Issuu

Lavina Suthenthiran Senior Retail Analyst

State of Grocery Retail 2026

Table Of Contents 04

1. Make everyday pricing beat the promotion

2. Make personalization reciprocal, not extractive

3. Treat membership as the start of the data strategy

4. Fix availability where it costs you most

5. Build workforce flexibility to offset rigid automation

6. Close the last-mile gap without handing away the customer

7. Run every week as if it's peak week

What grocery leaders do differently

Sponsored by:

Introduction: What happens when the most habitual category in retail becomes the most technologically strained?

They're becoming one system: the thing that actually earns the trip, builds trust, and turns into profitable growth.

That same connect-the-pieces test applies strongly to AI, and the industry isn't aligned on how to pass it. Nearly half of grocers expect That's where grocery sits right now. Same AI agents to assist with a third or more of stores, same brands, same routines, year transactions within five years, but consumer after year, and yet the ground underneath all trust hasn't caught up: willingness drops as that habit is shifting faster than the industry's AI's autonomy rises, and only 20% would let it built to handle. place an order with zero review. According to the Braze 2026 Customer Engagement The numbers back it up. U.S. grocery sales Review, the same gap runs straight through grew just 1.2% in 2025, driven entirely by higher the marketing side too, retailers are far more prices as unit volume declined. The slide has confident in their AI-driven personalization since gotten worse since: per analysis of than shoppers actually feel it's working. NielsenIQ data, unit sales were down 1.8% year-over-year in June 2026, a sharp reversal That gap is the narrative throughout this from flat growth just twelve months earlier, theme: loyalty programs that reward even as prices climbed 2 to 3%. Shoppers are transactions but never build a relationship, buying less. personalization retailers call sophisticated And the strain is only getting worse. By 2030, U.S. online grocery demand could outpace fulfillment capacity by $20B to $30B, per McKinsey, even after every expansion plan already on the books. The industry's back in growth, but that’s not enough this time. Shoppers have grown more intentional about value. Margins have gotten less forgiving. The profit pools worth chasing have moved. Retail media, ecommerce, private brands, fresh, loyalty, and AI used to be six separate agendas.

and shoppers call generic, AI investment outpacing AI trust. And underneath it all, lastmile, workforce, availability, operational categories that have quietly stopped behaving the way they did two or three years ago. Grocery is running at a speed it wasn't built for. Data, AI, and consumer expectations are all moving faster than legacy systems and org charts can absorb. These are the seven trends we're capturing in this report: 1. Make everyday pricing beat the promotion 2. Make personalization reciprocal, not extractive 3. Treat membership as the start of the data strategy 4. Fix availability where it costs you most 5. Build workforce flexibility to offset rigid automation 6. Close the last-mile gap without handing away the customer 7. Run every week as if it's peak week

1. Make everyday pricing beat the promotion Consumer confidence is nearly at an all-time low: 53.3 across North America in March 2026, per the University of Michigan. The index was benchmarked to 100 in 1966 and has averaged around 85 historically, just above its record low of 50, set in June 2022.

Confidence index 50 low

85 avg

100 base

53.3 Around 70% of shoppers say they're extremely or very concerned about rising grocery prices, per FMI, even as food inflation has begun to stabilize. Concern hasn't followed price back down. That backdrop is sharpening how deliberately shoppers approach every trip. Units per trip continued a multi-year decline in 2026, per Numerator, with consumers buying fewer products to offset higher prices, even as spend per unit rose. Shoppers are buying fewer things every visit, more deliberately, at a higher price per item. Bill Aull, partner at McKinsey, frames the shift as a move away from impulse: shoppers are cutting back on unplanned purchases and trading into private label as they build their baskets with more intention.

Per Forrester, retailers are treating profitability as this year's top imperative, rolling back generous no-questions-asked return policies to retain valuable customers while discouraging unprofitable shopping behavior. That same discipline is showing up in how grocers price. Shoppers want dependable everyday pricing now instead of promotions. Over 80% of grocers are responding by prioritizing Key Value Items, and Aull notes a parallel push into fresh pricing amid commodity and protein cost volatility. The pattern's already visible at scale, and across grocery formats. Kroger's lowered prices on more than 3,500 items while simplifying its promotional structure. Target's cutting prices on roughly 5,000 items concentrated in nondiscretionary grocery and household essentials. Two of the country's largest grocery sellers, moving on the same lever, in the same window. Private label's riding the same wave. Privatelabel sales hit $330 billion in the US in 2025, capturing 24% of retail food and beverage dollars, per Circana. It's outpaced national brands in both dollar and unit sales growth for three straight years running, a shift driven by shoppers increasingly rating private labels on par with name brands on quality. Kroger price cuts

Target price cuts

3,500+ items

~5,000 items

simplifying promo structure

grocery and household essentials

Private label has crossed a perception threshold: 85% of consumers now rate store brands as equal to or better than national brands, a sharp climb over the past decade. And the category has become a differentiator in its own right, with 69% saying leading retailers stock private label products they can't get anywhere else.

