Intelligent Website Testing Improves Performance

Data-driven website testing has a dramatic effect on user experience and digital revenue


There are millions of potential ways that content can be structured on your website. The layout, position of the content, and combination of ads all affect the way users behave on your website. If you can balance these millions of data points, websites can lower bounce rates, increase session duration, increase pageviews per visit, and improve revenue per user.

Testing is the only way to know how users will respond to different elements on your website.  Understanding just how user experience affects session length is critical in balancing things like ads and content. Finding the perfect balance of these elements is actually quite difficult without lots of data and the ability to test things in combination with one another.


website testing performance

(On the importance of testing) …At any given point in time, there isn’t just one version of Facebook running, there are probably 10,000.




website testing metrics
website testing platform


How content is organized, balanced with ads, and formatted on the page can affect user engagement just about as much as the content itself. This can be measured by looking at core user experience metrics like bounce rate, time on site, and pageviews per visit; which have statistical correlations with everything from search engine traffic to digital revenue.

Here, you can see the direct correlation that time on site and pageviews per visit have with digital revenue. This may seem like a no-brainer, but when making data-driven decisions are these factors always weighted the same? In fact, there are actually few things that have a more dramatic impact on digital revenue that these factors.

What if you are making short term decisions that you think are benefiting revenue but ultimately harming the length of user sessions — therefore ultimately harming your site’s long-term health and ability to maximize digital revenue? When optimizing for CPM’s, we’ve learned that is is very often the case…


It is common for digital publishers, website owners, and bloggers to become very focused on CPMs when attempting to optimize for digital revenue; however, it’s important to realize that CPMs are not directly responsible higher digital revenues. In fact, it is common for CPMs to be increasing while overall website revenue is decreasing.

As you can see in the graphic, the scenario with the lower eCPM (effective CPM) is actually is earning more per thousand website visitors than the site with a higher eCPM. This is because of the decreased pageviews in the second scenario with the higher CPMs.

This is is not uncommon as high-profile ad sizes & locations — and even more disruptive ads — often have higher CPMs. While those ads might earn more money, your visitors will abandon the site, costing you all the additional impressions that may have occurred had they visited other pages. The only way to know to the impact of this is by looking at how different changes affect EPMV (earnings per thousand visitors.

The only way to know to the impact of this is by looking at how different changes affect EPMV (earnings per thousand visitors). This is an easy metric that any publisher can calculate by dividing their total earnings per thousand website visitors.


Traditional A/B testing would say that if 60% of website visitors liked variation A best, A would win. Under that logic, you would implement an all A solution. But what about users that preferred variation B?

The optimal experience would actually be one that allowed for an A/B/C/D/etc. implementation — one that links different variables to the visitors who prefer the best. This is why multivariate testing and preferred implementation is so effective at improving user experiences on a publisher’s website. If you know how visitor behavior is affected by different variables, you can strategically deliver them a manner that enhances user engagement for everyone who visits your website.


It’s easy to make assumptions and optimize for data averages, but this can often lead publishers astray. Examining how ad impressions affect bounce rates is a great example of this.

In this example, you can see how ads affect the bounce rate on this particular page — on average — but some visitors may actually prefer fewer ad impressions than the threshold displayed here. Visitors that bounce when more ads are shown may need to see fewer ads to get them to stay on the site longer or visit additional pages (increasing EPMV). Knowing how different types of visitors will respond to ad impressions is paramount to optimizing your site’s user experience metrics & revenue.


As you might expect, testing isn’t complete in 24 hours; however, digital publishers using Ezoic see massive improvements in EPMV the longer they test using the platform.

This model of testing is never finished and progress never stops. The more data you have, the more a smart system like Ezoic can use to deliver better experiences to users. Furthermore, user behavior shifts all the time. Testing with a system like Ezoic is just as much about delivering every user the optimal experience as it is actually measuring their behavior.

With Ezoic, testing is one part measurement and one part delivery.


Publishers keep learning more and more about user behavior and user intent. One of the clearest things we’ve extracted from the data recently is that not all user experiences are equal. Visitor engagement has a massive impact on both revenue and website traffic. Often, user experience metrics, like pageviews per visit, can become artificially inflated; causing publishers to be fooled and revenues to take a dive.

Publishers should be accounting for artificially inflated user experiences when they begin testing. That’s why it’s important to track things like navigation bounces, engagement time, return visitor rates, and more every time a test is run.


This is exactly why Ezoic was constructed and exactly how Ezoic works. The platform contains a number of testing applications that gives digital publishers and website owners the ability to do comprehensive, data-driven website testing.

Ezoic allows platform users to automatically test ads, content, layouts, and much more. Artificial intelligence does the hard part, and you’re in total control of the experiments and how the solutions are delivered on your site.


The ability to test, optimize, and deliver every visitor a better experience is simply too hard for a human. There is too much data to sift through and analyze. That’s why artificial intelligence can deliver real value to website owners and digital publishers.

A.I. and machine learning have become industry buzz words, but few are actually leveraging real learning technology. This is crucial because it would be impossible to optimize all of these factors at scale over time without a technology that can actually learn as it goes.


You can start by only partitioning a very small portion to Ezoic for testing at first.