A portable seat cushion. Thirty Prime shoppers. One image took 63% of the vote in a five-way test. The winner had a small US flag badge in the corner. Not one shopper mentioned patriotism.

Here's a number that should stop you cold. In a five-way main image test, one image took 63% of first-choice votes. Nineteen of thirty shoppers picked the same one. Five options on the table, and one pulled six times the runner-up.

Five-way tests almost never break that hard. When one does, you want to know exactly why.

This test comes from Daniela Bolzmann, who's been mining hundreds of PickFu polls over in her Social Proof newsletter. She ran it to thirty US Amazon Prime shoppers on a portable inflatable seat cushion. Same product in every image. What changed was the angle, the badges, and the on-image copy.

The winning image had three things the others didn't: color swatches, a weight-capacity badge, and a small "US OWNED" flag badge in the top corner.

When the comments came back, roughly a third of the panel brought up that flag on their own. Nobody asked them to. They just kept circling back to it.

Then came the part that matters.

Nobody said patriotism

Read the reasons shoppers gave and you expect flag-waving. You don't get it. You get trust.

"US owned signals quality and safety," one wrote. Another said it told them a real company stood behind the product. A third just liked seeing it front and center because it mattered to them.

Notice what's missing. No pride. No "buy American." No allegiance to anything. These shoppers weren't voting their politics. They were answering a completely different question: is this a real business, or a drop-shipper?

That's the whole game.

Why the flag does the heavy lifting

On Amazon, a shopper makes a snap call with almost nothing to go on. They can't hold the product. They can't ask a question. They can't tell a real brand from a listing that got spun up last Tuesday. So they hunt for shortcuts. Anything that answers "can I trust this?" without making them stop and read.

The flag badge is one of the fastest shortcuts there is. It tells the shopper nothing factual. But it registers in half a second as "this one's probably legit." It stands in for a real company, real accountability, someone you could actually reach if the thing showed up broken.

That's why it converts. And that's exactly why it can turn on you.

The tripwire: quiet reads as trust, loud reads as suspicion

This is where most sellers blow it. They see "flag won the test" and crank it to eleven. Giant flag. Bold "MADE IN USA" across the image. A line about "no cheap Chinese ingredients."

That version loses.

A small corner badge says a real brand stands behind this product. A hammered-in flag says a company is trying way too hard to convince you of something. The us-versus-them framing, the jab at another country, pushed real buyers away in testing. Shoppers called it "try hard" and "insincere."

Here's the kicker. The words on the badge were identical. The only thing that changed was the volume and the framing around them. Quiet won. Loud lost.

"US Owned" is not "Made in USA," and that matters

One more thing worth flagging, because it can bite you.

The FTC requires anything labeled "Made in USA" to be "all or virtually all" produced in the US. That's a high bar, and it comes with real penalties for brands that claim it without backing it up.

It's a bar plenty of American icons can't clear. Levi's, Rawlings baseballs, Harley-Davidson: names built on Americana, with plenty of production happening overseas.

"US owned," "US based," or "founded in Austin" makes a claim about the company, not about the origin of every last component. You get most of the trust signal with far less exposure. If your product isn't genuinely all-or-virtually-all domestic, do not put "Made in USA" on the image. Pick the claim you can defend.

How to actually use this

Treat the flag as a trust cue, not a flag.

The flag isn't a patriotism play. It's a firm trust signal in badge form. Small, it tells a shopper a real company is behind the product. Loud, it tells them a company that needs you to believe it. Build for the first one.

Hat tip to Daniela Bolzmann and her Social Proof newsletter for the test data.

Building a massive e-commerce empire sounds like the ultimate dream, until you realize that your friends and family have absolutely no idea what it feels like to lose an $80,000 shipment or have a top-selling Amazon listing suspended overnight.

In this deeply authentic episode of Marketing Misfits, hosts Norm Farrar and Kevin King sit down with Drew Littlejohns, an e-commerce veteran who went from being homeless and living in a church basement to running 8-figure brands.

He pulls back the curtain on the hyper-aggressive optimization culture of e-commerce that often leads to severe anxiety, isolation, and addiction. Plus, the guys dive into the future of Amazon SEO and why the "Dunning-Kruger effect" in AI is making sellers dangerously overconfident.

If you are feeling the burnout of the Amazon grind or just want to know where the future of AI-driven product discovery is heading, you cannot miss this episode.

🔥 Also subscribe today to the Marketing Misfits Newsletter!

Amazon is shutting down Mechanical Turk. For 20 years it quietly powered your favorite research tools. Before that, it ran a whole generation of black hat ranking tricks.

On June 30, a small notice went up on the Mechanical Turk website. As of July 30, 2026, the service closes to new customers. Existing users can keep going for now. AWS also slid it onto its "Services in Maintenance" list, which is corporate speak for a service circling the drain. Security only. No new features. And no real reason given.

Most sellers reading this never logged into Mechanical Turk in their life. But you have almost certainly used it without knowing. And if you were selling in the early days, some of your rankings were built on it.

Here’s the story.

What Mechanical Turk actually was

Amazon launched it in November 2005, a year before AWS became the cloud giant it is today. The idea was simple. Some jobs are easy for humans and hard for computers. Tagging an image. Transcribing audio. Judging whether a sentence sounds happy or angry. So Amazon built a marketplace. Companies posted the tasks. Regular people around the world did them for pennies.

Amazon called it "artificial artificial intelligence." Bezos coined that phrase himself. The work looked automated to the buyer, but a human sat behind the curtain.

That name is a joke with history. The original "Mechanical Turk" was an 18th century chess machine that beat real players and amazed royalty across Europe. It was a fraud. A human chess master hid inside the box and worked the levers.

