In 1961, William Vickrey proved at Columbia that in a second-price auction, honest bidders are safe because they never pay their own bid. Over a million Amazon sellers built ad budgets around that promise. The FTC alleges the promise was hollow by 2019, and the system designed to protect honest bidding became the tool of extraction.

Here is what the proof means in plain terms. You bid ten dollars. The next highest bid is six. You win and pay six dollars and one cent. There is no reason to bid low. The system shields you from paying your own number. Thirty-five years after Vickrey wrote that proof, the Nobel committee gave him the prize for it.

Bid your true value and the system protects you. It sounds right. It is not.

Amazon told more than 500,000 small and mid-size businesses that its ad auction was second-price. Bid high. You will only pay a penny above the runner-up. The logic traced straight back to Vickrey's proof. Sellers set their ad budgets around this. They planned margins around it. The system ran as described from 2012 onward. For seven years, the promise held.

The FTC alleges that starting in 2019, Amazon added a hidden charge it called a "soft reserve price." Unlike a normal reserve, which is set before the auction and visible to all, this one was calculated after the auction had already cleared. After the winner was picked. After the real bids had set the competitive price.

If that reserve came in higher than what real bids had produced, the complaint alleges Amazon charged the reserve instead.

An Amazon senior vice president wrote it plainly in an internal document. The second price, he said, "isn't set by an actual bidder" but is "a proxy 2nd price that we calculate."

The internal name for the result: "surchargedSecondPrice." Another document used the phrase "invented auction participant." The FTC calls these what they look like: shill bids.

The numbers show what followed. In 2021, sellers paid their own full bid between 30% and 40% of the time. By 2022, that figure hit 70%. By 2024, it reached roughly 80%.

A system sold as second-price had drifted to first-price. No one told the sellers.

Internal documents quoted in the FTC complaint show this was not accidental. One described the goal plainly: tune the system to raise prices "while hoping that advertisers don't notice and decrease bids or ad spend."

Amazon believed disclosure could cause "irrevocable damage to advertiser trust." An Amazon scientist was told to remove the word "surcharge" from an internal slide. The team picked a new term: "performance premium." The practice was hidden from Amazon's own sales staff.

One grocery seller's account manager flagged the damage during Prime Day 2023. The client's cost per click was up more than 90%. Return on ad spend was falling apart.

The FTC and attorneys general from 22 states allege Amazon extracted more than $20 billion from 1.2 million sellers through these hidden charges.

Amazon says average cost per click stayed flat when adjusted for inflation. The company says the FTC is leaning on a handful of simplified documents pulled from 1.5 million pages. A flat average bid does not answer the question of what sellers actually paid. The FTC's case is about the gap between the competitive clearing price and the final charge. Amazon's response does not close that gap.

The problem is not that sellers bid too high. The problem is that the vendor running the auction also controlled a hidden input, one set after competitive pricing had already done its work. Any system built that way will drift toward higher costs. The pull is structural. It does not require bad faith. It only requires a lack of outside eyes on the dial.

Most people treat this as a platform problem. It is a system flaw.

The replacement is not "bid smarter." It is this: do not audit the invoice. Audit the mechanism that sets the price.

Once that is clear, three moves follow from it.

For any platform where you pay for placement, ask three questions. Who sets the price. Who profits when the price rises. Whether those are the same party. If all three answers point to one entity, the system has a built-in drift toward higher costs. That mapping takes ten minutes and changes how you read every bill that follows.

In a real second-price auction, a gap holds between what you bid and what you pay. Pull your ad data for the past 90 days. Plot bid versus actual cost per click. If the gap is closing, the mechanism changed. You do not need a federal complaint to tell you. The data will.

Cut paid ads for one product or one service line for 30 days. Measure the gap between paid reach and organic rank. That gap tells you what the ads are buying versus what the platform is taking as a toll. If your organic rank barely moves when you stop paying, the ads were buying access you already had.

Running these three checks over a single quarter does something the old advice never did. You see which platforms charge you to reach people who were already going to find you. You see where the spread between bid and payment has quietly closed. You see which line items on your ad spend are buying results and which are paying a toll to a vendor who also writes the rules.

At the end of that quarter, ask three things.

→ What moved the number: the bid, or the hidden input behind the price?

→ What looked like a cost of doing business but left no trace when you removed it?

→ What friction showed up more than once, on more than one platform?

That is the difference between checking the receipt and checking the register.

Vickrey's math still works. A second-price auction still protects the honest bidder, when the system running it is honest. The lesson applies anywhere the vendor who profits from the price also writes the rules that set it. The mechanism is only as clean as the hand on the dial.