It’s important to remember that all sites are different. The tests that work for one site, won't necessarily work for all sites.
Not everything the system tries is going to work. By the nature of testing, the platform will need time to find what your users respond to best.
For example, a site might have a naturally low time on site and high bounce rate (let's say it's a reference site, an online tool, or a calculator - where most users just get their information and leave). How that site is optimized is going to be different to how it tests a food blog with high page views per visitor and time on site. That might sound obvious, but it's about being realistic with what the users are on the site for. Stepping back from the content and asking questions like:
'If 90% of all your site's visits are 'new' vs repeat visitors, should the system test a higher than average ad density, or lower than average ad density?'
That's what the Ezoic platform works out. Eventually - an optimal layout emerges (in terms of ad placements and navigation) for each site; until of course it's replaced with something better.
That's what continual testing is all about. Testing is the only way to find out what works and what doesn't.
Example Site - Seasonality and Patience
This example shows how this publisher's patience was rewarded. In the first two to three weeks, the optimal layouts had not yet been promoted (so there was dilution from other test - detracting from the overall income). Then Spring Break/Easter Holiday kicked in and the site starts kicking in some very good income numbers when a new - better performing layout starts to win more traffic.
Desktop 'Variation-46' gets promoted to win more traffic after 6 weeks - kicking income improvement up again from 40% improvement in EPMV during the first 6 weeks, to 263% improvement.
Example 1 - Naturally High Bounce Rates & increased Ad density
The naturally higher bounce rate site will have more ads within the content when compared with an article site with high time on site and page views per visit.
This is because more 'new users' (who are not going to return to the site within 30 days), are coming to the site, see the content they want and will leave no matter what. Our system will learn that showing those users an ad before they leave makes sense.
If the bounce rate is naturally high anyway, then the number of ads on a page can be 100% sustainable. Take a look at any dictionary site and you can see this in action.
It's a counter intuitive point - but it is entirely predictable to be able to increase the number of ads on a web page, improve navigation at the same time and see both revenue and UX metrics improve at the same time.
Example 2 - Mobile is increasing. Better Mobile ads means improved income.
Mobile is increasing year on year. It is important to remember that tablet enabled and mobile enabled sites will get more traffic over time when competing against sites that are desktop only. Sites that have good mobile menus with Ezoic will also be automatically ad enabled for all screen sizes. The increase in ad income from adding correct size ads for mobile/tablet can be substantial.
Example 3 - Easy wins
We've all seen sites that look like they haven't seen any layout change since the 1990's. No menus, no social functionality, or mobile layouts, but still getting huge traffic because they have so much original content. These are easy wins for the Ezoic system, as they have both the traffic for the system to make quick optimization decisions, and will also add a ton of value because the old layouts are easily 'beaten' by testing newer layouts of the same content. These are slam dunk sites, and get quickly up to 4x-5x income improvement without the owner having to change a single line of code.
Example 4 - A typical site
Most sites that are plugged into the platform fall into an average income and UX improvement trend of 50%-200% increase in income. This is where there is a mix of mobile uplift, bounce rate optimization, and easy wins all combine to improve the overall site. Most people are happy, after monitoring the results daily for a week or so, to let the system do its work and come back in a month or so to see the results.