Mislabelled: Why retail calls its strategic failures execution problems

Do the conditions that made this solution work in its original context still exist in the context we are now proposing to invest into?

Before any retail technology investment is committed, that question belongs at the centre of the decision. In the four cases below, the conditions did not exist. Execution is where the failure became visible, not where it began.

The post-mortems of retail technology programmes share a vocabulary. Poor execution. Insufficient change management. Cultural resistance to adoption. They appear so consistently that the industry has stopped questioning the diagnosis. They describe what happened. They do not explain why.

Execution problems are real, but they are symptoms of the unexamined assumption rather than independent causes; the mislabelled decision produces the execution failure. The decision that matters, whether the conditions that made a solution work elsewhere still hold here, gets made by default the moment a programme is approved, simply by proceeding without checking it.

Coresight Research's 2026 in-store retailing study found store technology adoption close to universal, at 97% deployed or planned, while losses accelerate alongside it, a gap it attributes to systems going in before the operational foundation they depend on exists.¹ A June 2026 survey of more than 100 senior retail and brand leaders by PMC and Retail Economics found 35% citing return on investment and economic uncertainty as their leading challenge.² Retail is not underinvesting in technology. It is investing on the wrong basis.

A decade of mislabelled failures

Four cases, spanning more than a decade and differing in sector, geography and scale, share a point of origin. They are not a survey; what makes them useful is how little else they have in common.

Target Canada, 2013 to 2015. Target entered Canada by acquiring the leases of 133 Zellers locations, confident that a model proven across North America would cross a border between markets more alike than they appear. Product data inherited from the acquisition proved accurate roughly 30% of the time, against 98 to 99% in Target's US operations, causing cascading failures in stock availability and order routing.³ That was not a migration oversight but the consequence of assuming Canadian supplier systems would behave like their US counterparts. A former employee later told investigators there had been no discussion of accuracy during the rollout: “There was never any talk about accuracy.”³ 124 stores in under two years left no room to find the error before the investment was irreversible. Target exited in January 2015 on a $5.4 billion write-down and a net loss of approximately $2 billion.⁴

Bed Bath and Beyond, 2019 to 2023. Mark Tritton arrived as chief executive in November 2019 from Target, where owned brands had been a commercial success. He launched nine private-label brands between 2020 and 2022, displacing much of the national-brand assortment that defined the chain. Its customers, by the account of analysts at the time, came for national brands unavailable elsewhere, so removing them removed the reason to visit. Chapter 11 followed in April 2023.⁵ The decision was merchandising, not technology, and the logic was identical: a proven outcome moved to a different context without anyone testing whether the conditions behind it held.

A global prestige beauty retailer. An inventory visibility programme that worked across North America and Europe was extended to Asia-Pacific on that basis, without assessing whether the conditions behind it, among them consumer purchasing patterns and regulation, held there. In one territory, reconciling online and in-store product formats cost more than the programme's projected return. Nobody called it a failure. It continued and, by internal measures, succeeded. That is the version hardest to correct. A public failure gets investigated while a quiet shortfall gets absorbed into the budget and never looked at again.

Amazon Just Walk Out, 2016 to 2024. Amazon deployed its autonomous checkout first in Amazon Go convenience stores, where it performed as intended. The format suited the technology: a limited range and a controlled layout, shopped by people in a hurry. Amazon Fresh shared none of that. Grocery shoppers browse, reconsider and replace items in ways the system could not accommodate at scale. Amazon withdrew it from Fresh in April 2024. Reporting by The Information later found it relied on substantial numbers of remote human reviewers, a detail absent from the case for expansion; Amazon disputed that characterisation.⁶ The technology transferred cleanly. The assumption that grocery resembled convenience did not.

Three were written off publicly as execution failures. The beauty retailer's was not, because there was nothing to write off.

