Kimia Kabiri

For my retail buds from the “AI-personalized everything” days… Shopify is growing ~30% at $11B+… and the stock gets smacked ~28%. To SaaStr's recent post (below), multiple compression is relentless, nothing “wrong.” Just the market reminding everyone that: - The market is watching margin compression. - Multiples (for the high flyers) are unforgiving - Rise of AI into commerce, adding uncertainty to Shopify's future margins and growth as everyone scrambles to show AI-driven conversion, efficiency, and operating leverage. Retail AI used to be about personalization. Now it’s about productivity + profitability. There was an oversell, as with the #SaaSpocalypse, but businesses need to recognize the top-down pressure to deliver the #ExponentialEfficiency expected by the market in 2026. Welcome to the new scoreboard. Read more from SaaStr https://lnkd.in/gQSXBUBx #RetailTech #AI #Ecommerce #OperatingLeverage #BeenThere

2

2 Comments

I built 80% of the core features of Shopify in 30 hours. After 20 years of working in e-commerce, I wanted to ask myself a question: how hard would it be to recreate all of Shopify's important SaaS storefront tech with today's AI? I carved out some hours outside of my normal job over 6 calendar days. Armed with an idea and a blank slate, I began. This is what I found. What emerged at the end of the week genuinely surprised me: a working, multi-tenant e-commerce platform built entirely from scratch. When I say 80%, I mean the real, practical essentials an independent merchant relies on daily. It absolutely doesn't replace the vast, complex global ecosystem Shopify's world-class team has built. But for the core features 90% of standard stores need to launch and scale, the foundation is all there, and the business logic is real. Out of the gate, the platform includes: - A proper engine for taxes and layered discounts - An immutable inventory ledger - A full Order Management System (OMS) - A visual block-based theme editor so anyone can design a storefront without code The process itself was a fascinating shift. Every single line of code—all 41,000 of them—was generated by AI. But it wasn't mindless prompting. It felt much more like a deep, iterative conversation. I supplied the blueprints from two decades of experience; the tool executed the structure. I deliberately chose to build the entire backend in Rust—a language I had never personally used before. Because modern infrastructure tools handle so much of the heavy lifting out of the box now, I was able to stay completely immersed in the actual craft of the product, rather than wrestling with servers. Treating this as a 30-hour Version 1.0 makes me incredibly excited. It isn't a fragile prototype—it's structurally sound, compiles down to single binaries, and runs blisteringly fast. The speed limit of software development has permanently changed, and it opens up beautiful possibilities for small, dedicated teams. Imagine what could be delivered in a single year now. If you're curious to see how it works under the hood, you can try out the live demo or read my full reflection below. 📖 Read the blog post: https://lnkd.in/gjr2ZvcQ 🛠️ Try the Demo Admin: https://lnkd.in/gsFw-nfZ 🛍️ View the Storefront: https://acme.yep.shop/

158

26 Comments

Universal Commerce Protocol At National Retail Federation this year, Shopify announced the Universal Commerce Protocol (UCP), a new open standard co-developed with Google to bring commerce to agents at scale. Here is my summary and what it means for Shopify merchants: 🔶 UCP is an open, extensible protocol that establishes a common language for agents and commerce systems across the whole shopping journey, from discovery to post‑purchase support. 🔷 TCP/IP proved this and UCP applies this pattern to commerce, separating responsibilities into layers. Thoughtfully layered protocols survive and thrive by separating responsibilities, defining clear APIs, and enabling composition. 🔶 Shopify merchants declare and define what capabilities they support, including their own bespoke functionality. Agents discover these capabilities, negotiate what they can handle, and proceed to complete transactions. 🔷 Native shopping on Google is rolling out soon, which means Shopify merchants can sell directly in AI Mode in Google Search and the Gemini app.  🔶 The Shopify Catalog is now open to every brand through their new Agentic plan, expanding the commerce backbone that connects millions of merchants to buyers worldwide. Brands on any platform can now use Shopify’s infrastructure to sell on AI channels without needing to have a Shopify online store. 🔷 UCP was co-developed and endorsed by more than 20 companies across retail and payments, including Shopify, but also Etsy, Wayfair, Target, Walmart, Visa and others. 🔶 Shopify is releasing an update to our Microsoft Copilot integration with a new embedded checkout experience. These new integrations, in addition to ChatGPT, are managed centrally from the Shopify Admin through Agentic Storefronts. Read more about Shopify and the Universal Commerce Protocol here: https://lnkd.in/gRAJkr8U I plan to dive deeper into UCP next week in a further post to ensure Ambaum - Shopify Plus Partner merchants are up to date on what's coming.