Target's own grocery private-label strategy shows the split in real time: its Dealworthy line competes on price, while Good & Gather is positioned more on quality, evidence that even within one retailer's grocery assortment, price and quality have become separate. The infrastructure behind value is getting sharper too. Promotions are shifting from one-off events to a targeted system, expected to reach 55% fully personalized within two to three years, up from 35% today.

The retailers moving the needle in this area treat value as a coordinated system across pricing, promotions, loyalty, and private label, with profitability as the filter for which investments earn their keep. That discipline matters as shoppers weigh price, quality, and convenience more deliberately than they have in years.

Discounting doesn't fix the underlying problem. Broad promotions train shoppers to wait for sales rather than buy on habit, and mark down items that would've sold anyway, thinning already-tight margins without building loyalty. What shoppers read as real value is relevance.

$330B

24%

85%

69%

US private label sales, 2025

Share of food and beverage dollars

Rate store brands equal or better

Say retailers stock exclusive private label

2. Make personalization reciprocal, not extractive 92% of retail marketers say AI is helping them better understand consumer preferences, but just 53% of consumers say brands are getting it right, a 39-point gap between perception and experience. That’s the clearest evidence yet that grocery's personalization investment and its personalization delivery have come apart.

The confidence-experience gap 100

39-point gap

Retail marketers

Consumers

On the flip side, we're seeing this too: product discovery is quietly leaving the search bar. Shoppers are asking Gemini, ChatGPT, and Instacart to plan the basket for them instead of typing into a digital shelf. That’s a structural problem: personalization now has to land somewhere retailers don't fully control.

This is already playing out at scale. Instacart became the first grocery partner integrated with Google's Gemini in May, letting shoppers build a real cart through natural conversation. It's now embedded in ChatGPT and Claude too. Data fragmentation sits underneath the gap. 64% of grocers now describe trade, retail media, and joint business planning as mostly or fully integrated, up from a landscape where those conversations ran separately. But supplier-retailer integration is a different problem than integration across a grocer's own first-party data, loyalty history, and realtime behavior, and it's that second kind of fragmentation that keeps personalization generic even as the technology improves. The cost of getting this wrong is higher in grocery than almost anywhere else in retail. Grocery runs on routine, and routine runs on trust: a personalization oversight in a category shoppers visit weekly compounds every week it goes uncorrected, unlike a mistimed recommendation in a category people buy twice a year. Shoppers can tell when personalization is guesswork. Retail marketers are confident their AI-driven personalization is working. However, consumers aren't convinced, and the data shows a gap wide enough to carry real commercial consequences. Nearly half the customer base is telling retailers, in effect, that what brands think they know isn't translating into anything that feels relevant.

Grocery makes that gap more expensive. “The category is built on routine,” Meredith Gaiser Mitchell, Head of Retail and eCommerce Industry Marketing at Braze explains. "Shoppers expect you to know them. They've been buying the same brands, the same quantities, on the same schedule for years.” Miss that, and it erodes trust the brand spent years earning.

Share data, get something tangibly better back: a sharper offer, a real recommendation, a faster checkout.

The transaction has to be transparent, but it Shoppers are sharing less of the data personalization depends on, which makes this also has to be harder to fix than it sounds. Across the reciprocal...retailers broader retail landscape, per Braze, 27% of consumers refuse to share any data with AI who are asking for agents at all, even when promised a better experience in return. Mitchell argues grocery data but delivering the has less room to absorb that reluctance than same generic most categories, given how much personalization in the category depends on experience are exactly the kind of behavioral and purchase data shoppers are increasingly hesitant to accelerating that trust hand over. erosion. Roughly one in four shoppers, in other words, are opting out entirely, before personalization even has a chance to work. The smarter move is building a value exchange the shopper can actually see.

Meredith Gaiser Mitchell, Head of Retail and eCommerce Industry Marketing, Braze

3. Treat membership as the start of the data strategy Most grocers have built loyalty infrastructure. Few have built loyalty. Points, tiers, and digital cards are table stakes now, and the real gap, per Lily Varon, principal analyst at Forrester, is that grocery loyalty has been defined by points, tiers, and discounts long after the rest of retail moved past that definition. Varon points to Starbucks and Apple as the clearer model: both operate as loyalty companies. Her framing draws a sharp line between two different kinds of loyalty. Behavioral loyalty is habitual, driven by proximity and convenience, the shopper who returns because the store is on the way home. Emotional loyalty is what actually makes a relationship sticky, and per Varon's prior research on subscription retention, it's emotional loyalty, not behavioral habit, that keeps a customer through a price increase, a bad in-store experience, or a competitor's aggressive promotion. That distinction reframes what a grocer's loyalty program is actually competing with. Varon argues that grocers are, functionally, subscription businesses without a card on file, the shopper visits on a predictable cadence, buys largely the same list, and could churn to a competitor with the same ease as canceling a subscription. She points to Trader Joe's, Aldi, and Wegmans as grocers that have built emotional loyalty holistically, through a consistent identity and experience that gives shoppers a reason to stay beyond convenience.

