Welcome back to the Retention Edge.

As always, we’ve got the latest ecommerce, retail, marketing and retention news and content from the past week. Everything you need to know to stay at the top of your game.

Here’s what’s up in this edition:

  • Shopify lets merchants pay AI shopping agents extra commission, setting the groundwork for monetization in agentic shopping.

  • Uber counted four rides booked by an AI agent out of about 320 million trips and orders in Halloween week.

  • AI-referred visits now bring 83% more revenue per visit, Adobe says, but nearly half of shoppers who bought on an AI recommendation returned something.

  • Shopify's Q3 missed on GMV, and Sidekick's running costs squeezed its subscription margins.

  • DoorDash agrees to buy Instacart for about $17 billion and fold both memberships into one.

Now let's get into it.

AI & Agentic Shopping

Shopify lets merchants pay AI agents extra to sell their products

Shopify is apparently building the framework for paid discovery through AI agents.

Shopify already pays AI shopping agent developers 0.3% on purchases they drive through Promoted Placements, which lets those agents show sponsored products in their Shopify Catalog searches. Now merchants can add their own commission on top through Shop Campaigns. Shopify's example combined rate is 1.5%, credited to the last non-direct click within seven days.

There are limits for now. Promoted Placements is still an invite-only developer preview, and the higher rate only applies in the US when the buyer checks out in the Shop app or on shop.com.

So there’s nothing really happening now - but it’s an interesting sign of what might be coming in the future. Two things actually.

One - Shopify becoming the shared catalog for the internet. It feels like it’s becoming the new Amazon. You want to buy something, you’re going to be able to find it from someone who sells on Shopify.

Two - the monetization of agentic commerce (from the agents/developers side). Are we going to see every purchase or recommendation through an AI agent have a financial stake? It won’t surprise me. Almost every channel eventually becomes pay to play. It’s hard to imagine AI channels will remain completely impartial and stake-free.

Amazon opens Shop Direct to any eligible retailer, and its agent can buy from your site

Amazon launched the Shop Direct Merchant Workspace, a self-serve portal where eligible retailers submit product feeds and connect their stores. Those products can then show up in Amazon search and in Alexa for Shopping, which replaces the Rufus assistant.

The portal takes standard shopping feed formats and connects through Feedonomics, Salsify, CEDCommerce and VersaFeed.

Shoppers can tap through to your store, or use Buy for Me, where Amazon's AI completes the purchase on your website with their saved address and card. You keep inventory, pricing, fulfillment and customer service. You pick which products to include, there are no fees, and you can opt out at any time.

Google takes checkout inside Gemini to Canada

Canadian shoppers can now buy products they find across Google Search through a conversation with Gemini, paying with Google Pay and the details saved in Google Wallet. The retailer stays the seller of record and handles fulfillment. US shoppers got the feature on Oct. 15.

The launch partners are Lululemon, Shoppers Drug Mart, Simons and select Shopify merchants, with more expected in the new year. Shopify helped build Google's in-chat checkout, even as it rolls out competing AI shopping tools of its own. Google says access is free for retailers.

Almost no one is letting agents shop for them

Uber handled around 320 million trips and orders across Uber and Uber Eats in Halloween week, its busiest week in October. AI agents placed 500 of the Uber Eats orders and booked four rides.

"It is very, very early and very limited traction that we are seeing so far," CEO Dara Khosrowshahi said. DoorDash said the week before that it has received only a few hundred orders from AI agents.

Khosrowshahi pointed out that booking a ride in Uber's app takes about 50 seconds, and re-ordering a favorite meal about 90. An agent isn't saving anyone much time on that.

It’s the latest FUD for personal shopping agents. To be fair, it’s very early. But a lot of people are struggling to find a compelling use case for these agents. It’s not just whether or not they save you time - one also needs to consider the risk of agents messing something up. The benefits of using an agent to buy for you need to be strong enough to outweigh the risks as well.

Visa tested 22 AI models on shopping tasks, and they failed 42.8% of the time

This might give us an answer as to why agentic shopping hasn’t taken off yet.

