A store in Kansas executes the campaign exactly as headquarters intended. A store in New York does the same. Both receive a perfect compliance score. But what if the campaign was right for one market and wrong for the other?
Retail execution software was built to close the gap between a commercial plan and what happens where customers buy. Audits established what was present. Mobile workflows made the work accountable. Computer vision made conditions easier to measure at scale. Sales-linked systems began asking which activities actually produce results.
The problem has changed—not because consistency no longer matters, but because doing the same thing well does not guarantee the same commercial result. Store size, customer needs, competition, local demand and timing can change what the right action is.
Some of the industry's strongest companies already recognize this. The question is no longer whether localization is possible. It is how far the category has moved toward continuously learning what works where, under what conditions—and then executing accordingly.
The commercial return is the point. Execution is the means.
Brands, retailers and consumer-goods manufacturers invest in assortments, shelf space, product launches, promotions, field teams and store labor because they expect those investments to produce profitable sales. Retail execution is the machinery that turns those intentions into observable activity.
The complication is that retail networks aren't uniform. A brand may operate some stores, license others and sell through independent retailers it cannot direct. Even inside a wholly owned chain, a city-center convenience store and a suburban destination store can serve different shopping missions.
A sound strategy therefore has two distinct requirements: make the chosen plan happen reliably, and make sure the chosen plan is commercially appropriate for that location. The first problem helped create retail execution software. The second is increasingly shaping its evolution.
Define the standard. Distribute the plan. Inspect deviations. Correct the store.
Use local and network-wide evidence to learn which approaches work, where and when—then execute the appropriate plan.
This contrast describes a historical tendency, not every current vendor. Category management and sophisticated retailers have long used local insight. The competitive question is how central this learning has become to the software's business proposition.
How the category learned to ask better questions
Retail execution did not move through a neat sequence of generations. The approaches below overlap, and different buyers still need different combinations. But each added a more consequential question to the original one.
| Period | The new question | Technology and approach |
|---|---|---|
| 1930s–1970s | What sold or appeared in the store? | Store audits, market measurement and barcodes created better evidence about the point of sale. 0102 |
| 1990s | Can manufacturers and retailers make better joint decisions? | Category management and Efficient Consumer Response connected assortment, shelving and replenishment with shared economics. Learning was already an ambition. 03 |
| 2000s–2010s | Did the plan get executed? | Field-force software, mobile forms, photographs, tasks and compliance scoring made store-level gaps more visible and actionable. The underlying out-of-stock problem was already documented in 2002; a supplier account traces Opterus's task and communication deployment at Pet Supplies Plus to 2013. 0411 |
| 2010s–2020s | Which gaps or interventions are worth addressing? | Image recognition, on-shelf analytics, POS connections and field prioritization related observed conditions to commercial opportunities. 0506 |
| Emerging across the market | Which plan is right for which location—and when does that change? | Store clustering, localized assortment, experimentation and contextual recommendations move from auditing a fixed plan toward adapting it. 161820 |
The periods refer to overlapping waves of adoption, not the invention date of each capability.
From making the plan happen to knowing which plan is worth making happen.
This is not an argument against execution. A recommendation that never reaches the shelf has little value. The strategic shift is that execution quality and plan quality are separate variables—and optimizing only the first can make an organization exceptionally good at implementing the wrong choice.
Retailers are already adapting the plan
Localized retail strategy is not a speculative consequence of generative AI. Retailers and their advisers were pursuing it well before the present wave of artificial intelligence.
Assortment is becoming a location-level decision.
McKinsey documented REWE's work on automated localized planograms and Żabka's store-specific SKU selection. The business premise was not that every location should receive a better standardized assortment. It was that a location's shopper needs should inform which products earn space there. The report also warned that measuring local effects requires testing rather than simple before-and-after comparisons. 20
Two stores in the same channel can need different shelf principles.
Trax describes a manufacturer's finding that shelf flow drove sales in one store while brand adjacency mattered more in another. It explains using shelf and sales patterns to create cluster-specific action plans. This is unusually close to the alternative belief being investigated here. Reported commercial gains are vendor-authored and should not be taken as independently proven causal effects. 16
Even the purpose of a store may differ.