The payoff for closing that gap is measurable where it's built well. Loyalty members who redeem personalized offers spend 4.3 times more annually than non-redeemers, per Antavo, but that multiple depends entirely on the offer actually being personalized, not on the program existing

4.3x

more annual spend from loyalty members who redeem personalized offers, vs. non-redeemers

Loyalty is what holds the whole value system together. It's the consistent data, across every channel, that makes everyday pricing, targeted promotions, and private label actually work as a system, instead of pulling in different directions. Joshua Reuben, associate partner at McKinsey points to where that data is starting to show real economic weight, grocers turning first-party data into standalone analytics businesses that sit alongside retail media itself, citing Kroger's 84.51° and Walmart's Scintilla as examples. Loyalty data is starting to be a monetizable asset, but only for the retailers with the infrastructure and the emotional relationship to back it up. Varon's diagnosis is strategic and emotional. Braze shows what that gap looks like operationally, inside the systems retailers actually run day to day. Loyalty is earned in real time: most grocers built a loyalty program, but few built loyalty. That distinction is where most loyalty programs quietly fail.

The fundamental mistake we see is treating loyalty as a program rather than a behavior," says Mitchell. Grocers have built the architecture: tiered membership, points systems, digital cards. Then the data collection just... stops. Loyalty and CRM data sit in separate silos, and retailers are holding rich behavioral signals, purchase frequency, basket composition, channel preference, that their engagement platforms can't act on in real time. The outcome is a program that rewards transactions and never gets around to building the relationship underneath them.

The gap between what grocery built and what loyalty actually requires is the core issue this trend addresses.The retailers succeeding are asking what a membership actually earns a shopper emotionally, in real time, before they ask what it earns the business.

Per research from Braze: 29% of retail leaders say their marketing content is assembled for each customer at the moment of engagement. 42% describe their personalization as based on past transactional or behavioral data, which sounds sophisticated on a slide, but is actually reactive: telling customers what they already bought, not anticipating what they need next.

29% Real-time, at engagement

Grocery has the least room to get this wrong. It's a category built on high purchase frequency and deep habitual patterns, "one of the most data-rich retail environments that exists," Mitchell calls it. The retailers pulling ahead treat loyalty membership as the start of a data strategy, not the end of a marketing campaign, and they're investing in real-time orchestration that can respond to a signal, an abandoned browse, a lapsed visit, a shift in basket composition, in the moment it happens, not a week later.

42% Reactive, past data

4. Fix availability where it costs you most Retailers often trust their inventory systems more than the shelf actually deserves. The execution data suggests that trust doesn't always hold, particularly in the categories where availability matters most to a shopper's decision to trust a store again. Fresh is the clearest case study. Fresh departments strongly shape overall store trust for 91% of U.S. consumers, per Logile, with produce execution the clearest read on store quality. But delivering that consistency is a challenge, with grocers citing real, compounding barriers:

88%

78%

69%

cite cost pressure

cite labor availability and skill level

cite supply chain complexity, specifically perishability

Two operators diagnose the same in-stock/ on-shelf gap from different vantage points: one from the data model underneath it, one from what fixing it looked like at scale.

Anyone who's run stores knows the number in the system rarely matches what's on the shelf. That's why cycle counts and zero-sales reports exist, and why they don't solve it: new errors accumulate faster than a periodic count can clear them.

When “in stock” doesn't mean “on shelf”

Short of a real-time indication from the shelf itself, there's no way to know how wrong the record is without physically counting, which is why most retailers, in 2026, still fall back on infrequent, store-wide counts.

Retailers have learned to live with an inventory system they don't fully believe in because, as Aidan Mittra, co-founder of OrderGrid, puts it, they've stopped believing there's another way to run the business.

Here's the paradox: a retailer can sit in the high 90s on in-stock %, the system grading its own homework, and still have real gaps down the aisle. In stock means the system believes the item exists somewhere in the store.

On shelf means a customer or picker can actually reach it. In-store shoppers have always quietly absorbed that gap: walk to an empty shelf, grab something else, or leave. Per a consumer survey covered by EMARKETER: About one in three shoppers say they run into out-of-stocks at least "sometimes," especially during promotions. 83% of shoppers are satisfied with their primary grocery store, but only 48% are willing to actually recommend it, the gap between tolerating a store and being loyal to one.

