Shoptalk Fall 2026:
Key Themes & Strategic Insights
Three days, one recurring pressure point: the distance between a signal and a decision. This report captures the main themes from Shoptalk Fall 2026.
Overview
We reviewed all 58 content sessions from September 29 to October 1 and selected 18, six from each day. The headlines were about AI agents and discovery. The more important conversation was in planning, fulfillment and store operations, where retailers showed what it takes to act on signals quickly. The pattern was the same in every room: the technology is ready, and the operating model is what holds companies back.
Theme 1: Agility Is Measured in Response Time, Not Forecast Accuracy
Not one supply chain leader in Nashville claimed to have a better forecast. What they described was a faster response. Spin Master cut demand assessment from about a month to a couple of days. Wayfair handled tariffs with real-time reporting, daily cross-functional standups and daily executive communications. One panel defined operational agility as "the speed between the unexpected and the next best action," and argued that every hour between a signal and a response changes the financial outcome.
The leading indicators are getting more creative. Scotts Miracle-Gro watches forward weather patterns and wheelbarrow sales to see the lawn season coming. Spin Master uses staggered film releases to learn which SKUs sell before later markets launch. But the digital signal is only half of it. As Scotts put it, "the physical side still has some lag."
- Scotts Miracle-Gro: demand sensing and SKU rationalization reduced average carrying inventory by 52% over three years. Scotts now matches its demand plan with real-time commodity prices to choose formulations that protect margin.
- Scenario planning: merchants are running Monte Carlo simulations, 10,000 scenarios at a time, to stress-test tariff levels and shipping-lane closures instead of relying on a single forecast.
- FedEx: customs classification, duties and paperwork are handled before a shipment is created, which they call "shifting compliance left," so cross-border orders stop stalling at the border.
- Belkin: about 70% of a product's carbon footprint is locked in at the design stage. Upstream decisions shape cost, risk and resilience long before a disruption hits.
Key Takeaways
A forecast can't be perfect, but response time can be measured. Track the hours from signal to action.
Better demand signals are only worth something if production, inventory positioning and suppliers can act on them.
Decide on your non-negotiables before the disruption. The companies that adapted fastest already knew what they wouldn't trade away.
Theme 2: The Data Gap Is Now the Biggest Operational Risk
One retailer believed it was in stock 92% of the time. Interviews with customers leaving its stores put the real number at 75%. That gap between what the system reports and what the shopper experiences came up in nearly every operations session. Most organizations don't measure it.
Automation makes the gap worse. One brand described a Black Friday failure in which missing margin data sent products to a marketplace at badly wrong prices and created thousands of orders it couldn't fulfill. Its conclusion: "Automation doesn't make bad data safer. It makes your mistake happen way faster." The Overhyped! panel made a related point from the other direction: the most valuable AI investment right now is robust product data, not custom agent builds.
- Real-time ground truth: warehouse vision systems now refresh inventory and workflow status every 4.5 seconds. Yesterday's data isn't good enough to decide on.
- JustFoodForDogs: "We didn't build an AI tech stack. We built a tech stack and an enablement stack that AI can understand."
- Variant-level visibility: results that look strong at the style level can hide big differences by color, size and retailer.
- Analysis paralysis: merchants at Claire's and others described having more data than they can act on, and countered it with field observation: "our stores are our data."
Key Takeaways
Measure what the customer experiences, not only what the system says. That gap is where revenue leaks out.
Build data controls before you scale automation. Automation doesn't make bad data safer; it spreads the errors faster.
Clean, structured data is unglamorous, but it is the investment that pays off in every channel and with every AI tool.
Theme 3: The Store Is Now the Fulfillment Network
Ulta's 1,500-plus stores fulfill more than half of its ecommerce orders. A year after Best Buy opened a store in Bozeman, ship-to-home demand in that market was more than 20 times higher. Amazon said its fastest delivery speeds are also its lowest cost to serve, because inventory sits closer to the customer. Store, website and fulfillment network are increasingly one system.
The architecture behind that system came up too. Vitamin Shoppe moved off its legacy monolith step by step and then switched its point-of-sale system to the same composable architecture, so one cart can combine items shipped from a distribution center, shipped from a store and picked up the same day. Salesforce urged retailers to decouple checkout and fulfillment from their own storefronts so a sale can happen wherever the shopper discovers the product.
- Kohl's: replaced its Chief Digital Officer with a Chief Customer Officer, because "customers don't think in channels."
- Best Buy: is adding "did the customer achieve the intent of their visit?" alongside conversion, since many visits are for service, returns, fulfillment or inspiration.
