The claim
“Bid high, it doesn't matter. It's a second-price auction. You only pay a cent above the next guy.”
I have heard some version of this from almost every seller who has run Sponsored Products for more than a month. I used to say it myself. It is a comfortable belief, because it turns the bid into a safety cap instead of a price.
The usual reading
The reading goes like this: Amazon runs a second-price auction. The winner pays roughly the runner-up's bid plus a small increment. So your bid only sets the maximum you are willing to pay, and in practice you pay less, often a lot less. The conclusion people draw is practical: set bids generously, let the auction discount you, and don't worry about the gap.
Premises, laid out
Every conclusion below rests on these. Reject one and you can see which conclusions fall with it.
The reading rests on three premises. I mark each one with where it stands for me today.
- ⚠️Amazon's own documentation describes a second-price mechanism. It did, for years: the wording was that you "only pay the least bid amount needed to win." That describes the documentation, not the billing.
- 🔶The auction clears at the runner-up's price. This is now the subject of litigation. In late August 2026 the FTC, joined by 22 states, sued Amazon, alleging that after the winner and the competing price are determined, a "soft reserve" is calculated, and if it is higher than the competing price, the advertiser is charged the soft reserve instead. The complaint alleges that the share of Sponsored Products clicks charged at the full bid rose from roughly 30–40% in 2021 to about 70% in 2022 and about 80% in 2024. This is an allegation. There is no ruling yet.
- ❌In practice, you pay meaningfully less than your bid. This premise I can check against my own accounts, and it does not hold.
How we take it apart
Free: conclusion + one screenConclusion: in my accounts, the bid behaves like the price, not like a ceiling. The discount the second-price story promises mostly isn't there.
The full step-by-step decomposition, including how AI joins the reports, is in the course.
Evidence
Free: one exampleExperience
Fourteen days, 17,237 clicks across my own accounts, in a clean subset where the most the auction can charge is exactly the bid (fixed or down-only bidding, zero placement and audience adjustment). Click-weighted, actual CPC ÷ bid came to 0.944, and 75.4% of clicks were charged 95% of the bid or more.
What would prove this wrong
Free: one sentenceDisproof condition. If a forward measurement — today's bids recorded, the next weeks' CPC observed — showed actual CPC ÷ bid settling well below 0.9 at ordinary bid levels, I would reopen this.
How you run it
Course onlyReproduce this on your own account
- Step-by-step operations, including the AI prompts used to join reports
- Query templates and the exact fields to pull
- A checklist to rule out the common reading traps
- The full disproof conditions and the failed cases
Ledger status
Verdict: ❌ doesn't hold — as a working assumption for bidding. Your bid is, most of the time, the price you pay. Set it from your own unit economics: the CPC your conversion rate and margin can carry, not the CPC you hope the auction will give back.
Type, in detail: Experience (own accounts); the legal question is open, no ruling.
Have data that says otherwise? Share a case. Reversals are published.