Online orders picked from the store took away even that quiet workaround. The system said the item was there, so the order got accepted, and the customer only finds out the truth when a substitution or refund shows up. Nothing in the legacy system was ever built to catch that before it happened. Roughly two-thirds of these gaps, according to Mittra, are items not in the building. The rest are technically in the store, in the back, down to the last unit, misplaced, or already sitting in someone else's cart. That split holds up at the macro level too: IHL Group estimates puts the global cost of out-ofstocks and overstocks combined at $1.7 trillion a year, with out-of-stocks alone accounting for roughly two-thirds of that total. A meaningful share of the outs trace back to phantom stock: the system believes units exist that were never actually there. The fix is targeting the specific items the system is overstating: some are visible in data already on hand (time since last sale), the rest only findable on the floor. More frequent counting isn't it. Detection is only half of it; prioritization matters as much. Break gaps down by section and hour and the ranking changes fast.

1 in 3 shoppers say they run into out-of-stocks at least "sometimes," especially during promotions.

83%

of shoppers are satisfied with their primary grocery store

48%

are willing to actually recommend it

Bakery runs out often but the items are cheap, so it's near the bottom for value lost; meat runs out less often and loses the most each time. Rank by value, not raw frequency, and that's the difference between busywork and margin recovery. However you find the gap, it only closes if it's a task with someone's name on it. What makes that workable is holding the shelf record separately from the back room, so the moment a picker comes up short the system shows whether there's more to pull. The gap becomes a live task with a deadline in minutes, not hours: whoever takes it checks the shelf and either restocks or corrects the record on the spot. Most close in time, Mittra notes; the rest escalate. The advantage is knowing where inventory is wrong fast enough to fix it before the customer feels it, the real distance between the 83% who are satisfied and the 48% who'd recommend their store. Close that gap enough times, and satisfaction turns into loyalty. One retailer's answer to that same gap shows what closing it actually looks like operationally, at scale.

Availability is hospitality Grocers often believe their systems accurately reflect what's on the shelf. Reality frequently disagrees, and the gap between what a system reports and what's physically available to a shopper quietly undermines pricing, promotions, and the broader shopping experience.

That shift was powered by Vusion's connected-store technology: more than 11,000 electronic shelf labels deployed, shelf cameras automating on-shelf monitoring to detect empty facings and replenishment needs multiple times a day, and real-time alerts replacing manual gap scans entirely.

As stores scale, that gap only gets harder to manage. The Fresh Market, for instance, was completing more than 6,500 manual price changes every month before rethinking how it handled shelf execution, slowing teams down and introducing inconsistency at exactly the moment accuracy mattered most. Instock gaps and shrink remained persistent risks, quietly eating into both sales and the in-store experience. “Availability is Hospitality,” says Brian Johnson, president and CEO of The Fresh Market.

The outcome was a fundamentally different relationship between the system of record and the reality on the floor. Store teams stopped relying on periodic, manual gap checks and started receiving actionable alerts the moment a shelf went empty, correcting gaps faster and with far less manual effort. That shift also freed staff from repetitive, low-value tasks toward the highervalue, customer-facing work grocery increasingly depends on.

When our shelves are full and our teams have clarity, we deliver the experience our guests expect, The takeaway holds for other retailers too; a every day, in every record needs to be honest, in real time, about where it's wrong, and a mechanism that store closes that gap the moment it's detected Brian Johnson, president and CEO, The Fresh Market

rather than at the next scheduled count. Most grocers, even today, are still relying on the latter.

5. Build workforce flexibility to offset rigid automation As automation absorbs repetitive work, price updates, inventory counts, basic replenishment, the store associate's role is shifting toward higher-value, customerfacing work. That shift is real, but far from finished. Store-level execution and consistency is one of the top capabilities required to succeed in fresh, and that consistency depends directly on staff, not just systems. A more differentiated prepared food offering requires different skills, different training, and different hiring than a traditional grocery role, which is exactly why labor availability and skill level remain such a persistent barrier to scaling fresh. Technology is starting to close that gap rather than just adding to the workload. Digital tools for daily tasks are improving execution and consistency on the floor. The real ambition behind that investment is using AI to make every store manager as effective as the very best one. This shift is already reaching the board. “Labor availability is a strategic constraint on how many positions a grocer can actually fill,” says Renee Hartmann, retail strategy advisor, “That affects store formats, service levels, and staffing models. Ops is catching up to a reality boards already have to account for."

Where frontline AI fits in An associate who doesn't know which shelf holds a product doesn't need to send a shopper hunting through a 2,000-product aisle. Mark Ruston, global retail leader at Capgemini, points to Frontline AI as the answer. Ask the agent, and it flashes the shelf label. The friction that shows up, per Ruston, is about how the interaction feels on both sides. For customers, it's extra steps, an associate relaying a question to AI, then repeating the answer back. For employees, it's whether the tool reads as support or surveillance. His example, an associate asked about a promotion or loyalty offer gets an instant answer instead of digging through systems or memory. The same assistant handles returns, markdowns, and substitution policy too. None of it adds work, Ruston asserts; it makes existing work faster. Customers get quicker service and associates get support.

