Shopify’s merchants are selling more, while the shop-software provider is processing a larger share of their payments. That makes the services around a sale worth as much attention as the sales headline. Primark’s delivery plan poses another question: will extra convenience bring profitable new customers? Amazon, meanwhile, is changing the count behind advertisers’ performance ratios.
News
1️⃣ Shopify grows by handling more merchant payments.
2️⃣ Primark advances British home delivery to late 2027.
3️⃣ Amazon will count Store ads when they reach the screen.
Insights
4️⃣ Anthropic’s growth spreads computing costs, without proving net profit.
5️⃣ Two retailers could capture 81% of US off-site ad spending in 2027.
6️⃣ AI shopping fees: current US terms versus a 2031 scenario.
LET’S ENTER THE MAZE!
1️⃣ News
The Maze: Shopify merchants sold US$119.1 billion through the platform in July–September, up 29% from a year earlier. Shopify supplies software for online and physical shops. Its own quarterly revenue rose 32% to US$3.74 billion. These are global platform totals, not the growth rate of a typical merchant.
Source: Shopify. The company compares merchant order value, its own revenue and earnings with the same quarter a year earlier; amounts are in US dollars, millions.
The Details:
Merchant sales are not Shopify’s revenue. The US$119.1 billion measures order value after refunds, including shipping and taxes. Merchants receive the sales proceeds; Shopify earns subscriptions and fees for services around those sales. Merchant solutions, including payments, generated US$2.91 billion, up 35%, while subscription revenue rose 20% to US$836 million.
Payments capture more of the activity. Shopify Payments, its payment-processing service, handled US$81.4 billion of orders, up 35%. Its share of total merchant sales rose from 65% to 68%. Higher sales and greater payment-service adoption together added US$619 million to Shopify’s payments revenue. More use of a service can lift revenue alongside growth in the underlying shops.
AI’s contribution remains unmeasured. Shopify says orders originating from AI searches grew 3.3 times year over year, but describes the volume as small and provides no absolute count or share of sales. That cannot explain how much AI drove the quarter. For Q4, Shopify forecasts low-thirties revenue growth and operating expenses equal to 29%–30% of revenue. These are expectations, not results.
MarketMaze Take: Growing payment adoption gives Shopify a second way to expand alongside merchants: process a larger share of transactions as their sales rise. This makes checkout services increasingly important to its commercial position. Merchants comparing platforms should therefore examine the services and costs attached to each sale alongside the software subscription. The benefit to Shopify is visible in these results; a benefit to an individual merchant’s profit is not established. The next test is how much extra revenue Shopify retains after processing, infrastructure and operating costs, rather than how quickly an undisclosed pool of AI orders grows.
2️⃣ News
The Maze: Primark, the low-price clothing retailer, plans to start home delivery in Great Britain by December 2027. Its owner, Associated British Foods, has brought the timetable forward by about two years. England, Scotland and Wales are the announced market; this is not a live service yet.
Source: Primark. The collection counter pictured in Primark’s May 2025 rollout announcement. Customers pick up online orders and can pay for extra in-store purchases here.
The Details:
The store trip becomes optional. Primark already offers Click & Collect: customers order online, then pick up purchases in a shop. Home delivery would remove that journey. ABF chief executive George Weston says company research identifies shoppers who rarely visit because they live far from a store or prefer buying online. There is no published measure of how many will use delivery.
Sheffield supplies the missing parcel capacity. Primark acquired automation and a warehouse lease from Debenhams Group. It expects to take over the operation by mid-2027. Suppliers are preparing stock for individual shopper orders alongside their usual bulk store shipments. The faster launch depends on turning that capacity into a working service, rather than simply offering an online checkout.
Convenience has to cover its costs. ABF says delivery fees, minimum baskets and return policies have changed online retail economics. Charging for delivery or spreading its cost across a larger order could make low-price clothing viable to ship. Primark has not published its own charges, minimum spend or returns tariff. Mainland Europe has no delivery timetable.
