, we saw a couple of sites make what looked like some last-minute improvements. This year too, a couple of the slower sites managed to make some performance gains just ahead of the big day.
Figure 3 – Two pages got faster just before Valentine’s Day
Did either site benefit from a concerted effort to improve load times, perhaps after a flood of customer complaints? It looks unlikely. One site didn’t change until 13 February, which looks a bit too last minute. And both sites got faster thanks to a reduction in the (vast) number of product images. This is just as likely to be down to stock running out as it is to performance optimisation.
The positive story for this group of retailers was that none of them suffered performance problems as a result of a surge in traffic (of course, it’s not quite so positive if none of them received a surge in traffic). However, this exercise also shows just how much damage can be done simply by uploading an unoptimised or unnecessary image.
Eight ways you can optimise your page performance
- Set a performance budget
- Keep mobile in mind
- Optimise images
- Compress text files
- Make the most of caching
- Load style sheets early
- Merge files where possible
- Prioritise important content