Automation only pays off when the workforce can be flexible around it

That last one is the eye opener: the industry is already trying to build its way out of the exact problem Pátek describes, which makes it a real, recognized cost rather than a hypothetical one. In a business as fastmoving as same-day grocery, that's the central design constraint.

Automation gets sold as the fix for fulfillment economics, and in fairness, it mostly is. Grocery is buying that pitch: automation revenue in the category grew nearly 20% in 2024 alone. Dalibor Pátek, VP product, retail technology at Rohlik Group doesn't dispute that automation is close to a necessary condition for profitable same-day delivery. What he pushes back on is the assumption that automation is a clean win with no downside. It has a specific, structural drawback, and grocery hasn't fully priced it in. A manual warehouse scales like a dial. Automated capacity scales like a wall. "Imagine if you had a warehouse which is functioning fully in manual mode," Pátek says. "You have there a thousand workers. You can scale it up and down quite easily. You add 200 workers, you will not have 200 workers, it still works." Automated infrastructure doesn't grant that same forgiveness. It's a fixed asset, sized to a bet about future demand, and getting that bet wrong is expensive in a way adding or cutting headcount never was.

That bet is only getting bigger:

Hartmann sees the same downside Pátek is describing, just from the staff side rather than the systems side. "Automation brings fear for the staff first," she says. "Grocers who get it right show staff early that automation takes over the routine tasks, not their jobs, and frees them up for the parts of the role they actually enjoy, like working directly with customers. Once staff see that in practice, they embrace it. Skip that step and you get resistance no matter how good the technology is." That same tension shows up in how the tool itself gets rolled out. Whether automation actually frees store associates for highervalue work, rather than shifting the friction elsewhere, comes down to how the tool's introduced, not the technology itself, says Mark Ruston, global retail leader at Capgemini. One example: a deployment aimed at reducing equipment-related service callouts, where staff were guided step by step through self-diagnosis before an engineer was called. The target was a 10% reduction. The actual result was 27%, because staff experienced it as empowering rather than as extra workload.

25%

1/3+

of capital spending going to automation, on average, per McKinsey

of the capital budget is in logistics & fulfillment - the largest share of any sector

72% of logistics firms plan to shift to Robotics-as-a-Service, usage-based over fixed capital

The real lever is where the headcount stands Once the layout is right, the actual lever left is how work gets fed into the system hour by hour. Rohlik sees ~30% swings in hourly order and item throughput on a standard day. Build fixed capacity for the peak and you've overbuilt the rest of the day; build for the average and you're short at the peak. A static plan loses either way. Rohlik's answer was to provide a real-time solution. "It's super difficult with human brain to decide what exactly should I do to balance throughput across the sectors," Pátek says—too many goals and constraints, updating too fast for anyone to hold across a shift. So the decision no longer sits with a person. Rohlik's fulfillment platform Veloq takes: The live forecast

Shape of demand still coming

Each worker's productivity and skill profile

Orders already placed

Real-time sector assignment, all shift long

and assigns people to sectors in real time, continuously, all shift long. That’s a live reallocation that runs as conditions change. So the redesign is the routing—the workforce routed moment to moment to protect two things: the customer's order, and the fixed capacity behind it, which no one can argue with the way they can a headcount plan. The approach has a track record beyond Rohlik, too, cited by Intellectyx: Up to 25% improvement in workforce productivity for organizations using AIdriven workforce management Meaningful cuts to overtime and underutilization

Automation earns its keep through flexibility— holding speed, accuracy, and the customer promise steady as demand shifts. But the tech only decides where labor goes. Hartmann's point is what labor is for: human workers matter more because they deliver the shopper experience. The winning strategy comes down to knowing which tasks belong to technology and which belong to people.

Where AI confidence breaks down

71%

of organizations don't yet fully trust autonomous AI agents for enterprise use per Capgemini

Machine learning forecasting has run reliably for over a decade, now handling millions of SKU-location combinations. The breakdown, Ruston suggests, comes downstream—in diagnosing what went wrong once execution and forecast diverge. Lokesh Chawla, chief sales officer and AI lead for US consumer and retail at Capgemini, likens AI to a new employee: it earns trust through consistent, verifiable results, not a top-down mandate, and needs time to prove its judgment. The real question, he argues, is whether the organization is ready to act on what the model tells them. The same test applies to labor. Rohlik's routing system decides where staff go shift to shift. If planners won't trust a model on the forecast without proof, grocers can't expect managers to trust one on headcount without it. The workforce redesign has to earn that trust to keep moving.