Visa's new Agentic Commerce Benchmark gives AI models realistic shopping jobs, like buying a TV, a laptop or a personal care appliance. To pass, a model has to work out through conversation what the shopper needs, then buy the right product, or decide not to buy if nothing fits.

The first round put 22 models from 12 AI labs through 15 shopping challenges. The best model passed 82.2% of runs and the worst 28.9%.

Each model got three tries at every scenario, and all three passed only 34.2% of the time. In 54% of the failed runs, the AI bought something anyway, either a product that didn't meet the shopper's requirements or one chosen before it had found them all.

Visa plans to extend the test to checkout, payment and order placement. Its Trust Index in September found that 72% of US consumers had used an AI assistant, but only 23% trusted AI to handle payments for them.

The takeaway? It’s going to take some time until AI assistants can reliably do the job asked of them; and then some time still, after that, until everyone feels comfortable handing the keys over to their agent.

AI-referred visits now bring 83% more revenue per visit

In September, AI traffic to US retail sites (people clicking a link from an AI tool) grew 71% year over year, Adobe says. Those visits converted 90% better than non-AI traffic, the highest gap Adobe has seen. In August 2025, they converted 9% worse. AI visits also brought 83% higher revenue per visit.

For the holidays, Adobe forecasts AI traffic up 159% on Thanksgiving, 95% on Black Friday and 88% on Cyber Monday. The catch is visibility. Adobe's Citation Readability Score found the average US retail homepage scored 63%, meaning more than a third of the content isn't visible to AI systems. General merchandise homepages scored just 40%.

More brands are running ChatGPT ads, but the clicks convert at a third of Google's rate

Adthena counted 1,082 brands advertising on ChatGPT across the US, UK, Australia and the EU in June, and 6,333 in August. The mix is widening too. Shopping's share of US ad spend on ChatGPT fell from 38% in April to 20% in September, Sensor Tower says, while financial services grew from 2% to 13%.

Paid ChatGPT traffic still converts poorly. In a working paper under review by Maximilian Kaiser and Christian Schulze, based on data from ecommerce sites, paid ChatGPT visits brought US advertisers $0.89 in revenue per session, against $2.45 for paid Google Search visits, and converted at 0.37% against 1.17%.

Nearly half of shoppers who bought on an AI recommendation sent something back

In a Blue Yonder survey of 2,000 US shoppers, nearly a quarter had bought a product based on an AI recommendation, and 46% of them had returned an item in the past year. The top reason was that it was different than described (36%), ahead of size or fit (23%). For returns in general, size or fit leads at 42%.

"An AI assistant is only as good as the product information it can find, and most product data was written for a search bar, not for a machine suggesting products for someone," said Blue Yonder CEO Duncan Angove. And 51% of respondents said they've picked one retailer over another purely for a better returns policy.

The Shopify Beat

Shopify's Q3: growth, a miss, and the cost of Sidekick

Shopify’s GMV rose 29% to $119.1 billion, short of the roughly $120.1 billion analysts expected. Subscription revenue grew 20% to $836 million, below the $850.2 million forecast.

Total revenue climbed 32% to $3.74 billion, and net income rose to $462 million from $264 million. Shares fell as much as 19% before the market opened and were still down about 7% that afternoon.

CFO Jeff Hoffmeister said spending on Sidekick cut into subscription margins, because the cost of delivering those services rose 37%.

Shopify Protect blocks likely fraud before it reaches the bank

Shopify Protect is a set of fraud and dispute tools that runs automatically, at no extra cost, for every merchant on Shopify Payments. Shopify now declines payments it's nearly sure are fraudulent before they go to the bank, so you never end up with an order to cancel or refund.

Risky orders that get through now come with a plain-language reason, and when a customer files a dispute, Shopify's AI gathers the evidence card networks look for, including carrier delivery photos. Shopify already covers fraud chargebacks on eligible Shop Pay orders and plans to extend that beyond Shop Pay.

Send each sales channel its own product data

A new channel mapping page lets you choose what product, variant and metafield data each channel gets, without touching your storefront.