- Rhone: would open four stores at once rather than one. A single store doesn't generate enough to learn from.
- Walmart+, Lowe's and The UPS Store: "Loyalty has evolved from just a marketing tactic to truly an operational discipline."
Key Takeaways
Measure stores by the total market they create, including the digital demand they generate.
Unified commerce means shared services across channels, adopted in stages. It isn't a single platform replacement.
If incentives and P&Ls are split by channel, they'll work against the growth the store network is creating.
Theme 4: Fulfillment Speed Is a Design Choice, Not a Budget Line
Wamsutta tested offering multiple shipping speeds at checkout, and conversion didn't change. That one result sums up the session: speed matters only where the customer needs it. The fix for a poor delivery experience is usually the network design, not a bigger logistics budget. As one panelist put it, "that does not mean we need to be able to ship a comforter next day to everyone."
The panel also reversed the usual order of automation. Before automating a process, give the people running it room to solve problems by hand, and use what they learn to decide what's worth automating. On fast delivery, the operators agreed that reliability, not speed, is what separates the best: accurate picking, precise in-store directions and simple returns.
- Amazon Prime: every benefit has to meet a real need, scale operationally and be financially sustainable, and every unit shipped has to add to contribution profit.
- DoorDash: when orders are picked inside large stores, guiding the picker to the exact aisle and shelf improves both speed and accuracy.
- Set promises jointly: one retailer reviews delivery times with marketing every quarter and advertises only what operations actually delivers.
- Know the deadline: a delivery that's technically fast but arrives after the event the customer bought it for is still a failure.
Key Takeaways
Test whether customers value speed in a category before paying for it.
Decide what problem the network is solving (speed, reach, acquisition or opening hours) and design backwards from that.
Automate a process only after it has proven itself by hand.
Theme 5: AI Transformation Is a People-and-Process Problem
The change-management session put a number on it: 10% of AI transformation value comes from models, 20% from technology and data, and 70% from people and processes. Retailers seeing results build feedback loops into the work, push decisions down within clear guardrails, and shift leadership from being right to being willing to test. One team used that approach to cut an 18-month product launch to five months across 10 countries.
The most effective deployments start by learning from the people who already do the work. Artisan Lane's routing AI watched a veteran dispatcher for six months before recommending a single route. Lowe's designs its agents to "infer, confirm, remember": anticipate the need, but never override the customer's choice.
- Sequence the work: JustFoodForDogs prepared operations, data and the organization before touching the customer experience, then replaced its stack in about five and a half months.
- Ownership: a standalone chief AI officer can slow adoption by signaling that AI is someone else's job. Leaders in every function have to own it.
- Governance: Vitamin Shoppe has its security team review every AI use case and builds AI questions into every RFP.
- Measure the whole interaction: one retailer cut handoffs from chatbot to live agent so far that customer satisfaction fell.
Key Takeaways
Scope and staff AI programs as operating-model change. The 70% is where they succeed or fail.
Start with a ranked list of business problems, not with the technology.
Keep a human at both ends of every workflow, and increase autonomy only as performance is proven.
Closing Perspective
Across three days, the same point surfaced in different rooms: most of the progress, and most of the problems, were in operations. Demand signals are arriving faster than supply chains can respond to them. Inventory systems report numbers that customers don't find on the shelf. Stores have quietly become fulfillment networks while P&Ls and incentives are still split by channel. Retailers are paying for faster delivery that customers never asked for. And AI programs are being run as technology projects when most of their value depends on people and process.
The retailers furthest ahead aren't the ones with the most pilots or the newest tools. They know what customers actually experience, they decide their non-negotiables before a disruption hits, and they treat response time as something to measure and improve.
The common thread: technology is no longer what holds retailers back. The constraint is how quickly the operating model can turn a signal into a decision and then into action.
Incisiv at Shoptalk Fall
At Shoptalk Fall, some of the most useful conversations happened after the sessions ended. Incisiv hosted two evenings in Nashville to give retail and technology leaders room to keep talking.
Incisiv Community Happy Hour
On Tuesday, September 29, we brought together our community of retail tech thought leaders for an evening of good drinks and candid conversation. People compared notes on what they were hearing on stage and what it means for the year ahead.
The Speakeasy @ Shoptalk Fall
On Wednesday, September 30, The Speakeasy brought retail leaders together at Nelson's Green Brier Distillery for an evening of food, spirits and candid conversation, in one of Nashville's most historic settings. Many of the themes in this report, from response time to the data gap, came up again over drinks. Thank you to Manhattan Associates and Talkdesk for making the evening possible.