6. Close the lastmile gap without handing away the customer Same-day delivery has moved from differentiator to baseline expectation, and the math behind it is getting harder, not easier. 45.3% to 63.3%: the share of online grocery orders delivered rather than collected, 2018 to 2025, a 28.4-point swing, per Coresight Research data. That shift accelerated sharply in just the last few years: delivery and collection were roughly split as recently as 2022 and 2023.

a 28.4-point swing toward delivery — roughly split as recently as 2022–23

45.3%

63.3%

2018

2025

Removing friction from the shopping mission is the lever that matters here. That's the gap retailers keep missing when they compete on delivery fees instead of on how effortless the whole mission feels. That growing demand is on a collision course with fulfillment capacity. Across 182 major US metro areas, current online grocery demand sits at $85 to 95 billion, projected to reach $123 to 142 billion by 2030. Existing capacity can absorb only another $13 to 17 billion, and currently announced expansion adds just $2 billion more, leaving a projected $23 to 28 billion gap by 2030, one expected to widen further as online penetration increases.

$85–95B

$123–142B

+$13–17B

+$2B

Existing capacity can absorb

Announced expansion adds

Current online grocery demand

Projected demand by 2030

$23–28B projected fulfillment gap by 2030, across 182 major US metros

Convenience is driving it:

67%

Saves time

Free delivery

52%

Fits their schedule

52%

Discounted delivery

12%

There's no single fix for delivery unit economics as delivery keeps outpacing cheaper options like in-store pickup. Cost per order, order value, and fulfillment model all have to move together. The real levers: cutting last-mile cost through third-party partnerships and pre-negotiated volume, fulfilling closer to the customer, growing basket size, and automation.

The grocers doing this best take a missionbased view. When a delivery drives meaningfully higher loyalty or value, it's worth absorbing more of the cost. When it doesn't, that cost passes to the shopper through fees, minimums, or assortment limits This same recurring-order economics problem connects directly to the payments infrastructure. Traditional card mechanisms were never built for recurring, card-on-file delivery and subscription orders. Card expirations, lost cards, and fraud all drive payment retries and failed recurring charges. That's precisely why network tokens and emerging account-to-account rails like variable recurring payments are becoming load-bearing infrastructure for last-mile economics as noted by Varon. The economics are trending in the right direction, if not yet resolved, per McKinsey. 46% of grocers say ecommerce is already profitable, and 72% expect it to become more profitable than in-store within two to three years. More than half expect to increase third-party and centralized dark-store fulfillment specifically to help capture the delivery-driven demand ahead. The upside extends well past the delivery transaction itself: when a store-only customer becomes an omnichannel customer, retailers see a 5-plus point increase in that customer's overall share of wallet, evidence that fulfillment strategy is becoming a competitive advantage.

Outsourcing the order means losing the signal Most grocery operators frame delivery as build-versus-buy. Pátek says that framing was broken from the start: there are three doors, not two. Hand a platform the delivery and the customer relationship, and you get speed at the long-term cost of the channel. Build the entire stack yourself, and you get control, at the cost of years and internal capability most grocers don't have. The third door keeps the thing that matters, the customer relationship, and buys the execution underneath it. It's also the hardest to walk through cleanly, Pátek notes, because everything rides on picking the right partner. The headline number looks modest: Instacart, for example, charges retailers 5% to 8% per order. But that commission fee doesn't show up on the invoice for what it actually costs. RETHINK Retail studied Rohlik Group's approach to owning last-mile delivery and found third-party platforms can drain up to 40% of order revenue once everything past that commission line gets counted. The real price is control: You don't control how your offering gets displayed. You don't control whether the platform keeps favoring your listing once a competitor with a similar offer shows up. You don't own the demand signal, or the reaction to your own capacity.

“Platforms can actually optimize for their cost and not to make your customer happy,” Pátek says: basket steering, promotion sequencing, all engineered around what serves the marketplace, not the store. The erosion is slow enough to miss in the moment and expensive enough to matter in three years. Every order makes the platform smarter about your customer. It doesn't make you smarter about anything.

You can definitely find partners which will give you advanced automation of the warehouse, but you will Run the unit economics past a single order and the real problem shows up: there's no hardly find the partners cost bend-down. As volume grows, the retailer on a platform rarely keeps the savings which will give you that scale should generate. The platform automation, does. Pátek has the counter-example on orchestration of the hand: Rohlik’s Veloq platform drove last-mile cost down 35% in one region through fulfillment, coordination operational and execution improvements, not just volume. A platform can realize that same with last mile, scale-driven saving over time, but "most likely forecasting, marketing, you will not be [able to extract it]," Pátek says, "because you will become dependent on all in one package. them, and they will keep all the value."

Rohlik's own-delivery story, commercialized as Veloq, the platform Rohlik now offers to other grocers gets cited constantly as proof that owning last-mile at scale is possible. Pátek says the model transfers; the shortcut doesn't. Rohlik built vertically integrated from day one: forecasting, fulfillment, and delivery owned end to end, holistically optimized rather than stitched together.

