Google Optimize sunset: 5 ways to keep your experimentation program on track.
Well, after months of speculation, it’s finally happening: Starting Sept. 30, 2023, Google Optimize, Google’s tool for website experimentation and A/B testing, will no longer be supported or available.
With over 500,000 domains currently using Google Optimize—16% of which are among the top 10,000 most-trafficked sites—this is troubling news for the digital marketing world. As if the migration to GA4 wasn’t enough for digital teams to navigate, this is going to cause headaches for product, ecommerce, and marketing teams for months to come.
But all is not lost. As the great Jack Welch said, “When there’s change, there’s opportunity.” You could argue that Google Optimize played an important role in moving the industry forward; it was great for helping companies dip their toe in and start optimizing.
However, what it never did—or set out to do—is help product teams and experimentation leaders build testing into their culture. It had major limitations around page performance/flicker, targeting, and integrations as well as support that made it hard to operationalize at complex enterprises.
The good news is you now have the opportunity to rethink your investment in testing, uplevel your tech stack, and build a culture around experimentation. As Brian McIntosh, Chief Consulting Officer at BlastX, said, “Marketing technology is constantly evolving: as one solution is retired, new, better solutions emerge in the marketplace. With the announcement of Google Optimize’s retirement, now is the best time to explore new solutions for testing and optimization. Take a long-term view and identify solutions that will support the business outcomes you want.”
But where to start? Below are some ways you can increase ROI around experimentation, justify investment in world-class solutions like Optimizely, AB Tasty, and a data stack built on BigQuery, and create a culture of experimentation.
5 ways to uplevel your experimentation program.
Uncover hidden bias.
Which test won should always be a function of the quality of the creatives or segments. But too often, it’s biased by other factors such as page performance, traffic volume, errors, etc.
Having out-of-the-box tracking for these hidden biases drives confidence in experimentation outcomes and allows you to minimize the risk of the dreaded Type I or Type II errors.
Go beyond the winning test with behavioral analysis.
Understanding which test won based on statistical analysis is table stakes. Going beyond that to understand why is critical to ensuring that you capture appropriate learnings and apply them to future experiments. Leveraging behavioral analysis through tools like heatmaps and Session Replay is critical to building an experimentation culture
Leverage all of your data.
Personalization is still in its infancy, and it’s only as good as the data collected. Leveraging your organization’s first-party data to predict a user’s needs to trigger a real-time rescue before they do something adverse like abandoning or calling the contact center is one of the most leveraged ROIs you can get on an experimentation program.
You can also use data to form hypotheses (and to avoid testing bad hypotheses). Often one of the most complex parts of testing is coming up with a good pipeline of ideas to try, so don’t frivolously test stuff that won’t move the needle.
Remove bottlenecks by running more tests with auto-track.
We hear from experimentation leaders that manual tracking code is one of the biggest bottlenecks to experiment velocity. Needing to implement code to track every little piece of an experiment is a tax on development resources and slows down the ability to learn. Leveraging an auto-track technology allows you to focus your tagging on the primary success metric and let all secondary metrics be captured automatically.
Tap into BigQuery analysis.
Analyzing clickstream data, offline data (CRM, in-store purchase, call center, etc.) is a hallmark of a mature experimentation program. It allows you to identify hidden cohorts, build LTV and predictive CLV models, and prioritize opportunities for improvement. This can be more trouble than it’s worth if it requires length ETL, batch processing, and data merging.
Having data natively available in BigQuery from solutions like Quantum Metric, Optimizely, and AB Tasty makes this process easy and allows your data scientists to focus on the fun stuff.
Start generating better results with Quantum Metric.
The coming months promise macro and micro headwinds, but investing in a stack for the long term can help you grow through economic storms and emerge as a leader. Interested in learning more about how Quantum Metric helps digital teams generate better results? Request a demo today.