Were you forwarded this email? Sign up here to get the next edition.

Well, it's not every week that you get told there's a 10% chance of you dying within the decade. Anyway, here's some more AI ecommerce news. Hope it helps you sell some stuff.

CHEERS, BUSHY

P.S. Last call. I'm hosting a webinar with StudioHawk today where they'll share results from 100 Australian brands they've studied. We've got almost 300 people joining. Save your seat here.

01

Up until this point, ChatGPT ads have essentially been water-gulping search ads. This week's release of sponsored agents, though, starts showing us what the future of conversational advertising might look like. When users get recommended a product or a brand on ChatGPT, they can now choose to start chatting with that brand within the platform. Essentially, the bots will have concession bots within the platform. Maybe it's the new sales model for Myer?

ChatGPT aren't alone with their conversational ads either. Google announced this week that they are embedding business agents directly inside YouTube ads so shoppers can ask product questions without leaving the ad. If the YouTube ads agent can make me a cup of tea and empty my bladder mid-ad, then I'll be impressed.

Both of those are US-only for now, sponsored agents with a handful of selected advertisers and the YouTube agent in beta for eligible US retailers. But they'll come.

As an aside, the ChatGPT Ads app is now on Shopify and syncs your catalogue straight into ChatGPT Ads Manager. Reviews so far: two five stars and four one stars, averaging out at 3. The bots will call that a perfectly above-average result. It is available here in Australia on 23 September.

02

Not gonna lie, it's a Google-heavy week because Google made their pre-holiday season announcements. But they've heard what you well-behaved retailers have put on your Christmas list. They've updated the Universal Commerce Protocol so that when products are added to the Google cart through AI Mode and Gemini, retailers can enable cart transfer and send that cart to their own site to check out. Big news if you want to fold customers into loyalty, subscription or upsell. I know it's one of the biggest issues people have with agentic checkouts. Cart transfer rolls out in the US first, with Australia and Canada flagged for early next year. Maybe the website isn't dead after all.

What you can action straight away is AI performance insights in Merchant Centre, which is live in Australia now. It gives you share of voice against other brands across AI Mode and AI Overviews, split by discovery, evaluation and ready-to-buy intent. Head over to your Merchant Centre account, then Analytics > Products > AI performance at the top of the page, and you'll see how you're showing up.

03

According to SimilarWeb, 11.4% of shopper journeys now use AI and those journeys carry a 23% conversion rate when combined with search. With 89% of AI shoppers saying they use both AI and search in a typical journey, I expect all my conversion figures to be in the double digits from now on. Dreaming.

I've always really liked SimilarWeb's State of Ecommerce reports, however, it's getting harder to give them credit. The journey number covers desktop and mobile worldwide, but the conversion and referral data is US desktop only, which doesn't reflect how customers are behaving. Their view on AI influence still hinges on clicks. They did include some qualitative data worth a look though. AI was seen as twice as useful as search engines for discovery and research. Australia also made the top 10 countries for visits to marketplace sites.

WPP Media's Hidden Pathway research shows how hard it is for companies like SimilarWeb to understand the customer journey now. We're not using our AI tools and clicking through immediately or directly.

WPP put the question to 2,500 Australians in August and found 3 in 4 already use AI search to help make purchase decisions. 40% go straight to LLMs such as ChatGPT, Gemini or Claude. WPP found buyers complete an average of 18 AI actions when making a purchase decision and 94% of that is happening at the start of the journey. SimilarWeb claimed the middle was the most important part for AI. If these research and data companies can't map the AI customer journey, what hope do we have?

04

Earlier this year, Shippit told us that only 11% of retailers now offer free returns, down from 50% in 2018. We now have additional data from Loop on the impact of that on customers.

Loop's new report, called the Returns Revenue Gap, found that 58% of consumers have walked away from a brand over its return policy. (Keep in mind Loop have a very good reason for publishing these stats.) Contrary to common belief, 75% of customers would rather a full refund but 87% will still take an exchange instead if the experience is easy enough.

Australia is handling returns pretty well and (according to Loop) is 12 to 18 months ahead of the US and UK - primarily because we embrace exchange over refund. 53% of Australian retailers offer instant exchange, the highest rate of any market, and we've got the lowest refund share at 42%. If you're sitting on the fence with your return policy leading up to peak, exchange could be the make-or-break difference, especially if you get the sales in peak and the exchange happens at full price.

05

Gahhhh… You thought that was it from Google. This week they also shipped Gemini 3.8 Live, a speech-to-speech model that holds a real-time voice conversation and executes tool calls in the background while it's still talking to you. It's a developer release through the Live API rather than an update to the Gemini app you use.

And while my first thought is how we can enhance our customer service here, I love this Modern Retail article showing the power of real customer service via phone chat as a differentiator.

I've said for a long time that if I were a small-to-medium brand in Australia looking to break through, I would open up phone lines for customers to call. No menus, no bots, just the team on the other end ready to chat. After all, you've presented them with your best content. Given them a great offer. Chased them around the web. Paid for the privilege. And then you won't let them call you?! ARE YOU NUTS.

ATC PLAYBOOK #663

We're all in a race to acquire new customers right now. But how do you measure if that new customer was actually profitable in the long term?

You can probably tell me how many new customers you brought in this month. Far fewer of us can say whether they came back, how long it took, and whether they were worth what we paid for them. It gets harder for the customers you bring in during a sale because a customer who came in at 30% off is not the same as a full-price customer even though they look identical in your reporting.

Today's Playbook is anchored by Rob Ward, Co-founder at Quad Lock, who bootstrapped a novelty bottle opener into a business doing more than $200 million in revenue, then sold it to Thule for $500 million. He calls his framework the three gates.

Examples from this Playbook

🚪 Rob from Quad Lock runs new customers through three gates: profitable on the first purchase using nCAC and new-customer AOV, how long until the cash comes back, and what they're worth across the whole relationship. Not every gate has to be green. Quad Lock failed different ones at different times and still sold for half a billion.

🎯 Mark from Mark Baartse Consulting says the work is figuring out who your one-and-done customers are and who your twenty-purchase customers are. Once you know, upload only the good group as your lookalike seed instead of every purchaser you've ever had.

📊 Carla from Profit Peak is more bullish than Rob on that first gate and wants new customers profitable from day one unless you have real subscription revenue. Blend new and returning together and you'll end up spending your whole budget on people who already know you.

💸 Danny from Kelly+Partners has seen where the other version ends, with an operator carrying a $200,000 debt to Facebook for ads that were never profitable and no path out of it. Payback has to be observed, not hoped for.