Session Highlights
Day 1 · Tuesday, September 29
Retooling Your Ecommerce Tech Stack (JustFoodForDogs)
JustFoodForDogs replaced its stack in about five and a half months. It got there by going slower at the start: it aligned operations, data and the organization before changing anything customers would see. It didn't assemble an AI stack. It built a clean, structured foundation that AI can read, and bought proven tools rather than building its own. The panel's view was that enterprise AI is moving out of the experimentation phase and into a phase where every investment has to show what it contributed.
Why it matters: Most re-platforming projects are slowed by the organization, not the software. Sequencing is the strategy.
Track Keynote: Beyond Ecommerce: How Digital Services Redefine Customer Value (Amazon)
Prime's fastest delivery speeds are also its lowest cost to serve, because inventory sits closer to the customer and spends less time in transit. Amazon tests every new benefit against three filters: it has to meet a real need, scale operationally and be financially sustainable. Prime isn't designed to profit from members who don't use it.
Why it matters: Where inventory sits drives both cost and customer experience. That makes network design a growth decision as well as an operations one.
Case Studies: Adapting to Fast-Changing Market Conditions (Wayfair, Amazon, AG Jeans)
When tariffs hit, Wayfair set up three things: real-time reporting, daily cross-functional standups and daily executive communications. Underneath them sat a clear list of non-negotiables. AG Jeans found that static customer segments were holding it back. Personalizing search and merchandising to what each shopper was trying to do lifted conversion 16% in three months.
Why it matters: Resilience comes from operating rhythm, not from a better forecast. The companies that adapted fastest had agreed on what they wouldn't trade away before the disruption arrived.
Insights: New Research on Shopper Needs and Expectations (Salesforce, ChangeUp)
Shoppers increasingly start with AI, and Salesforce expects agents to drive about 20% of holiday ecommerce traffic. At the same time, annual loyalty-member churn rose from 13% to 23%. The research recommended decoupling checkout and fulfillment so a sale can happen wherever the customer discovers the product.
Why it matters: Commerce is moving off the retailer's own site. If order management and fulfillment are tied to a single storefront, that becomes a constraint.
Measuring and Improving Company Sustainability (Belkin)
About 70% of a product's carbon footprint is set at the design stage. Supply-chain (Scope 3) emissions are roughly 11 times a company's own operational emissions. Belkin made its packaging plastic-free at roughly no extra cost. A recycled-plastic program, started because a customer asked for it, won about $8 million in new business, and it now saves money as virgin plastic prices rise.
Why it matters: Sustainability makes a business case when it is treated as a design and sourcing decision. Treated as a reporting exercise, it doesn't.
Track Keynote: Creating Shopping Experiences for Today's Consumer (Ulta Beauty)
Ulta's 1,500-plus stores fulfill more than half of its ecommerce orders through ship-from-store, buy online and pick up in store, and same-day delivery. Data from 47 million loyalty members decides where new stores open and what format they take. As Ulta put it: "Stores are a growth engine. They are not a legacy channel."
Why it matters: The store network is now Ulta's fulfillment network. Retailers still running stores and ecommerce on separate P&Ls are measuring the wrong thing.
Day 2 · Wednesday, September 30
Tech-Enhanced Merchandising (Artisan Lane and others)
One retailer reported being in stock 92% of the time. Interviews with customers leaving its stores put the real figure at 75%. Merchants are now running 10,000-scenario Monte Carlo simulations to test exposure to tariffs and shipping-lane closures. Artisan Lane's routing AI watched an experienced dispatcher for six months before it recommended a single route.
Why it matters: The gap between what the system reports and what the shopper finds is where revenue leaks out. Planning with real-time inventory data is a basic requirement, not an upgrade.
A Change Management Toolkit: Managing Successfully in Fast-Evolving Situations
The session split the value of AI transformation as 10% from models, 20% from technology and data, and 70% from people and processes. Building feedback loops into the work cut one product launch from 18 months to 5, across 10 countries. The session also called for a shift in leadership mindset, from being right to being willing to test.
Why it matters: Most AI programs are scoped and staffed as technology projects, but the 70% is where they succeed or fail.
Tech Spotlight: Solutions Enabling Operational Agility (FedEx and others)
"Operational agility is the speed between the unexpected and the next best action." FedEx's approach to cross-border shipping is to settle customs classification and paperwork before a shipment is created. Warehouse vision systems now update inventory every 4.5 seconds. The vendors argued that a deployment should prove its value in weeks, not quarters.
