STUMP BEZOS
Total U.S. retail sales for the second quarter of 2026 were estimated at $1,986.5 billion. What were total e-commerce sales for Q2?
[ Answer at bottom of email ]
👀 CHECK YOUR PRICE HISTORY BEFORE SATURDAY
BDSS Dream 100 member Liran Hirschkorn shared a warning this week that has a deadline attached. Black Friday is November 27. The 90-day window Amazon uses to judge your holiday deals opens Saturday, August 29. What you charge in September and October decides what your Black Friday discount looks like.
Liran's cautionary tale: a brand raised prices across its whole catalog. Three weeks looked great. Week four gave it all back. Conversion fell from 11.6% to 8.6% with zero change to ads. Organic rank slid a little more each week. Ad cost went from 22% to 28% of ad sales.
That's the customer side. Amazon's systems punish a bad price increase three more ways.
1. It shrinks your Black Friday discount.
The crossed-out "before" price on your deal is Amazon's Typical Price: the median price shoppers actually paid over the last 90 days. Coupons, Subscribe and Save, and Prime Day sales don't count.
Raise slowly and you sell more units at the old price, which drags the median down and leaves less room to show a discount in November. Best Deals need 15% off. Lightning Deals need 20%. Plus an unpublished minimum Amazon layers on top. A sagging Typical Price can knock you out of qualifying entirely.
2. It caps your deal price.
Amazon runs a second lookback at 30 and 60 days to set the highest price you're allowed to run a BF/CM deal at. That window reaches into late September.
3. It can kill your offer.
Jump too far above what the product recently sold for, on Amazon or anywhere else, and Amazon can pull the Featured Offer or deactivate the listing. A product up 30% in three weeks is competing against its own history. Traffic stops, and the conversion data you were waiting on never arrives.
Do this now:
Pull the last 90 days of price history on every ASIN you raised this year. Look at the units-weighted median, not the current price.
Check Pricing Health in Seller Central before your next increase.
Pick 2 or 3 products to test. Not the catalog.
Move in small steps, one step every 14 days. You have 49 days until October 15. That's three tests. Two if you wait until after Labor Day.
Hold ad spend steady while you test. If price and traffic move the same week, you can't tell which did what.
Check conversion weekly, not at the end of each step. Conversion moves before profit does.
Judge on profit per unit, not revenue.
Write down the profit per unit you need after October 15. Peak fulfillment adds about 32 cents per unit on top of the 3.5% fuel surcharge, plus higher storage. An increase that works in September has to be bigger to work in November.
Submit holiday deals by September 5 for the lower fee. Hard cutoff is October 20.
How you get to your new price matters more than the price itself. Amazon is grading the path, not the destination.
A few months ago my Chris Rawlings hosted a live workshop for Amazon sellers who want to replace themselves with Claude workflows.
It was so popular BDSN subscribers are still asking me how to get access after it’s been over for weeks.
Now he’s doing it again starting this Monday.
The AI Amazon PPC Challenge is coming up and early bird tickets just went live.
You know you need to learn how to use Claude in your Amazon brand and he will show you the specific skills to run and how to connect Claude directly to your Amazon data.
Click here to join the AI Amazon PPC Challenge while seats are available.
🔭 YOU GOTTA SEE THIS
In this episode of Marketing Misfits, Norm Farrar and Kevin King chat with Henrik Johansson, co-founder and CEO of Gembah and former founder of the $100M+ promotional products company Boundless.
Henrik breaks down why simple white-labeling is dying and why custom, differentiated product development is essential to building a defensible e-commerce brand today.
He shares the brutal realities of trying to source in the US, Mexico, Vietnam, and India, and explains how AI tools like Alibaba's Accio and Gembah's "Scout" are disrupting initial factory discovery.
Plus, discover why on-the-ground factory inspections, Triple-N agreements, and pre-launch customer validation are critical to preventing costly manufacturing nightmares.
Join the Marketing Misfits Newsletter
🌎 INTERESTING STATS
🚰 CLAUDE IS WATERMARKING YOUR LISTING COPY
Anthropic now embeds a statistical watermark in text generated by every Claude model launched on or after that date, with older models being added over the coming months.
No opt-out. Applies worldwide, across the app, the API, Claude Code, and Cowork. If a current Claude model touched the text, it's marked.
The trigger is the EU AI Act's Article 50, which requires AI output to be identifiable as machine-made. Google, Meta, Microsoft, and OpenAI signed the same code of practice, so expect this across every major model, not just Claude.
How it works. This is not hidden characters. Anthropic is explicit: nothing is added to the text. When the model has several equally good word choices, it normally picks at random.
The watermark (Google DeepMind's SynthID-Text method) replaces that randomness with a keyed pattern, so the path through hundreds of tiny word choices becomes the signature.
The mark is the text itself. A hex editor finds nothing. Find and replace can't strip it. Light editing won't remove it. A full rewrite will, and heavy paraphrasing or translation can too.