Dalibor Pátek, VP product, retail technology, Rohlik Group

The magic is coordinating every piece at once, so no one component can compensate for weakness in another. That's the warning for any grocer eyeing same-day as a bolt-on: no single silver-bullet vendor hands you this, which is why Rohlik is offering its Veloq platform to other grocers. A retailer not ready to own the last mile fully should treat third-party delivery as a channel, Pátek argues, not a fulfillment strategy: use a platform or a marketplace listing, but keep it to one channel among several, with a backup plan if the terms deteriorate. Using a platform isn't the mistake. The strategic question is who gets smarter from every order and who owns the relationship when the customer comes back. Every order builds your advantage or someone else's. Delivery is the transaction; the data and the relationship are the asset, and whoever owns those owns the customer's next order.

Two operators, two different fulfillment models, converge on the same warning: the moment you hand off delivery, you hand off the signal that would let you improve it.

When the picking isn't yours, neither is the data The pressure of on-demand grocery doesn't land evenly. And increasingly, it's not landing on the retailers with any real ability to manage it. When a retailer's own staff pick online orders, something valuable comes along almost for free: every short pick or substitution is a measurement of a real gap on the shelf, generated the instant it happens instead of surfacing in an exception report the next morning. That signal can route real-time gap scans to exactly the sections and hours running short, closing the loop between online and in-store instead of treating them as two separate problems. Most North American grocers didn't get there. They came to on-demand through a thirdparty marketplace, where the picking isn't theirs, says Mittra. Per McKinsey, here's how grocers currently fulfill e-commerce orders: through thirdparty partners

17% 26% through centralized or dark-store fulfillment

57% through their own stores

That arrangement takes the operational burden off the table. It also takes away the one signal that would let a retailer fix the underlying problem. And the burden doesn't disappear so much as change shape. The sale is still the retailer's, but the picking is done by people it doesn't employ, working its aisles at the busiest hour of the day, using its tills and its staff's time. The store keeps the cost of the peak while handing away the signal that would help it manage the next one. A marketplace only shows what a retailer feeds it, which requires real-time inventory and the APIs to push it, infrastructure Mittra says most grocers still don't have. In practice, that tends to mean batch-updated feeds sitting between demand spikes, quietly showing stock that already left the shelf. The customer blames the retailer. You can outsource fulfillment, but you can't outsource the customer's experience of your brand. The real cost is structural. The marketplace keeps the demand data a retailer would need to forecast against. It keeps the short-pick signal that would tell the retailer exactly where the shelf is going empty for the next customer, who, Mittra notes, is statistically more likely to be shopping in the store than online. That's still the safer bet: even with online grocery growing, it made up just over 19% of total U.S. grocery sales in early 2026, per Brick Meets Click data, meaning roughly four out of five grocery dollars are still spent in a store. Hand off the picking, and you're paying someone else to take away a genuinely valuable signal. The opportunity is to be intentional about what gets outsourced. Own enough of the fulfillment experience to keep the signals that improve availability, forecasting, and the next customer order, then use marketplaces where they add capacity and reach. Every signal a retailer keeps is a chance to make next time's promise more trustworthy. Every signal handed to a marketplace is a chance the retailer never gets back.

7. Run every week as if it's peak week Grocery used to plan around predictable seasonal surges, holidays, back-to-school, the occasional weather event. That rhythm has broken down. Demand now spikes unpredictably and continuously, driven by ultra-fast delivery expectations, socialcommerce-driven trends, and shifting consumer routines that don't map to a calendar the way they used to. The fulfillment infrastructure gap explored in the last-mile theme compounds this problem directly: a system already projected to fall $23 to 28 billion short of demand by 2030 under normal growth assumptions has even less slack to absorb an unplanned spike. What it takes to operate as if every week could be peak week is less about capacity for the biggest day of the year, and more about the flexibility to absorb volatility on an ordinary Tuesday. A spike exposes existing problems, and two operators have built very different answers to the same pressure.

When the shelf doesn't match the screen, spikes are when it shows A demand spike compresses a normal day's ordinary failures into a few high-pressure hours, says Mittra. The data shows the pattern: about one in three shoppers report hitting out-of-stocks at least "sometimes," most often during promotions, per EMARKETER, exactly the high-demand windows Mittra describes. What a spike adds is time pressure, and time pressure changes behavior predictably.

A picker under pressure reaches a shelf, can't spot the item, and hits the "not found" button, the fastest way out. In most stores nothing happens next: the order gets substituted or refunded, and no one checks whether the shelf was actually empty or the item just got missed. In stores that do route "not found" into a follow-up check, real gaps get buried in false ones, generated by the same time pressure that created the first problem. More staff sounds like the obvious fix, but Mittra is candid about why it usually isn't: retailers are already short-staffed exactly when spikes hit. They're walking into peak demand with the fewest seasonal hands in 15 years, even as holiday spending crosses $1 trillion for the first time, per National Retail Federation. The more realistic gain comes from the labor already on the floor: batching multiple on-demand orders into one trip through the store instead of picking them one at a time. The structural fix is about what pickers are measured on. If speed is the only metric, pressing "not found" and moving on is the rational choice every time. Mittra's answer: make accuracy count alongside speed, and rank the check-task queue so limited capacity goes to the checks most likely to be real. Ahead of the spike, a meaningful share of on-demand volume can be forecast at the item level, letting a retailer roster staff and pre-position stock before the pressure hits.

