📊 Pre-BFCM (October): Get Your Baseline - Run MMM to understand your saturation curves - Pick ONE channel you're suspicious about (we chose branded search) - Test it in 15% of geos for 2 weeks - Finding: Only 10% incremental at normal spend → likely 2% during BFCM Here's the key: Use October findings to optimize BFCM in real-time: - Branded search barely incremental? → Cap it at 1x while others scale - Found your breaking point at 2x spend? → Don't chase 3x vanity metrics - Retargeting only 15% incremental? → Reduce frequency from 10x to 3x - That's $300K saved to reallocate to channels that actually drive growth 🚀 During BFCM: Embrace the Chaos (Strategically) Two approaches that don't hurt revenue: Option 1: Natural experiments Your budget caps, geo differences, and inventory constraints create natural tests. Use them. Option 2: Positive scale test 80% of geos: Normal 3x scaling 20% of geos: Aggressive 5x scaling Compare the incremental difference 📈 Post-BFCM: Connect the Dots Compare your elasticities: - October baseline: 0.25 - During BFCM: 0.08 "But what if we miss revenue by testing?" Fair concern. That's why we suggest positive scale tests or using natural variance. No revenue loss. "This seems complex during our busiest season" It's really just 3 analyses: October MMM, BFCM documentation, December comparison. Plus, October tests give you a real-time playbook - not more work, just smarter decisions. "How do we know this applies beyond BFCM?" BFCM is like a stress test for your marketing. The patterns you discover apply to every promotion, product launch, and peak period. The Bottom Line: You're going to spend millions during BFCM anyway. Why not know which parts actually work AND optimize while you're spending? Pre-BFCM tests aren't just measurement - they're your optimization playbook. Same effort, 30-40% better efficiency.

https://vinaykarode.github.io/streamlit_mmm_101/granger_causality_reality_check.html Check out the video in the MMM course https://www.skool.com/marketing-science-made-easy-3352/classroom/659d7790?md=aecdaaee1b1048798a2df6d40850dec2

https://vinaykarode.github.io/streamlit_mmm_101/vif_math_whiteboard.html Full video in MMM course page - https://www.skool.com/marketing-science-made-easy-3352/classroom/659d7790?md=ff4446505dfd49498da12459039d7af8 "Should we trust our MMM?" My framework: - VIF < 5: Green light, but verify with holdouts - VIF 5-10: Yellow light, need experiments - VIF > 10: Red light, fix before using - VIF > 20: Your model is hallucinating

When your marketing channels move together, making it impossible to tell which one actually drives sales. https://vinaykarode.github.io/streamlit_mmm_101/multicollinearity_101.html

Here's a guide on what saturation curves are with some examples. https://vinaykarode.github.io/streamlit_mmm_101/saturation-curves-guide.html