A 2026 retail analysis argues that stores increasingly need distinct missions—convenience, discovery or fulfillment—with layouts, assortment and staffing matched to that role. A uniform network can be easier to administer while leaving individual locations optimized for none of those missions. This is an analysis of market direction, not proof that a particular execution platform delivers it. 21
The executive complication remains: retailers and partners have constraints and interests of their own. Nike's wholesale reset, McDonald's planned franchisee investment support and Adidas's comments about retailers allocating their own space show why a centrally chosen plan may also need economic legitimacy and operational capacity. Those cases concern channel economics, not direct evidence of store-specific merchandising gains. 121315
The perfect store was never going to be one perfect store.
The historic Perfect Store idea made a complicated problem manageable: define the desirable assortment, availability, shelf position and promotion, then measure how closely each store meets that picture of success. Consistency matters. Safety, brand identity and other nonnegotiable standards should often remain common.
But a universal commercial scorecard can hide a problem. A store might achieve the highest possible compliance score against a plan that is poorly suited to its customers. Or a store may depart from a guideline in a way that produces better results under specific conditions.
What if a perfect store is not the store that most faithfully follows the plan, but the store executing the right plan for its market?
That does not imply 500 entirely independent stores inventing 500 strategies. It might mean six recognizable market types, a small number of tested variants or a recommendation that changes when inventory, competition or customer behavior changes. It also doesn't imply that local managers always know better: they see important signals, but may miss network-wide economics.
Adaptive execution means combining local evidence with network-wide learning. The organization learns where each approach works, not simply which approach wins on average. It standardizes what must be consistent and adapts what evidence shows should differ.
Who can do this—and who is leading with it?
The market already offers significant pieces of adaptive execution. But a product capability, a documented customer application and a vendor's lead strategic argument are three different kinds of evidence.
The table below takes each provider's current homepage hero at face value, then asks how far its documented work goes toward learning which approach works in which location. It is not an exhaustive product audit or a ranking of revenue, share or customer satisfaction.
| Vendor | Homepage hero | What the product or a customer has done | G2G reading of the strategic conversation |
|---|---|---|---|
| Trax / FORM | Trax: “The Future of Retail Excellence is Here”FORM: “Enterprise-Ready AI for the Frontline”2223 | Trax explicitly describes identifying different sales-driving shelf principles by store and delivering action plans by cluster. FORM combines image recognition and frontline execution. 1607 | Closest to articulating the seam. Trax goes beyond better inspection toward category strategy and locally differentiated execution. The merged offering's end-to-end continuous learning is not proven by that claim alone. |
| Repsly | “The Future of Retail Execution Starts Here”24 | UNREAL Candy tested placement, pricing and coupons, compared execution with sales data, and scaled a winning holiday promotion across participating accounts; the supplier reports a 23% Northeast lift. 17 | Experimentation demonstrated. The example learns from different tactics, though it documents a winning approach being scaled—not a sustained system of different plans for different market types. Hero still leads with execution. |
| YOOBIC | “Your stores are costing you millions. And you can’t see it.”25 | Store Manager Copilot uses local data and signals to recommend actions; its published design includes comparable-store benchmarks and inventory checks. Platform messaging promises to scale what works and recognize local versus systemic issues. 34 1925 | Strong context-aware capability and an evolving performance proposition. Its homepage emphasizes costly blind spots and repeatable execution. Its product explanation also describes comparable-store benchmarking, inventory-aware actions and manager judgment. That is meaningful overlap with this seam; whether it continually improves distinct commercial playbooks remains a comparative proof question. |
| FranConnect | “FranConnect Connects the Entire Franchise Lifecycle in One AI-Powered Platform”26 | Compares unit performance and joins operational context to plans and coaching across independent businesses. 1026 | Different adjacent problem. Valuable evidence of differentiated unit management; less directly about learning which merchandising or campaign strategy works where. |
| Opterus | “Retail communication software for multi-location retailers”27 | Coordinates headquarters-to-store tasks, messages and execution, including a documented franchise-heavy network deployment. 1127 | Primarily coordination. Can communicate different instructions, but the homepage proposition doesn't center experimental learning from locations. |
| ParallelDots | “Unparalleled Visibility into Your Retail Execution”28 | Uses shelf images to detect product placement, availability and competitor conditions and give field teams corrective guidance. 28 | Observation-led. Rich local signals could feed adaptation, but the published lead proposition focuses on knowing and fixing shelf conditions. |
Homepage positions were reviewed against vendor materials in October 2026. Trax image-recognition operations merged with FORM in February 2026. Customer sales effects are vendor-reported unless independently identified.
