Short reads, full premises.
Each note judges one popular claim or exposes one reading trap. Premises first, one example, and the condition that would prove it wrong.
Claims, judged
A popular claim tested against data, with a verdict that goes into the ledger.
TYPEReading traps
What a report actually counts, and how that fools a careful reader.
TYPEOur own claims
Things I have said myself, put through the same test. Reversals stay on the record.
Claims, judged 6
A popular claim tested against data, with a verdict that goes into the ledger.
- "It's a second-price auction, so your bid is just a ceiling"❌ Doesn't hold
On my accounts, click-weighted actual CPC ÷ bid was 0.944. The FTC allegation of a soft reserve is unresolved; no ruling yet.
- "Raise the bid and you'll get more impressions"❌ Doesn't hold
Bid sets how actively you reach shoppers you can already be matched to, not who they are or how ready they are; own campaigns compete for the same auction; capped budgets turn higher bids into fewer clicks.
- "Ad sales push up your organic rank"⚠️ Partly
Conversion rate carries more durable signal than click volume, but both effects are smaller than weekly rank noise. Ads → organic orders is invisible in observational data, yet a two-week full switch-off on one account cut non-ad visits about 35%. Judge per account.
- "Attributed sales are the sales your ads brought in"⚠️ Partly
Close to incremental when ads reach people who don't already buy from you; overstated when they reach people who do. Differs by campaign type.
- "Top of search converts best, so push your placement there"⚠️ Partly
Cheapest per order on high-repeat products, but most expensive per new customer. Its impressions mostly reach new shoppers; returning customers dominate at conversion. You can't buy who stands in a slot, only how much you buy there.
- "Lower ACOS is always better"❌ Doesn't hold
ACOS is a verdict line for harvest campaigns, not a goal for every campaign. A low ACOS can mean the ad stopped finding new customers or ran out of budget.
Reading traps 3
What a report actually counts, and how that fools a careful reader.
- Reading trap: the ad report shows zero on the 1st of the month— Reading trap
Report days follow the account's time zone. Ahead of it, the account's 'today' hasn't started yet.
- Reading trap: the Business Report doesn't match the number you saved last month— Reading trap
Revised for weeks: units drift down, sessions drift up. Store the pull time; re-pull before comparing; prefer ratios.
- Reading trap: AMC says 100% of your ad buyers are new customers— Reading trap
Privacy masking blanks small cells, and returning customers live in small cells, so fine-grained new-customer shares come out too high.
Our own claims 1
Things I have said myself, put through the same test. If one flips, the reversal stays on the record.
- Our own claim: "It's not ACOS, it's TACoS" — does it still hold?⚠️ Partly
The principle holds — judge ads by what they do for the account. The ratio is a gauge; proving ads drove growth takes holdouts, controls, and placebo arms.