You could keep a short title on your site and send Google and YouTube a longer one that combines the title, color and key features. Each field shows a coverage percentage, so gaps are easy to spot, and you can still override values per product in bulk.

Shop and Shopify Collective aren't supported, and prices can't be remapped.

Shopify now checks your barcodes against the GTIN registry

When you add a barcode that looks like a GTIN, Shopify looks it up in the GS1 registry and marks it as a GTIN if it's registered. Saved custom barcodes that look like GTINs get a suggestion you can accept or dismiss, one by one or in bulk. Google and other channels use GTINs to match your listings to the right product. Nothing changes unless you accept.

Sidekick gets web search, app data and simulated shopper tests

Sidekick's suggestions can now draw on web search for market and industry context, and on data from installed apps that support it. It can also draft a storefront change and compare it with your current store using AI-simulated shoppers. Shopify is clear that these aren't tested on real shoppers and aren't live A/B tests. You review every change before it goes live.

Marketing and Retention

Klaviyo grows 25% on bigger brands, and opens its data to Claude and ChatGPT

Klaviyo's Q3 revenue rose 25% to $387.2 million, and it raised full-year guidance to $1.537 billion to $1.541 billion. Growth is coming from the top end: 105 customers now generate $500,000 of ARR, up 57%, and 390 generate over $250,000, up 49%. 22% of ARR comes from customers using 3 or more products.

It also announced Klaviyo Headless, which lets customers access their Klaviyo data from any AI system, including Claude and ChatGPT, and launched Klaviyo in Shopify, an embedded app that brings Klaviyo directly into Shopify.

TikTok Shop now credits livestreams with sales on your own website

Brand LIVE Impact, a new Seller Center module for some US brand merchants, credits a livestream with sales up to 14 days after a viewer interacts with a product in the live room. That includes purchases on the brand's own website, as long as a TikTok Pixel is connected.

The number keeps updating for 14 days after the stream ends, and TikTok says to wait for it to settle. Its older Live Diagnosis tool uses a 24-hour window, so the same stream can show two different sales figures. TikTok also warns of duplicate counting when several shops share one Pixel.

DTC Brands

Rhode's European launch drives a 51% sales jump for e.l.f. Beauty

e.l.f. Beauty's net sales rose 51% to $520.8 million in the quarter, with US sales up 25% and international up 165%. Rhode's launch into Sephora in 19 European markets was "the biggest launch in Sephora's history," CEO Tarang Amin said.

DTC is still pulling its weight. A Rhode collaboration with influencer Alexandra Leclerc, with limited-edition colors sold only on Rhode's own site, drew a 200,000-person waitlist, and most products sold out in under an hour.

Cotopaxi swaps early discounts for limited drops

Cotopaxi is releasing limited-edition products every other day from Nov. 1 to Nov. 9, under a campaign it calls "Cotodays," with free shipping on Nov. 11. The drops sell on its site and in its 20 US stores, and include a $175 bundle with a Fujifilm Instax camera. It's teasing them on social, email and SMS.

CEO Lindsay Shumlas said the idea is to keep "sustained consumer interest, product discovery and engagement around newness earlier in the month," before the brand has to compete in the promotional window.

Funding and M&A

DoorDash agrees to buy Instacart for about $17 billion

DoorDash will pay $65 a share, a 26% premium over Instacart's price on Oct. 22, before news of the talks broke. Instacart keeps its own name and app, DashPass and Instacart+ get combined under one fee, and grocers can still use rival delivery platforms. The deal needs antitrust clearance in the US, Canada and Israel, and the companies expect it to close between late 2027 and mid-2028.

The owner of Hydro Flask and Osprey gets a $1.6 billion takeover offer

Empower Brands, a joint venture of Spectrum Brands and Oaktree Capital Management, made an unsolicited, non-binding offer of $37 a share for Helen of Troy, which also owns OXO. That puts Helen of Troy at an implied enterprise value of about $1.6 billion. Empower already owns about 4.5% of the shares.

