Newsletters
Marketing measurement insights, delivered weekly.
The Halo and The Prescient Perspective — two newsletter series on MMM, halo effects, and budget optimization.
All editions
Meta shrank the window in January. Furniture and jewelry brands feel it first.
Meta removed 7-day and 28-day view-through attribution from its Ads Insights API on January 12. TileCloud's story shows what happens when a brand measures itself through someone else's decision — and what it looks like when it doesn't.
The memes about OpenAI Ads have not aged well. Real accounts are hitting 1.49x.
OpenAI Ads are hitting 1.49x ROAS in 15 days. Prescient launched a native connector to measure it. Plus how Quiet Owl scaled BillieBars 600% on YouTube by ignoring the dashboard.
Three new native connectors. And the BrüMate story that explains why they matter.
BrüMate shows how they scaled Amazon 85% by leaning into a channel the platform reports ranked last, CTV. A note on industry context this issue. Retail media spend in the US hits $71B in 2026, and most MMMs still cannot ingest those surfaces without a CSV. We shipped three native connectors on June 22 to close that gap.
Math matters. Here is what we bet on it.
Eight weeks of structural failure modes, all traced to the same architectural cause. The system we built instead, what it does, and the bet we are making.
Same model, different question. Catalina Crunch compressed CAC 24% in 30 days.
How Catalina Crunch used halo measurement to reallocate budget from underperforming channels and unlock 22% more efficient spend.
How DECKED found awareness was their best performance tactic
DECKED's worst-reported tactic was actually their best. Retargeting held 3% of its reported return.
The model gives credit by surface. The channel does its work across surfaces.
What halo actually is, why standard MMM cannot see it by construction, and what changes when the architecture can model what the channel is doing.
How TV channel share scaled 2.5x while media ROAS rose 11% in four months
How Beekman 1802 scaled TV 2.5x while watching media ROAS rise 11% across four months.i
The forecast inherits the bias, the math imposes a ceiling.
How today's error becomes tomorrow's plan, and why models lose fidelity exactly when losing it costs the most.
Why Global Healing almost cut the channel that turned out to be #1.
Global Healing found YouTube hiding in plain sight
Standard models don't always credit paid correctly at peak.
When calculating the revenue your channels generated, look closer at the equation. That is where the mismatch lives.
Linear TV CAC -22%, without changing spend
What Coterie changed about Linear TV without changing spend.
Lift test works, but only for a specific purpose.
What no test can correct when the underlying equations were designed for conditions different from yours.
Where does CTV revenue actually land?
What changes when you measure CTV with the right commerce mix as your reference.
There is a point where the next dollar starts costing more than it returns
Why every campaign has a curve, and why dashboards take weeks to show the bend.
We went quiet. Here's what we were doing.
You're on a rocket. Is the math powering it going to get you to your destination?
Adding a new channel can improve your entire portfolio. Or destroy it.
What dashboards miss when you scale, and how Portland Leather found out.
Why peak is where MMM math gets tested hardest.
Three things our research surfaced, in time for BFCM planning.
Most marketing teams are flying blind. Not you.
Why ROAS is the wrong metric for subscription brands, and what to measure instead.
No editions in this series yet.