The competitive picture is neither an empty opportunity nor a finished transition. Trax has already articulated differentiated plans. Repsly has demonstrated experimentation. YOOBIC is embedding contextual actions in operations. Their hero narratives emphasize different problems, but the capability-to-conversation distinction is not permission to ignore what their sales teams may already be saying to customers.
From localized plans to a network that keeps learning
Knowing that Kansas and New York need different promotions is one improvement. A more demanding capability is knowing which conditions make each approach work—and detecting when the answer changes.
Measure compliance; fix gaps before the selling opportunity closes.
Differentiate assortment, merchandising or promotion by local conditions.
Test interventions, observe outcomes, revise market-specific guidance and learn over time.
The third step is not simply more analytics. It requires a repeatable decision process: selecting the right comparison stores, distinguishing signal from coincidence, retaining what the organization learned, and translating that learning into the next set of local actions.
This may mean using store clusters rather than writing a unique plan for every address. It may also mean protecting nonnegotiable brand and operating standards while varying pricing, assortment, promotion or shelf priorities within permitted limits.
The strategic research does not end with the emerging intelligence race. It asks which alternative business proposition could change the executive buyer's decision: choosing the right local plan, making it economically executable, or learning from the stores that challenge it. The seams section tests these separately.
Why this is harder than an adaptive dashboard
The idea is attractive. The operating model is demanding, for at least four reasons.
- Differences aren't necessarily causes. A high-performing Kansas store may have better inventory, lower competition or a different customer mix. Controlled tests, suitable comparisons and repeated measurement matter.
- Local variation has a cost. More planograms, materials, promotions and instructions can overwhelm store staff and supply chains. A few robust clusters may outperform hyper-personalization.
- Brands and partners don't control the same levers. A manufacturer may recommend locally appropriate shelf treatment without having the authority to impose it. Recommendations must work within commercial agreements.
- Not everything should adapt. Safety, regulatory requirements, core brand promises and certain operational standards may need uniform execution, even when localized alternatives appear profitable in the short term.
McKinsey's work on granular retail decisions similarly warns that local assortment requires changes to analytics, planograms, operating responsibilities and rigorous testing to measure impact. More data is not itself a learning system. 20
This is also why there are legitimate businesses at every level of the category. A narrowly focused store audit tool may deliver more dependable value than an ambitious intelligence platform whose recommendations add complexity without improving returns.
Three fault lines—and three possible strategic reframes
The G2G seam method starts with an inherited way of explaining the customer's business problem. We look for occasions when that explanation leads to an insufficient or incorrect decision, establish what the customer actually experienced, and test an alternative belief against what vendors can already deliver. A provocative question should express that strategic possibility; it should not substitute for the investigation.
Three materially different changes in the business conversation emerge: which store plan is commercially right; whether a network has made that plan executable; and who gets to change the plan when stores discover something headquarters did not know. Each implies different buyers, capabilities, proof and economics. None is an uncontested opportunity.
| Candidate seam | Inherited answer | Alternative business proposition | Current standing |
|---|---|---|---|
| A / The plan | Set standards, verify compliance and fix deviations. | Continuously establish which commercially effective plan applies in each type of store. | Real business need; already pursued by Trax and others. |
| B / The economics | When a store doesn't comply, improve enforcement and support. | Design and sustain a plan that stores and commercial partners can execute profitably. | Executive-level problem; vendor control of remedy is limited. |
| C / The governance | HQ improves instructions with better visibility and analytics. | Make local exceptions and managerial knowledge inputs to governed, store-context-specific decisions. | Independent operating evidence; incumbent solutions address parts of the loop. |
| Cross-cutting / The moat | More images and execution records automatically generate superior intelligence. | Demonstrate that accumulated evidence improves future commercial decisions beyond what other providers can reproduce. | A possible defense and proof requirement, not a fourth seam by default. |
These are hypotheses about distinct ways to change competition—not a ranking of which commercial approach will win.
What if the perfect store is not a standard, but a hypothesis?
Inherited belief. Management defines the picture of success; retail execution technology helps headquarters verify that the picture has been implemented. Under this answer, raising compliance is a rational way to improve performance when the standard is appropriate.
