Measurement capabilities of world class advertisers, incrementality testing, media mix models, and when to use attribution | Andrew Covato (Growth by Science, Snap, Netflix)

Andrew has over fifteen years of Ad and MarTech experience at eBay, Google, Meta, Netflix, Snap, and fintech startups.  He has tackled hard challenges in product, strategy, growth, go-to-market, and partnerships across many verticals and geographies.  He is highly strategic, and has experience building and applying buy- and sell-side ad technology at scale. He has been advising companies from start-ups to Fortune 500 enterprises since 2012. 

---

We talk about: 

• Varying measurement methods and when to lean in on one or two of them 
• How the most sophisticated advertisers use a persistent hold out 
• The best way to approach media that has to run nationally 
• The differences between measuring advertising and channel effectiveness and optimization impact 
• The variables needed to structure effective measurement methods  

--- 

Where to find Andrew: 

Andrew's LinkedIn Profile: https://www.linkedin.com/in/andrewcovato/

Growth by Science: https://www.growthbyscience.com/

---

We reference: 

https://www.measured.com/blog/geo-testing-series-test-design/

https://www.measured.com/blog/geo-testing-series-part-2-test-management/

https://www.measured.com/blog/geo-testing-series-part-3-analytics-and-reporting/

---

Where to find Patrick:

Patrick's LinkedIn Profile: https://www.linkedin.com/in/pcmoran/

---

(2:07) History of advertising and contextualizing the ideal measurement stack 

(7:40) Options for measurement - Geo testing, using media mix models, time series analysis, and newer versions of multi-touch attribution 

(12:55) Using a measurement tax effectively to value advertising and the consequences of using attribution 

(15:16) The most effective ways to impact advertising using creatives, reach, and frequency and not playing around with bidding 

(16:50) Using upper funnel metrics to inform optimization approaches 

(18:54) Situations where attribution is applicable 

(21:46) Budget approach required to run these tests and committing to the process to detect signal 

(25:44) Approaching marketing level impact first and then measuring channel impact 

(30:26) Using a constant holdout, when it works and when it doesn't work

(33:39) Simplicity and specificity when setting this up