Amazon named its service after a famous hoax about a machine secretly run by hidden humans, then built a business doing exactly that.

Each task was a HIT, short for Human Intelligence Task. The workers called themselves Turkers. Pay ran from a penny to a few dollars. The Turk pre-dated Fiverr, Upwork, and the whole gig economy. It was the original.

How your tools ran on it

Here’s where it touches you.

Think about how PickFu works. You upload two main images. You get back real human opinions in minutes, with written comments. Feels like magic. For years, part of that magic ran on Mechanical Turk. PickFu operated as a Turk requester, so some of those "real shoppers" reacting to your hero image were Turkers picking up a HIT for a few cents.

PickFu has since moved to professional consumer panels, the same providers big CPG brands use. But the early version of that whole category leaned on the Turk.

Same for a long list of survey tools that promised quick opinions from "real people." Even academic research leaned on it. So the next time a tool hands you fast human feedback, remember there is always a human somewhere, and that human costs money. Which is a big part of why this is dying.

The black hat era

Now the part every old-school seller remembers.

Before Amazon got smart, the Turk was a weapon. Sellers, and the shady services they hired, posted HITs that had nothing to do with honest research. Search a keyword. Scroll past the competitors. Find this exact product. Buy it. That was search find buy, and the algorithm treated those keyword-driven purchases as proof shoppers wanted your product for that term. Rankings jumped.

Other HITs told workers to add to cart, add to wishlist, sit on a listing, then come back and buy days later to fake a shopping journey. Then the reviews. Buy through a normal-looking Turk purchase, get reimbursed, leave a glowing five star review that now counted as "verified." Multiply that across a crowd and you launched a product with a review wall that looked organic and was anything but.

This was not a rumor. A well-known Stern School study found more than 40 percent of Turk tasks were spam. The flagged categories read like a black hat menu. Fake ratings. Fake reviews. Fake likes and votes. Fake accounts. Fake clicks. Real humans doing fake things, which is exactly why the spam filters of the day missed it.

Amazon caught on. It rewrote the terms, purged reviews by the tens of millions, built detection, and sued the fake review brokers. The tricks that built empires in 2014 became suspension in one email. That is the history, not a how-to. Run these plays today and you lose the account.

Why it is finally dying

The irony is thick. Mechanical Turk existed because computers could not do human tasks cheaply. AI can now do most of them for a cheap API call.

Add years of bots, fraud, and silent bans, and the platform was already a ghost town. Amazon also has replacements it would rather sell you, like SageMaker Ground Truth. The Turk is not needed anymore, and Amazon rarely keeps what it does not need.

Amazon just retired a 20-year-old service that shaped an entire era of selling, with a one-line notice and zero explanation. That is exactly how it operates. Tools vanish. Programs sunset. Loopholes close overnight. The platform under your feet is always moving.

The Turk was born as a story about a machine secretly run by hidden humans. Twenty years later, machines are killing it because they no longer need the humans at all.

Build on what you control. The rest is on borrowed time.

Sixteen months ago, SPS Commerce paid $210 million for Carbon6.

Last week it sold the third-party seller piece of that business for $9.5 million.

Read that again. $210 million in. $9.5 million out. And it booked a $20 million loss on the sale.

That is not a typo. That is a fire sale.

What actually happened

On June 30, SPS Commerce (NASDAQ: SPSC) announced it completed the sale of its "3P Revenue Recovery business." That is the part of Carbon6 built for third-party Amazon sellers. Think Seller Investigators, the FBA reimbursement tool that files claims for you and takes 15-25% of whatever it recovers.

SPS kept the 1P side. That is ChargeGuard, which helps vendors selling wholesale to Amazon fight invoice deductions. It also kept SupplyPike, the Walmart deduction tool it bought in 2024.

The buyer of the 3P business? SPS did not name them. Right now nobody knows who owns the seller-facing reimbursement business.

CEO Chad Collins framed it as focus. His line was that dropping the 3P portion lets SPS zero in on 1P suppliers who trade with multiple retailers. Translation: the Amazon seller tools were never the real prize. The wholesale supplier network is.

The rest of the old Carbon6 stack (SoStocked for inventory, PixelMe for external traffic, ManageByStats, PPC Entourage, ScanUnlimited) was not spelled out in the announcement.

SPS only talked about the "3P Revenue Recovery business." So the status of those other tools is a question mark. If your operation runs on any of them, watch your inbox. Do not wait for a surprise email.

This is the aggregator model breaking in real time.

Carbon6 raised $66 million, then snapped up more than a dozen software companies to build an all-in-one suite for Amazon sellers. It was the Thrasio playbook applied to tools instead of brands. Buy everything, bolt it together, sell the bundle.

SPS bought the whole thing in early 2025. Now it is unwinding the seller side and eating the loss.

And here is the kicker. SPS itself is on the block. In late June, Reuters reported the company hired Morgan Stanley to explore a sale of the entire business. Activist investors Anson Funds and Irenic Capital have been pushing for it. The stock is down roughly 60% over the past year. Revenue growth slowed from 18% in 2025 to a guided 6% to 7% in 2026.

So the seller tools some of you depend on now sit inside a chain of owners who are all selling.

Nobody is coming to save your tool stack. Own your leverage.

Keep a running list of every third-party tool your business relies on. Know who owns each one today. Know how to export your data. Have a backup for anything critical, especially reimbursements and inventory.

The lesson from Carbon6 is simple. Software companies get bought, bundled, and dumped. Your business does not run on their roadmap. It runs on yours.

✌🏼 Have a great weekend.

See you again on Monday.

The answer to today’s STUMP BEZOS is
Adding prompts to Sponsored Brands jumps conversion by 6%