Emerging business reality. In its First 50 initiative, Macy's tested changes to staffing, merchandising, presentation and events rather than simply auditing stores against an unchanged plan. Its 2024 results showed a fourth consecutive quarter of comparable-sales growth in those stores, and key departments outperformed the broader fleet by roughly 320 basis points. Because the locations and interventions were selected, this does not identify a single causal lever.29 McKinsey describes REWE's locally optimized planograms and Żabka's store-specific assortment system, and stresses that more localization can create complexity that must be economically justified.20 An independent working paper using 280 million shifts across 20 U.S. retail chains found that the scheduling characteristics associated with employee turnover varied across firms, regions and worker groups. It reinforces the danger of assuming that one operating intervention creates the same result everywhere.38
Candidate new belief. The perfect-store plan is a commercial hypothesis to be tested by store type and revised as conditions change. Management should preserve standards that must remain uniform while learning which assortment, display, promotion or shelf priority improves contribution in comparable local environments. The goal is not a unique playbook for every address, nor a lesson discovered in Kansas imposed universally on New York.
Economic consequence and buyer. A retailer's chief merchandising or operations executive, or a manufacturer's commercial/category leader, could allocate shelf space, field effort, inventory and promotion spend according to measured returns rather than compliance alone. Any improvement must exceed the cost of added variation, logistics and frontline complexity.
Who could pursue it. Trax / FORM is already especially close. Trax describes clustering stores by commercial drivers, changing shelf priorities by cluster and generating different action plans; its published cases claim economic improvement but are not independent trials.16 Repsly's UNREAL Candy case documents repeated promotion experiments and a 23% regional sales lift from a selected holiday program; it does not establish indefinite store-specific adaptation.17 YOOBIC brings contextual recommendations into frontline tasks. The opportunity cannot be claimed simply by adding “intelligence” to the hero.
What would validate an opening. Repeated matched-store or cluster tests showing incremental contribution after inventory, promotion, labor and merchandising costs; an observable process that revises the appropriate local playbook rather than forcing one winning tactic everywhere; and customers selecting one provider for a distinctive commercial answer over other capable platforms.
Who can persuade a commercial leader to treat the perfect store as a continually tested business decision—not a permanent compliance score?
What if execution fails because the deal does not work?
Inherited belief. Headquarters sets a commercially sound standard; a failure to comply is principally a problem of communication, field supervision, training, accountability or task completion.
Emerging business reality. Nike CEO Elliott Hill described how Nike's direct-channel strategy, discounting and declining investment in its brand had weakened wholesale relationships. He committed to restoring mutually profitable growth and product demand with retailers—not auditing their shelves more intensely.12 McDonald's September 2026 NEXT program explicitly makes franchisee capacity and payback part of the rollout decision. It plans about $8.5 billion in partner support through 2036, and says markets will determine their paths based on local insights and investment returns.13 These are executive business decisions, not evidence that a particular missed in-store task was caused by a bad agreement.
Candidate new belief. The job is not merely to drive compliance; it is to make commercial plans executable by the people and companies expected to carry them out. In a wholesale or franchise network, standards, capital, economics, authority, incentives, product availability and local relevance are part of the execution system.
Economic consequence and buyer. For a franchisor or manufacturer, the issue may concern franchisee returns, shelf access, distribution, sales and relationship durability. The accountable buyer could be a chief commercial officer, franchise president or CFO—someone able to change terms, investments or policy. That expands the conversation beyond the retail-operations software budget, but may also remove much of the remedy from the vendor's control.
Who could pursue it. FranConnect already joins unit performance, financial and operational signals with business plans and coaching.10 Trax provides evidence useful in commercial negotiations; YOOBIC helps operational teams prioritize support; Opterus coordinates partner communications. None of those capabilities proves a provider can diagnose or renegotiate the underlying economics. There may be a valuable services-led advisory opportunity before there is a new software category.
What would validate an opening. Cases where a recurring compliance or availability failure was traced to financial incentives, insufficient capacity, conflicting ownership or unworkable terms, then measurably improved when those conditions—not simply supervision—changed. Also needed: evidence that a particular provider can reliably identify the driver and is preferred by executives empowered to act.
When the store isn't following the plan, who can tell whether the store needs to change—or whether the economics of the plan do?
What if the local exception is evidence that headquarters should learn?
Inherited belief. Better central visibility, forecasting and instructions reduce local inconsistency. The store's role is to follow the plan, document exceptions and complete corrective tasks.