Graphic Courtesy Jo Lambadjeiva
Why it matters for Amazon sellers, per AI expert Jo Lambadjeiva:
Short, factual copy carries almost no signal. A 75-character title or a spec-heavy bullet has no room for the word choices the mark lives in. Anthropic says detection is weak on small samples and factual passages.
Long, expressive copy carries the most. Your 600-word brand story, your A+ content, your buying guides, your Klaviyo flows. Exactly the stuff most sellers bulk-generate.
It marks involvement, not authorship. A page you wrote yourself and asked Claude to proofread gets the same mark as one generated from scratch. A detection hit proves Claude touched it, nothing more.
The watermark travels. Paste it into your Amazon listing, Shopify theme, or email tool, and the provenance signal goes with it. Platforms that decide what gets seen could, in principle, read how it was made.
Nobody can check this today. Anthropic holds the key. Its detection API hasn't shipped. No marketplace has stated a policy. Anyone telling you Amazon can now detect your AI copy is making it up.
What to do:
Audit your catalog now. Know which copy was model-drafted and which was human-written or heavily reworked. Do it while it's boring, not urgent.
Decide where the model sits in your workflow. Brainstorm and outline with AI, write the final page yourself, and you carry a very different signal than publishing the raw draft.
Keep hero copy and flagship pages closest to human hands. Let AI run wild on internal briefs, research, and first drafts.
Ask your agency the question. "Fully human" creative claims now sit against a verifiable signal. AI-disclosure clauses in contracts are about to get technical teeth.
Don't panic and don't quit AI. Your archive is being signed as you build it. You're not choosing whether your copy carries a signature, only whether you know.
🛠️ BDSN SOFTWARE TOOL of the DAY 🛠️
Here’s a tool worth a look this weekend. Alexa doesn't return ten results like Google. It picks one product, recommends it, and moves on.
Buyers who shop through Alexa convert 2.74x more often than browsers. So if Alexa is choosing your competitor, you lose your highest-intent customers and never see it happen.
Before Alexa recommends anything, it runs four checks: who the customer is, what problem they want solved, who the competitors are, and which product solves it best. The listing that answers those questions best wins. The rest are invisible.
This tool reads your listing the way Alexa does. Paste one ASIN. A 24-page strategy deck lands in your inbox. The model behind it is trained on 100,000 top Amazon listings, and it scores how well your listing answers Alexa's four questions versus the products next to you.
🏪 AMAZON WANTS YOU TO BID FOR WAREHOUSE SPACE
Amazon just opened its Sub Same Day network to 3P sellers on a pay-to-play basis. This is the network of small fulfillment centers parked inside dense metros that delivers in 2 to 5 hours.
Until now, Amazon picked which of your ASINs went in based on its own demand and supply signals. That still happens for free. The new part is you can nominate additional ASINs yourself, and you bid a per-unit price to get them in.
The pitch: products in Sub Same Day sell 12% more on average versus standard FBA.
How it works. You set a max price per unit. You only pay for units that actually ship through Sub Same Day. Never more than your bid. Space is limited because these buildings sit in expensive urban real estate, so higher bids win the slots.
Eligible ASINs are in Capacity Manager under "Download ASINs eligible for Sub Same Day."
What should you bid?
Start with the math. If a unit nets you $10 in profit and Sub Same Day lifts sales 12%, the incremental value is not $1.20 per unit. You pay the fee on every unit that ships through the network, including the 88% that would have sold anyway.
So the fee is spread across all units, but the upside only comes from the extra 12%. Rough break-even: bid ceiling = (0.12 × margin per unit) ÷ 1.12. On a $10 margin, that's about $1.07 per unit. Bid above that and the placement loses money even if Amazon's 12% claim holds.
Sellers can bid for extra space up to 20% of their initial allocation or 2,000 cubic feet, whichever is greater.
Three things push the number around. Amazon's 12% is an average across products it already chose for the network. Your self-nominated ASINs are the ones Amazon didn't pick, so expect less lift, not more.
Second, this is a first-price auction with a limited-supply twist. If nobody else bids on your category in your metros, you win at the floor. Bid low and see.
Third, the value isn't only the 12%. Faster delivery badges can lift conversion, Buy Box share, and ad efficiency in ways that don't show up in Amazon's headline number.
Treat this like a PPC bid, not a commitment. Start at 30 to 40% of your break-even ceiling, pick your highest-margin, highest-velocity ASINs, and let the data tell you whether to go up.
🔥 MORE HOT PICKS 🔥
🥃 PARTING SHOT
“The measure of a life is not what you consume, but what you contribute.
Spend less time owning, consuming, taking. Spend more time giving, creating, making."
✌🏼 Have a great weekend.
See you again on Monday.
The answer to today’s STUMP BEZOS is
Total US e-commerce sales for Q2 were $340.2 billion
