Pillar 03: Amplification
Each week's The rest of the field read in one place, with the full log of standing rules and the claims we retired.
Trends
Trend charts start at 4 weeks of data. 1 week logged so far.
Week by week
Standing rules, the full log
7 rules, newest first. A rule that replaces or refines an earlier one says what it supersedes.
| Rule | Area | Change | Logged | Supersedes |
|---|---|---|---|---|
| Add these channels after Meta and Google are working against your contribution margin target, not before. | Sequencing | new | Nothing | |
| Every new channel test gets a control group: some regions, households or weeks with the channel off. No control, no scale decision. | Measurement | new | Nothing | |
| Don't judge streaming TV, Pinterest, Reddit or Snapchat on the platform's last click. Read new customers, branded search and MER over the same window. | Measurement | new | Nothing | |
| A lift study a platform runs on its own ads is a claim until your own control group repeats it. | Evidence | new | Nothing | |
| Read Amazon and DTC together. Sponsored Products tends to take credit for shoppers already looking for you, and streaming TV often shows up on Amazon before it shows up on your site. | Retail media | new | Nothing | |
| Don't start a new channel test between Black Friday and Cyber Monday. Prices and noise wreck the read. Launch in January. | Timing | new | Nothing | |
| With automated campaigns (Amazon DVA+, Pinterest Performance+, AppLovin), pick the products and the goal before launch. Changing either later restarts the learning. | Automation | new | Nothing |
Retired claims, the full log
2 claims we stopped repeating, and why.
| Claim | Why we retired it | Retired |
|---|---|---|
| Netflix ads now offer 4 million data segments. | The 4 million is the size of The Trade Desk's whole data marketplace, not segments built for Netflix. | |
| AppLovin's ecommerce ROAS is comparable to the major social platforms. | AppLovin's own claim, repeated by agencies. No figure, period or method has been published. |