Emerging business reality. A 2026 study by researchers Caleb Kwon, Ananth Raman and Jorge Tamayo examined algorithm-generated labor schedules in approximately 500 grocery stores and 46 million shifts. Store managers overrode 72.9% of algorithm-generated shifts. Its instrumental-variable analysis, which depends on the validity of its identification assumptions, estimated a 5.7% labor-productivity increase from a one-standard-deviation increase in overrides—even though the resulting schedules were less aligned with predicted demand. This suggests local managerial information can improve central optimization; it does not mean every override is beneficial.37
Independent qualitative research published in February 2026 also records retail staff contending with faulty digital data, unforeseen customer interruptions, competing assignments and staffing constraints. It documents the context software instructions can miss, without establishing how prevalent those problems are across all retailers.31 Starbucks CEO Brian Niccol separately reported improvements in transactions, sales and service times where more employee hours were placed at busy times, illustrating that changing operating capacity can matter as much as assigning work more efficiently.32
Candidate new belief. The store should not simply be a monitored endpoint. Its judgment, overrides and explanations can be evidence that improves future rules and priorities—for that store or the comparable cluster where the lesson applies. The strategic move is governed two-way learning: understand why the local exception worked, validate the result and decide who can revise the central plan. Unrestricted local autonomy is not the answer either. An adjacent 2026 randomized study in smart vending shows why discretion needs a design: unconstrained worker overrides reduced inventory but also reduced sales, while a constrained override policy lowered inventory without the same sales penalty. The application is not conventional store execution, and the paper is a preprint, but the result supports governed local input rather than unrestricted autonomy.39
Economic consequence and buyer. A retailer COO or head of field operations might reduce wasted labor, mistaken replenishment, lost sales and unnecessary tasks. More importantly, management can distinguish a store breaking a good rule from a store revealing a bad one. The expected benefit must offset evaluation and governance costs.
Who could pursue it. YOOBIC's Store Manager Copilot reads POS, stock, footfall and execution signals, then recommends actions while deliberately leaving judgment with the manager.34 FORM / Trax describes intelligence flowing upstream into category and trade planning.35 FranConnect supports network performance investigation. The still-unproven distinction is whether local exceptions systematically change headquarters' future decisions, rather than only help clear the current task queue.
What would validate an opening. A traceable series of local challenges or overrides that led to changes in the correct store or cluster's plan, forecast, labor allocation or operating policy; counterfactual or matched-unit evidence that those changes worked; clear approval rights; and evidence that the learning reduces frontline burden instead of adding another layer of reporting.
Who will turn store-level exceptions from deviations to be corrected into knowledge that makes the right future plan better?
More store evidence is not automatically a learning advantage.
FORM / Trax emphasizes the scale of its image-recognition data and its shared models, which can improve SKU identification across customers; Trax says it analyzes more than four billion items annually. That could be a genuine advantage for observation. A separate question is whether added history improves the ability to choose which commercial intervention is most valuable in a particular store context.2235
The data may also be incomplete or misleading: better sales can reflect price, local demand, labor, competitor conditions or seasonality, and a photograph cannot capture all of them. The durable moat, if one emerges, would be longitudinal evidence connecting context → decision → execution → incremental economic outcome → revised plan, together with legal access to reuse and learn from that evidence. Proprietary image volume alone does not establish that moat.
Proof required. Show that the system's recommendations improve against fresh stores or future periods, outperform conventional category management and can be reproduced without heavy bespoke consulting. Test transferability across different retail formats and ownership arrangements. Keep the buyer's right to the data, the vendor's rights and the provenance of inferred effects distinct.
Why ordinary execution still matters
An independent 2024 analysis of approximately 15,000 promotional displays across roughly 5,000 stores found that 29% never reached the sales floor; the authors measured real sales benefits from displays that were placed. It would be a mistake to dismiss compliance as obsolete.33
Nor should every store be free to ignore regulatory, safety, contractual or brand standards. A centrally designed plan may be perfectly sensible and still require better execution. The seams are useful precisely because they separate three diagnoses that an apparent execution gap does not settle by itself.
For G2G, the next work is not to crown the smartest platform. It is to identify where one of these alternative beliefs changes a costly executive decision, where a vendor can credibly provide the answer, and where its product, narrative and proof give the customer a reason to choose it. Until then these remain candidate seams, not validated openings.