Also This Week

  • A Spanish florist gets 1 in 5 online orders from ChatGPT - every one arrived tagged utm_source=chatgpt.com, but only about one in eight showed ChatGPT as the referrer. Read more →

  • Home Depot will let AI agents shop, if they prove who they work for - it's talking with Meta about how Muse should identify itself, while Amazon has blocked Muse. Read more →

  • Amazon proposes changes to Google's agent checkout standard - so carts and orders on multi-seller marketplaces show which seller actually sells each item. Read more →

  • Zalando's own AI agent will check out on other shops from 2027 - through Zalando Extended, which adds products from shops Zalando doesn't sell to its app, starting with bulky home goods and Westwing as the first pilot partner. Read more →

  • 1 in 4 shoppers mark holiday promo emails as spam - 33% blame too many emails, and 22% want a snooze or lower-frequency option instead. Read more →

  • Meta says AI ranking upgrades lifted Facebook ad conversions 15.7% - in Q2, with partnership ads with creators now past a $15B run rate. Read more →

  • Faire prepares for a possible IPO - annualized revenue up 49% to $750 million, and its first quarter of EBITDA breakeven. Read more →

  • Amazon now requires $1M of liability cover in safety categories, whatever your sales - including children's products, cosmetics, supplements and lithium batteries. Read more →

  • Etsy's new terms let it train AI on sellers' listings - from Dec. 16, with no clear US opt-out, while it reserves the right to block AI shopping agents. Read more →

  • Whatnot makes US sellers pay USPS weight adjustments - from Nov. 4, with 45 days to dispute a charge. Read more →

  • Reddit plans to require logins to browse - starting with tests outside the US, as its search referrals keep falling. Read more →

The Social Sphere

What DTC's been talking about on social media this week.

Matching a competitor's price can cost you your best customers

Thomas Lalas warned founders about cutting prices to match a competitor: conversions go up, then 90 days later retention, LTV and margin are down while CAC hasn't moved. "Lower the price, lower the commitment," and you start attracting people who were always going to leave.

Don't waste the two moments every customer sees

Lalas also pointed out that the thank-you page and the moment a customer opens the box are the only two points in the funnel with a 100% open rate. His advice: ask one tap-to-answer question on the thank-you page, "What made you buy this today?", so every email and SMS after it can speak to that reason, and put a single QR code or insert in the box.

Plan BFCM email and SMS like you plan paid

Feras Khouri of retention agency New Standard Co estimates that only 5 to 10% of brands will forecast retention revenue, with contingency plans, for this BFCM. "Everyone has a plan for paid but almost nobody has one for email and SMS."

How to segment your abandoned cart campaigns for better results

Max Sturtevant splits checkout abandonment with two questions, whether the shopper has bought before and whether their cart is above your free-shipping threshold, and sends each of the four groups its own email. Everyone hears exactly where they stand on free shipping, new buyers also get the welcome offer, and shoppers below the threshold often add another item, which lifts AOV on top of the recovered sale.

A new way to build ecommerce mobile apps

The Retention Edge newsletter is brought to you by MobiLoud.

MobiLoud lets ecommerce brands launch custom native apps, powered by the same underlying code as their website.

Once upon a time, businesses used to have two separate builds for mobile and desktop web. Then the tech improved, development processes evolved, and now no one maintains desktop and mobile separately.

The same used to be true for mobile apps. You needed a separate codebase, a whole new framework, to have a native app. You could launch an app, but it would be at the cost of maintaining a separate storefront in parallel to your site.

Not anymore. MobiLoud lets you keep one source of truth, shipping to three unique channels at once: desktop, mobile web, and a native app.

You get all the benefits of a custom, fully native app, with a modern UI built exclusively for app users - without the overhead of running a separate storefront.

It’s the engine behind the VIP channels from brands like John Varvatos, Jack & Jones, Tadashi Shoji and many more.

If you want to learn more about how it works, and see what’s possible, get in touch and book a free preview and consultation now.

That’s all for this week.

I’ll be back in touch soon, with an eye on what’s happening in the ecommerce and retail world, so you can keep yours on growing your business.

Until then,

Pietro and The Retention Edge Team

PS: shoot me a DM on LinkedIn if you’re interested in what a custom mobile app could do for your brand.