This talk is about a quest for data. It’s important to own your own quest for data rather than relying on third party data.
We believe that slow sites have a bad UX and that faster sites convert better. We know that site speed is a ranking factor. But we don’t want to rely on belief.
We can’t rely on belief because SEO is a competition. We need to know what we’re competing against. How good are our competitors?
This led Nichola’s company to create the mobile site speed benchmark, available now for free . All of the raw data is available to have a look around.
So how did they get there and what can we get out of it?
It began with an R&D project where they tried to turn their agency site into a progressive web app. They wanted to show their clients what’s actually achievable. The key thing about PWAs is that they can be added to smartphone home screens and they can send push notifications.
They then wanted to see how many other people are using PWAs. They audited the top 1000 UK websites to look for 5 key criteria, including PWA-specific stuff and general good practice stuff, including service workers (PWA-specific), HTTPS) and responsive content.
They made use of Lighthouse on those 1000 sites manually and discovered after the fact that they could have done it automatically! But it wasn’t worth it…barely anyone is doing any of those things. Httparchive also has all these results already in BigQuery.
Using a lot of data from HTTParchiveErudite had data from 5000 UK sites organised by 7 different critteria. They checked the sites for currency (200 status code), big outliers and spam, which got them to 5010 sites. They then had to classify them by some broad categories to allow them to compare meaningful chunks of data, which they did using Majestic’s topic data.
The biggest categories they found are retail, business services and news & media. These categories can give us some meaningful information.
In terms of time to first byte there’s not a lot we need to know. It’s not much help except as a benchmark for the future. If we now look at visually complete and consistently interactive points, we see that the big three categories are the slowest, as they all have loads of assets to load.
If we look at retail sites, the biggest category, in more depth, we can see some more interesting data points. The top quartile for FMP (first meaningful paint) loaded in 2seconds. For content loading, the median is 3.75, top quartile is 2.71 and bottom is 5.31. These are benchmarks that we can use to compare our clients sites with and build a business case.
At the consistently interactive point, the median was 9.4 seconds and the top was 6, with the bottom quartile at 14.2 seconds! The recommended speed from Lighthouse is 3 seconds, so this is pretty shabby.
But how do we actually use all this data?
We can set smart goals. We now know what we’re aiming for so we can aim for certain, measurable increases based on what our top competitors are doing. We can then use Google’s calculator to work out how much more revenue our clients could get with the speed boost we’re targeting.