Preparation beats reaction every time demand spikes Most companies frame peak capacity as two bad options: over-provision and eat the idle cost, or under-provision and eat the missed SLA and overtime. Pátek calls that the wrong trade-off. "It's much more costly to do the reaction," he says, "than to be prepared in advance." Rohlik plans for the variance, not one optimal scenario: a labor curve shaped to the demand curve, with enough buffer to hold utilization if demand dips and enough slack to absorb a spike without blowing the SLA. Neither buffer is free, and the model's job is deciding how much of each to carry. The payoff shows up only under pressure: demand spikes, and the order still arrives on time, at the promised price and quality, because the buffer existed before the demand did. That planning has to happen before the spike, which is where automation adds a complication most grocers don't see until they've already built for it.

The spike is the magnifying glass. The retailers that perform through it can see demand coming, know where inventory actually sits, and help their teams make the right decision before the customer feels it. The real reliability test arrives at 6pm the Sunday before Thanksgiving, not on an ordinary Tuesday. A vertically integrated operator faces the same crunch from the capacity side rather than the shelf side.

8. What grocery leaders do differently The grocers closing the real gap, between what they collect and what they actually act on, are doing it in real time, at the moment it matters to the customer. Five imperatives separate them from everyone else:

Make targeted value investments

The grocers closing the real gap, between what they collect and what they actually act on, are doing it in real time, at the moment it matters to the customer. Five imperatives separate them from everyone else:

Choose winning battlegrounds

Pick the few areas — fresh, prepared foods, wellness, private label, price, delivery, or experience — where a banner can credibly win, and over-deliver there with enough consistency that shoppers actually notice.

Turn technology into an execution edge

Use digital tools and AI specifically where they improve fresh quality, in-stock rates, substitutions, and associate decision-making, not just where they digitize existing complexity.

Drive ecommerce profit by mission, not as one blended business

Some fulfillment options deserve subsidy to deepen loyalty; others need fees, minimums, or redesign to protect margin.

Build one commercial engine

Connect customer insights, pricing, promotions, loyalty, and retail media into a single operating model and financial algorithm, or risk ending up with a stack of promising side projects that never change how the business actually runs.

Getting there starts with a simpler question, Ruston: says who are you serving, and what role do you want in their lives? Get clear on whether you're competing on value, service, or convenience, then use AI to strengthen that position, not blur it. From there: Big, Deep, Narrow. Pick a problem big enough to matter, deep enough to cover a full workflow, narrow enough to actually measure, then scale. Loyalty sharpens that last imperative most. Trader Joe's, Aldi, and Wegmans got there by building an emotional relationship first and letting the data infrastructure follow, rather than the other way around, per Forrester. That's the thread running through every theme in this report. Value-seeking shoppers, the personalization gap, the loyalty reset, AI's uneven trust, availability, workforce, last-mile, continuous peak, none of these are unrelated problems. They're all the same question: is the data a grocer collects actually connected to what it does next? The retailers answering yes, structurally and consistently, are the ones building real advantage this year.

Editorial Contributors Lavina Suthenthiran, Senior Retail Analyst Jeremy Goldman, VP of Editorial and Insights Annie Petrova, Graphic Designer

Appendix A: Interviews Bill Aull, Partner, McKinsey Lokesh Chawla, Chief Sales Officer & AI Lead for US Consumer and Retail Industries, Capgemini Tom Kilroy, Senior Partner, McKinsey Alexandra Kuzmanovic, Partner, McKinsey Renee Hartmann, Retail Strategy Advisor Brian Johnson, President and CEO, The Fresh Market Meredith Gaiser Mitchell, Head of Retail and eCommerce Industry Marketing, Braze Aidan Mittra, Co-founder, OrderGrid Dalibor Pátek, VP Product, Retail Technology, Rohlik Group Joshua Reuben, Associate Partner, McKinsey Mark Ruston, VP and Global Retail Lead, Capgemini Thaddeus Segura, Chief Product Officer, Vusion Lily Varon, Principal Analyst, Forrester

Appendix B: Sources Accenture Antavo Bain & Company Braze Capgemini Coresight Research Food Business News Digital Commerce 360 EMARKETER- Pricing mismatch, AI adoption in grocery in accelerating fast FMI Forrester — profitability predictions · AI payment autonomy · payment tokens · general Grocery Dive (Kroger) IHL Group Intellectyx McKinsey & Company National Retail Federation Numerator PYMNTS Retail Dive (Target) RETHINK Retail University of Michigan (via FRED) Visa Vusion