You just checked the PageSpeed score on your ecommerce site, and it’s not great. Is this something to be concerned about? Probably not. Remember that a PageSpeed score is not actually speed. It’s just a useful bit of guidance to help find some areas for improvement. On its own, it means very little. Ultimately, your site needs to be competitive and provide a good user experience. As long as you are on par with your competition, there’s probably nothing to worry about. And for ecommerce sites, how much effort you should put in to outperforming your competition should probably be a financial decision, not a technical one.
Speed and Organic Rankings
Google has made it clear that speed is a ranking factor for organic search. But it’s not the only factor. There are hundreds of others as well. The difference in ranking between a slightly faster site and a slower one might be a few percentage points of organic traffic, everything else being equal. But it generally won’t be orders of magnitude in difference. That could be worth a lot of money. But if you’re weighing that against ecommerce site features that are common and expected by users, like having a JavaScript-heavy chat interface on the front end, a nice menu system, or an image-rich site search, the features are probably more important than the slight speed improvement you would get without them.
For example, you might have an amazing chat interface that clearly helps close more sales. For many sites with complex products and users with lots of questions, chat users convert at twice the rate of those who aren’t engaged by chat agents. I wouldn’t take down that amazing chat in order to get a minor bump in organic traffic. You’d need a massive improvement in organic results to offset the financial impact of taking a highly-converting chat system offline. In a case like this, if you’re concerned about the PageSpeed score impact of the chat system, reach out to the vendor and ask them what they can advise to help improve the speed of their system. Perhaps it can be implemented a bit differently, or the vendor is already planning on improvements to clean up their code more.
The bottom line is that the speed impacts on organic rankings have to be balanced against the benefits of those things which reduce your PageSpeed scores. And in the case of an ecommerce site, this should be a calculation based on the potential financial outcomes, not a technical decision based on trying to improve a PageSpeed score.
Do Conversion Rates Vary with Speed?
Google has published a lot of statements about how conversion rates for ecommerce are lower for slower sites. While this effect is obviously there when you compare a site with a 1-second page load to a site with an 8-second load time, they likely overstate the impact on sites that load in reasonable timeframes. The difference between a site with a 1-second page load and a 2-second page load is minuscule and may not even exist.
It seems like everyone quotes maxims such as “for every one second in page load time improvement, your conversion rate will go up by 15%”. Those statements are complete rubbish. Read the fine print. It will always say something like “source: internal study”. Nobody ever publishes their data or methodology for review. More likely, highly profitable sites have more money to spend on improving their performance. It may not be the improved speed that is driving additional sales, but quite the opposite effect. Sites with more sales invest more in speed. We have no evidence that any reputable study has been done on this that could actually tease out the direction of causality between speed and conversion rate.
We’ve even done some A/B tests in the past, purposefully slowing down ecommerce sites in order to see if that had an impact on conversion rates. After all, if we expect a slight improvement in speed to translate into improved conversion rates, shouldn’t we see a slight reduction in conversion rates if we slow a site down by a similar amount? These experiments showed that within an “acceptable range” of site speeds, there is no statistically significant change in conversion rates, and convinced us that tiny improvements in site speed aren’t worth the costs in many cases.
Try it yourself. Add a 1-second delay to your page loads with some JavaScript and let that run for a week on a fraction of your traffic. If you don’t see a significant reduction in conversion rate and sales, then you’re probably not going to see a significant increase if you improve your site speed by a similar amount. If you are faced with a high-cost decision around improving speed, using a test like this to gauge the impact of variation in site speed on your conversion rate is a fast and inexpensive way to get the information you need in order to make a solid financial decision.
How Should We Use PageSpeed Scores?
In general the right answer is to use PageSpeed scores and things like them as research tools. If they recommend some obvious and easy improvements, take that low hanging fruit. But don’t waste time and money on complex optimizations that cost a lot and make the site harder to maintain or harder for customers to use. Bottom line… if the site has a similar feature set and similar load times to competitors, it’s not going to have a ranking penalty relative to competitors. But definitely take action on any obvious and easy improvements, or if your site clearly underperforms competitors.
Above all, remember that PageSpeed scores are not actual speed. This is a metric that purports to be a proxy for speed. Some of their analysis and recommendations are flat out wrong, because they can’t take into account all the nuance of individual site architectures. This is also an automated calculation that doesn’t accurately capture the actual user experience for a real person. PageSpeed scores apply the same ranking system to all sites without regard to legitimate differences in real sites and legitimate trade-offs.
There’s no such thing as perfection, just trade-offs. But when making trade-offs and trying to figure out how much further you should go in site optimization, it can be useful to contemplate questions like “if we could improve organic search revenue by 1% by having a much faster site, would the labor and complexity required to do that be worth it?” For smaller organizations, it is not worth it past a point. On the largest sites, it clearly is worth it, given the scale. When you’re somewhere in the middle like most of us, it can be a bit murky.
Perhaps You Shouldn’t Bother?
Keep in mind that after loading the first page on your site, many elements are cached, so all subsequent page loads are faster than what is being measured by these metrics. The real user experience is not just loading a single page. PageSpeed scores and other site speed metrics tend to ignore this reality and merely sample a single out-of-context interaction with your website. Implementing costly improvements to shave off half a second of load time may only save that time on the first page load. If you think about this spread across a 5-page average session, that is only a slight 100ms average load time improvement.
Don’t worry about going crazy with complex optimizations that only shave off 100 milliseconds. Those just don’t matter for the vast majority of sites. Such small improvements in average page load times are not perceptible by users.
Finally, consider that the costs of improving PageSpeed scores may come in complexity. If you have to restructure the architecture of key systems on your site and modify core parts of your ecommerce platform to make it work differently, you are pushing more costs into the future. For example, platform updates may not go as smoothly, and may break functionality in the future. The ongoing cost in increased developer time may be higher than you expect.
Consider Amazon
A really low PageSpeed score, down at the bottom of the “this is bad” zone, is… well… bad. That indicates there is a lot of low hanging fruit that has probably been neglected. I’d tackle some improvements that are low cost and see where things stabilize after that. But at that point if you’re solidly in a middle to upper tier, I wouldn’t spend much time or money to improve scores on any one particular arbitrary scale like PageSpeed.
Amazon has a terrible PageSpeed score. That’s because PageSpeed scores are not, by themselves, important. Giving your users a great experience, helping them find what they want, and making checkout a breeze are far more important. If Amazon can be as successful as they are with a PageSpeed score in the low 40s, your site will do just fine with a score in that range as well. If you can score higher, go for it. But don’t sacrifice important user-focused features in your quest for a higher PageSpeed score.