Marketers are always looking for ways to improve the response to their marketing content. And between the increase in connected devices and the ubiquity of programmatic campaigns, that need is greater than ever before.
That highlights the need to vet and optimize marketing content aimed at attracting customers and to meet that need, analytics providers have been steadily improving the quality of their A/B testing solutions.
Integrated Testing and Analysis
Google Optimize, a testing solution Google has added as part of its Analytics 360 Suite, is a good example of how the testing and analysis functions are becoming more integrated, marketing-friendly and easy to execute.
Originally offered as a tool to compete with Adobe Analytics, Google has recently expanded Optimize to mesh more closely with Google Analytics.
When users initialize Google Optimize, they add it to their websites as a container, much like how they would add a script or a tag manager container. An Optimize container can be labeled with text up to 255 characters long to identify a campaign.
Once Optimize has been added, users can then design experiments to test website pages or web apps. Through these experiments, marketers can determine which page or app changes have been most effective at achieving their specific marketing objectives.
Page Variants Without Recoding
Probably the most straightforward benefit from Optimize is easier usability. Landing pages can be set in a visual editor, which allows marketers to create new variants of pages without requiring site recodes each time a test is planned. Mobile variants of tests can be planned within Optimize, and the editor also allows for easy code diagnostics.
3 Types of Testing
Optimize allows for three types of experiments:
- A/B: These are random tests of two or more variants of the same website page. Each variant is served separately but at a similar time, so that the relative performance of each variant can be observed while holding other external factors constant.
- Redirect: This is a split test that pits separate website pages, such as different landing pages or page redesigns with different URLs, against each other.
- Multivariate (limited): This form of testing is used to inspect two or more elements simultaneously to determine which combination will perform best.
Formulating Testing Objectives
Once a user has decided which type of experiment to perform, Google Optimize prompts the user to identify objectives for that experiment, which can be thought of as hypotheses or educated guesses about the expected results. Up to three pre-selected objectives can be chosen within a given experiment.
Objectives in Google Optimize play the same role as goal setting in Google Analytics reports, namely explaining how well particular website or web app elements meet the original hypotheses.
Targeting Audiences and Actions
Targeting offers a way for marketers to add parameters of limits to a specific audience or factor in other critical influences on an experiment.
Google Optimize then lets marketers select the percentage of site visitors who will see an experiment. Optimize also allows for URL or path rules to be set so that participants must navigate or conduct important actions before seeing the experimental test page or web app element.
Enhanced User Forum Support
To gain some insights into what others have experienced with tests, Optimize also makes a user forum available.
This forum support offers a prime example of how Google has continued to improve its solution-specific help center resources, building on its improvements to Google Analytics over a year ago.
Evolving Testing Through Machine Learning
Optimize is among the latest analytics tools to introduce machine learning techniques to testing. Google has already adopted machine learning in some of its analytic reports such as Smart Goals.
Adobe applied a similar machine learning upgrade with its A/B test solution, Test and Target, and more solutions are likely on the way.
For example, at F8, Facebook announced the inclusion of artificial intelligence and machine learning updates to its Automated Insights, with the goal of helping marketers reduce analysis time and create more rapid action plans for deep-dive analysis.
In addition, analytics providers are increasingly deploying features that allow for enhanced decision-making in conducting optimization testing.
This approach improves testing accuracy by running tests where the appropriate sample sizes are determined by real-time conditions. It also allows marketers to avoid manual errors such as stopping a test too early.
Contextual Statistical Analysis
Machine learning protocols also introduce the possibility of using Bayesian statistics to support test conclusions. Bayesian probability determines outcomes based on the context of the data, analyses made possible by the better computational capacity associated with machine learning protocols.
By contrast, most A/B testing solutions produce results based on frequency, which can lead to test results that measure frequency of arriving at a given website page. Bayesian statistics, however, allow for consideration of conditions that influence when that page was accessed.
Connecting Better With Customers
Overall, Google Optimize offers a great way to optimize user response to digital content through testing. And as Optimize democratizes testing, better experiments will lead to better analysis and decision-making and ultimately, better connections with customers.