MarketMaze Take: The commercial test is whether Primark reaches extra shoppers without making existing purchases more expensive to serve. A customer far from a store could add a sale. A regular shopper switching from collection to a parcel could add shipping costs and remove the chance of extra purchases during the store visit. For value retailers elsewhere, the useful benchmark is profit across the whole order after delivery and returns, plus how much demand is genuinely additional. The warehouse enables the choice; the eventual terms and customer mix will decide its value.
3️⃣ News
The Maze: Amazon plans to change display-ad counting on its shopping pages from January 1, 2027. An ad will count when it appears on screen, instead of when it starts downloading. An advertiser notice reproduced by PPC Land forecasts about 63% fewer recorded impressions on affected placements, on average. That does not mean 63% fewer ads delivered or 63% less spending.
Source: Amazon Ads. Amazon’s DVA+ campaign illustration shows objective selection. It depicts the existing buying workflow, not the January impression-counting change.
The Details:
Lower-page ads face the biggest count change. A downloaded ad can sit below where a shopper scrolls. The notice covers display placements bought through DVA, Amazon’s Display, Video & Audio advertising product, on Store home, product, search and order pages. Campaign effects vary. No country list is supplied; sponsored search ads, video, audio and third-party sites are not identified as covered.
A higher reported price can be arithmetic. Hypothetically, $500 spending across 100,000 impressions produces a $5 cost per thousand impressions, or CPM. Reduce only the count by 63% and reported CPM becomes $13.51. Spending and shopper behavior have not changed. Click-through rate, the share of impressions producing clicks, would also rise if clicks stayed fixed. The notice leaves invoice calculations open.
Better results need separate evidence. Return on ad spend measures attributed sales per advertising dollar; cost per action measures spending per attributed action. A new impression count alone improves neither. Amazon claims better results in tests but supplies no quantified uplift or method. It also gives no minimum visible area or duration for its new screen trigger. A separate industry standard generally calls a desktop display ad viewable when half its pixels are visible for one continuous second. An opportunity to see an ad does not prove someone paid attention.
MarketMaze Take: Amazon controls both these Store ad placements and the count brands use to compare them. A stricter count can improve reporting, but the commercial benefit depends on what advertisers pay for the newly defined unit. Until billing is clear, January’s higher reported CPM or click-through rate could simply reflect a smaller count. Brands should compare spending and outcomes by placement before moving budgets. The useful gain would be better exposure for the money; a cleaner denominator alone cannot demonstrate that.
4️⃣ Insight
The Maze: Anthropic, maker of Claude AI, is growing revenue faster than computing costs. Its October 26 filing shows why fuller capacity helps—and why 85% does not mean net profit.
Demand spreads the computing bill. Quarterly revenue rose from $376m in January–March 2025 to $11.6bn in April–June 2026, roughly 31 times. Compute expense rose about six times to $5.6bn. Capacity is commonly bought 12–18 months ahead, so faster sales can spread that committed cost across more usage.
Growth estimates need a label. Revenue grew 144% from the first to second quarter of 2026. Philipp Klöckner's estimated floor of about $16.7bn for the third quarter implies at least roughly 44% growth. He multiplies about 5,300 employees by more than $6.6m annual revenue per employee, then subtracts the known revenue from the previous three quarters. Those rounded inputs give an approximate floor, not reported third-quarter revenue in the original filing. A lower bound cannot prove an actual slowdown. Meanwhile, more than 11,900 customers spent at least $100,000 with Anthropic in the year to September 2026, versus more than 950 a year earlier: large-account demand is broadening.
The margin changes with the bill included. The filing's costs imply about 85% of first-half 2026 revenue remains after functional cost of revenue: mainly computing used to serve customers, plus some staff and overhead. It leaves model-development computing and cloud-distribution fees in other expense categories. Klöckner's separate illustrations leave 66% after deducting that cost of revenue plus sales, marketing and partnerships, or 45% after deducting all compute from revenue instead. The 45% case replaces the other deductions; it includes development computing but does not deduct all other expenses. Those different definitions are not net profit.
