Statistics
Bayesian Angels and Frequentist Demons
Why do Bayesian and Frequentist claims sound so different?
Learn more
Marketing data is more abundant than ever, however, it often feels out of reach. Eppo’s latest launch of GeoLift revolutionizes how marketers assess and grow your campaigns.
During our discussion with Greg Dale (Marketing Experimentation Leader at Eppo), and Bryce Casavant (Sr. Data Scientist at Whoop) we shared insights on the core challenges marketers face, discussed the transformative potential of incrementality testing, and illustrated the power of experimentation in building a data-driven culture. Here's what you need to know.
Greg hit on something we've all felt - that end-of-quarter moment when you're staring at attribution reports and MMM models, trying to figure out where your budget actually made an impact. His take? Stop guessing and start proving it through rigorous testing.
"At the end of the quarter...we all tend to look back and say where did my money go? Like, where's the actual lift in revenue and the metrics that I care about?" - Greg Dale
The reality check from Greg was refreshing: at Eppo, they've found that across all types of business experiments, roughly a third work as intended, a third do nothing, and a third actually hurt performance. Without proper testing, you're flying blind on which bucket your campaigns fall into. This isn't just about better marketing - it's about building a culture where decisions are based on evidence, not hunches.
Bryce walked us through how Whoop evolved their approach from relying on channel-specific data to focusing on actual business impact through GeoLift testing and MMM.
Many marketing channels make it easy to test and evaluate success using the metrics provided by that channel, but there are always issues. As Bryce stated, "One person might be on Facebook, go to Google, and then go to Pinterest and TikTok... It actually ends up being four [attributions] versus one." Unlike metrics controlled and manipulated by third-party platforms, business metrics from the data warehouse provide a precise picture of ROI. As Greg mentioned, "We actually test using business metrics—your actual purchases, your GMV [gross merchandise value]."
Some practical lessons from Whoop's experience:
The old way of testing geographic impacts usually meant comparing a couple of cities and hoping they were similar enough. Eppo's approach is much more sophisticated - they use data from multiple markets to build synthetic control groups based on historical patterns. As Greg put it, "Instead of picking one market, we're picking many markets and using past behavior to assemble a synthetic control."g
This opens up new possibilities. As Bryce noted, you can test incremental spend changes (like going from $1k to $2k in daily budget) multiple times to map out saturation curves and optimize your spending.
The big themes that emerged:
Final Thoughts
Marketing measurement isn't getting any simpler, but this conversation showed there's a path forward. By combining smart testing tools with a focus on real business outcomes, companies can start making decisions based on what actually works, not what they think might work. As Bryce concluded during the discussion, “Ultimately, if you're spending a little here [and] losing, you won’t be doing that again if you weren’t measuring it... Learning about your business ensures better outcomes, every time.”
Interested in learning more? Get a personalized walkthrough of Eppo’s GeoLift and experimentation platform.