Why it matters: Every hour between a signal and the response changes SKU-level margin. Agility should be measured as response time.
Brand Strategies for Leveraging Marketplaces (Sally Beauty)
About 75% of Sally Beauty's sales through delivery marketplaces are incremental. The panel advised looking at marketplace performance by variant, not just by style, because results differ by retailer. One brand shared a Black Friday failure: missing margin data sent products to a marketplace at badly wrong prices and created thousands of orders it couldn't fulfill.
Why it matters: "Automation doesn't make bad data safer. It makes your mistake happen way faster." Data controls have to come before automation is scaled.
Case Studies: How AI Agents Elevate the Customer Journey (Lowe's, Estée Lauder, Crocs)
Lowe's design principle for its agents is "infer, confirm, remember": anticipate the need, but never override the customer's choice. Estée Lauder groups its agents into three layers: ecosystem agents that influence discovery, consumer-facing agents on brand sites, and foundation agents that run catalogs and inventory behind the scenes. One retailer cut handoffs to live agents too far, and customer satisfaction fell.
Why it matters: Agent programs work when they are designed around the whole interaction. Optimizing one metric, like chatbot containment, can make the experience worse.
Overhyped! Blunt Takes on Hot (or Not) Technologies
Shoppers are barely buying through agentic storefronts yet. The panel's advice was to invest in robust product feeds, which pay off now in Google and Meta and will pay off later with agents. Vibe coding gets an idea to a prototype quickly, but it doesn't replace the engineering work needed to deploy and maintain it. A standalone chief AI officer role can slow adoption by signaling that AI is someone else's job.
Why it matters: The most useful AI investment right now is clean data. It also happens to be the investment most likely to be skipped.
Day 3 · Thursday, October 1
Demand Sensing for Agile and Resilient Supply Chains (Spin Master, Scotts Miracle-Gro)
Spin Master cut demand assessment from about a month to a couple of days. Scotts combined demand sensing with SKU rationalization to reduce average carrying inventory by 52%. It now matches its demand plan with real-time commodity prices to choose the formulation that protects margin. Weather forecasts and wheelbarrow sales serve as early demand signals.
Why it matters: The digital signals have improved faster than the physical supply chain's ability to act on them. Flexible production and inventory positioned in advance are what close that gap.
Fulfillment and Delivery: Exceeding Customer Expectations (DoorDash, Wamsutta)
When Wamsutta tested offering multiple shipping speeds at checkout, conversion didn't change. The panel's argument was that speed matters only where a customer actually needs it, and that fixing the experience is a network-design question, not a budget question. Automation should be the last step, after a high-touch process has shown where the real problems are.
Why it matters: Spending on faster delivery without testing whether customers value it is one of the most common ways retailers overspend on logistics.
The Store–Ecommerce Growth Connection (Best Buy, Rhone, Kohl's)
A year after Best Buy opened a store in Bozeman, ship-to-home demand in that market was more than 20 times higher. Kohl's replaced its Chief Digital Officer with a Chief Customer Officer, because "customers don't think in channels." Rhone said that if it were starting over, it would open four stores at once rather than one.
Why it matters: A store's value shows up across the whole market. If incentives reward channels separately, they'll hold back the growth the store is creating.
Driving Loyalty Via Omnichannel Convenience (Walmart+, Lowe's, The UPS Store)
"Loyalty has evolved from just a marketing tactic to truly an operational discipline." The panel urged retailers to fix the operational foundation before adding technology, to watch early proxy metrics that predict retention, and to pair every KPI with countermetrics that expose trade-offs.
Why it matters: Convenience is something you build into operations. A loyalty program can't deliver it on top of inconsistent fulfillment.
Ask Me Anything (Vitamin Shoppe)
Vitamin Shoppe moved off its legacy monolith step by step, then switched its point-of-sale system to the same modular, composable architecture. Today one cart can mix items shipped from a distribution center, shipped from a store and picked up the same day. AI use cases are reviewed by its security team, and AI questions are built into every RFP.
Why it matters: This is unified commerce in practice: shared services across channels, adopted in stages, with governance built in from the start.
Holistic Merchandising: Leveraging Org-Wide Data Streams for Better Planning (Claire's and others)
Merchants described data overload ("analysis paralysis") and countered it with field observation: "our stores are our data." Automated pricing grids still need weekly human refinement. The consensus was that AI speeds up analysis, but the merchant remains accountable for the assortment.
Why it matters: The constraint is no longer data. It is how quickly an organization can turn data into a decision it will commit to.