Why it matters: More big accounts fill capacity bought in advance. The test is whether demand also funds new capacity, cloud distribution and continued model development.
5️⃣ Insight
The Maze: Buying retail media can mean buying ads shown outside a retailer's shopping sites and apps. That wider reach may still send most spending through two retailers.
Amazon and Walmart could capture 81% of US off-site retail media spending in 2027. That is EMARKETER's forecast, not an observed result. It leaves a calculated 19% for all other networks combined; the available evidence does not disclose the two companies' individual shares.
Off-site describes where the ad appears. Brands buy through a retailer's advertising network or buying service, but the ads run elsewhere online. More places to reach shoppers can therefore coexist with fewer dominant buying routes.
Smaller networks must earn a separate budget decision. The combined share suggests a challenge: offering similar reach may be insufficient when brands can expand existing relationships. Distinctive audiences or useful advertising opportunities could give brands a reason to add another network, but the forecast does not measure campaign returns.
Why it matters: Advertising beyond retailers' sites could extend Amazon and Walmart's role in brand budgets. Wider reach does not necessarily mean more buying partners.
6️⃣ Insight
The Maze: Shopify's ChatGPT route adds no selling fee for eligible US sales. Could shopping assistants eventually charge retailers a commission for the customers they send?
Shopify's ChatGPT route sends shoppers to the merchant's own checkout. Its 0% means no extra ChatGPT commission; ordinary payment-processing charges remain. It applies to eligible stores selling to US customers.
US beauty sellers already pay TikTok Shop a 6% referral fee. Amazon charges 8% or 15%: 8% for total prices of $10 or less, 15% above, with a $0.30 minimum. These are transaction percentages, not shares of profit. Different fee bases and separate charges mean they are not a total-cost ranking.
Investment firm Coatue's 2031 scenario assumes agents keep 10% of purchase value: one dollar in ten. That is an illustrative global-excluding-China model, not an announced fee. It shows the potential business incentive: an assistant that influences purchases could charge retailers for the sales it sends them.
Why it matters: If assistants charge for sales, retailers must judge whether they bring additional profitable customers or collect margin on purchases they would already win.
🧰 Resources
Useful stuff, handpicked by the team every day
🤖 AI prompts: 12 AI Prompts for Short-Form Video Scripts or 👉 All here
📖 Glossary: Google Ads, Decoded: 18 Terms You Actually Need to Know or 👉 All here
📅 Events: DTC Live Connect LDN 2026 or 👉 All here
🔓 Get the full resource library with MarketMaze+
🗞️ Quick hits
Everything else you should know
👖 Brands
Authentic completed its Lee acquisition, with regional partners handling product and distribution. The up-to-$1bn deal includes a possible $250m performance payment; new categories remain planned.
💳 Payments
Fiserv doubled its savings target to $1 billion and began talks to sell peripheral businesses. It has cut 3,000 employee and contractor roles, but cannot yet estimate sale proceeds.
China proposes wider trade-payment access for payment firms beyond e-commerce. Providers would still have to identify customers and verify trade; the rules are not final.
📦 Delivery costs
DHL plans a €2.99 fee for initiated second parcel-delivery attempts for German business senders. Retailers object that they may pay without being able to verify who caused the first failure; final contract details remain unclear.
Whatnot shifts extra USPS postage for underdeclared parcel weight to US sellers from November 4. Sellers whose first sale was on or after September 5 are initially exempt.
🛒 Shopping platforms
Reddit is testing ads in search results with selected advertisers, using query context as one matching factor. Countries and commercial results are undisclosed.
Zalando’s Q3 sales rose 5.1% to €3.2bn, while business services grew almost 28%. Its Westwing pilot would add other stores’ products in-app, with AI checkout planned from 2027; countries and fees remain unclear.
THAT’S IT FOR TODAY!
You’re the reason our team spends hundreds of hours every week researching and writing this email.
What do you think of this issue?
See you next time in the maze!
MarketMaze team