10

3 Comments

One surprising trend our results reporting picked up: advertising cookware products on cooking channels are some of the worst performing ad placements in the Canadian market. This is just one out of many counter-intuitive market observations that we've discovered over the last 23 years of advertising experience in the Canadian market. When we dig into the reason and discussed with our advertising partners two reasons surfaced: 1) The consumers watching the cooking shows are aspirational viewers, and they actually don't end up cooking. 2) The second factor is that the other segment of the audience are ones who typically have a fully stocked kitchen and are not on the market for new brands. For this particular market, we typically may have advertisers come in and intuitively they may request advertising on food-based channels. We'll instead steer them towards other channels that are lower cost and high response, such as the W network or Makeful, where more of the audience typically are women and we see better response and conversion rates. There is no Wikipedia or database that covers this. This is just one of the many surprising layers of earned knowledge that we developed helping our clients deploy hundreds of millions worth of ad spend in Linear and CTV in Canada over the last 23 years. If you are a challenger brand looking to expand in the Canadian market, happy to share more of these edges that we’ve discovered with your team.

7

TLDR: the industry has solved outcomes measurement. Good work team. I was at the Fifty5Blue (formerly KantarMedia) Summit in Montreal last week and got to go back to my roots, spending time thinking about things from a research, and not a product, perspective. One of my conclusions: the industry has solved outcomes measurement. What do I mean by solved? As an industry we now know what outcomes are, how they can be measured and who should do that measuring. While there are complexities in implementing those frameworks within an organization, that is about applying context and nuance, not a gap in frameworks or tools. What are outcomes? With the proliferation of digital data, the industry has been able to map a series of proxies, or signals, to revenue. Whether it’s an attitudinal metric, or a shopping funnel metric like branded search, the question is never “is this metric an outcome?”. The question has become “is this the right outcome metric to measure for me?” How do we measure them? MMM, MTA and causal methods. Yes, there are flavors of all of the above, and yes each of them does some things better than others, but importantly we know what those things are. So again it’s not a question of “what methods exist to measure my outcome?”. The question is “what method should I apply for this specific question?”. Lastly, is the question of who? In the twenty-teens the answer appeared to be media companies and platforms. Their argument was “we own the audiences that drives the outcomes, so only we can tell you precisely what’s happening.” There was an explosion of first party measurement solutions from brand and conversions lift to attribution modeling. But in the post-ITT and AI world their message has shifted to “how can we help you measure? What do you need from us to help you triangulate?” And who is the “You”? The marketer. Marketers own outcome measurement, as they should, because even with all the AI at the media and platform companies, marketers will always know their business better. More importantly they know how decisions in their businesses are made, and outcomes measurement has moved from “how do we do it?” to “how do we use it to make decisions?”, something only marketers know. Yes, vendors, media companies and agencies will all play important parts. But ultimately marketers are driving the future of outcomes measurement.

28

8 Comments

If your campaigns are still relying on pixels alone, you’re only scratching the surface of what your data can do. True performance means sending real outcome signals back to ad platforms so bidding algorithms can optimize for quality, not noise. Learn how to unlock better #ROAS and more predictable growth in our latest blog 🔽 https://hubs.li/Q043Gsyb0

6

Thanks IAB Canada for outing together this great starter kit for everyone looking to integrate Agentic AI technology into advertising

4

Prelude to a video post coming tomorrow. In CPG, even the emergence of AI cannot change these immutable realities: - Medium & Heavy brand and category buyers require a lot of attention both for conquest and retention - New to Brand (NTB) buyers are highly coveted but often inaccurately identified due to fragmented retailer signal (bought at Walmart one week, at Kroger the next) - Consumer shopping behavior changes quickly, not only by channel (TikTok Shop) but by form, retailer and brand Context or intent and even demographics are great, but unless you know what brands and categories consumers and households are buying, you’re flying blind. Many years ago, people used to say, you can’t get fired if you buy from IBM. They were the low risk vendor everyone knew. Today it’s Google and Meta. It’s so easy. They’ll even tell you how great you’re doing. But you have to start with seed data and audiences where you know what the consumer is buying. Are they loyal? Do they look like your best customers by the other items in their basket? Are you totally missing a group of people buying in the category or at another retailer but not from you? These are the questions that becausal can answer. Legacy CPG purchase data providers don’t cover Walmart like we do (20% of our file). That’s key because Walmart is still 40% of most CPG clients’ ACV. We’re multi-sourced from receipt scans, retailer POS systems and loyalty card programs. So we have great scale at 20MM MAUs. Skew younger than Circana. Good retailer signal, too. We provide the granularity that is essential for success in CPG. Case in point in the comments, below. Hit me up. #cpgmarketing #purchasedata #retailmedia #cleanrooms

22

8 Comments