What connects them is not that each was a technology failure; Amazon's functioned as designed. It is one decision error, made at approval, that recurs regardless of which department owns it. Technology is the focus here because that is where the error is now most expensive, not because it belongs to technology alone.

The conditions that make this inevitable

These are not isolated misjudgments but a system working as designed. The conditions below come from direct observation across retail programs in North America, Europe and Asia-Pacific, not from literature.

By the time a program is visibly in trouble, capital is committed and vendors contracted. The organization can only execute better or exit. Every post-mortem inherits that constraint, examining what could still be changed rather than what was fixed at approval.

The approval process itself has nowhere to put the question. A program is reviewed by technology, finance, strategy and an executive committee chaired by the chief executive, separately or together, each through a different lens as they should: feasibility, return, strategic fit, and then, drawing on all three, whether to proceed. No layer is asked or equipped to judge whether the environment being invested into resembles the one where the solution worked. The question is not thought irrelevant. It has no owner.

Better governance would not fix it, because the decision is usually made before formal approval begins. Advisors are engaged and business cases constructed to support a conclusion already reached. The approval process then does what it was designed to do: it evaluates the case presented to it.

Four conditions produce the error. Each is sufficient alone. They rarely occur alone.

The reference point problem. The primary evidence is success elsewhere, a competitor's deployment or a pilot under different conditions. That is not irrational; it is the most available evidence. But it answers whether a solution can work, not whether it will work here, for this customer. The two questions look alike, and nothing in the calendar rewards separating them.

Budget cycle pressure. In most large retailers, allocations not deployed within the window are returned and rarely restored at the same level. A program needing three more months of validation risks its funding, so the incentive rewards deployment over rigor. That alone would be manageable. What makes it structural is that it operates inside an organization already competing with itself.

Internal competition. Store operations, e-commerce, supply chain and technology each hold a budget and compete for priority. A function that does not deploy risks losing influence to one that did, so investment gets weighed against the cost of not investing and rigor is the casualty. That reduces the incentive to ask. Distance decides whether a question, once asked, reaches the room.

Practitioner distance. The people who know the data infrastructure is incompatible or the timeline impossible sit several layers below the approval. That is true of any large hierarchy. What makes it specific here is what their knowledge would have supplied: not a warning about execution risk but a direct answer to the question the review never asks. Disagreement travels slowly upward where seniority is read as certainty.

The problem is not who is in the room. It is what the room was built to do.

AI adoption: The same logic, larger stakes

Retail is now committing to artificial intelligence at a scale exceeding any previous investment cycle. The reference points are familiar. The approval structures, budget cycles and competitive dynamics have not materially changed and the practitioner distance has not closed.

A June 2026 study of 200 retail decision-makers found 91% feel moderate or significant pressure to adopt AI, while 46% have a defined AI strategy backed by a clear value case.⁷ Global retail technology spending overall is forecast at $232.8bn in 2026, rising toward $287bn by 2029.⁸

The organizations making these allocations are largely the same ones that could not demonstrate return from the previous decade's programs. Those conditions were never addressed, only inherited. What changes with AI is not the nature of the failure mode but the cost of discovering it.

None of this argues for slower investment. It argues for giving that question an owner it does not have: before the business case is built, someone formally responsible for naming which conditions of that success are assumed to travel here, and confirming whether they do. That is a structural fix, not a courage problem, which is why it does not depend on the right person being in the room. Its absence, four times over, cost billions.

REFERENCES

1.    Coresight Research, “The State of In-Store Retailing 2026,” in partnership with Simbe and RELEX Solutions, May 2026.

2.    PMC and Retail Economics, “ROI and economic uncertainty top UK retailers’ digital transformation challenges,” Retail Technology Innovation Hub, June 1, 2026.

3.    Joe Castaldo, “The Last Days of Target Canada,” Canadian Business, 2015.

4.    Target Corporation, Form 10-K, fiscal year ended January 31, 2015, US Securities and Exchange Commission.

5.    Bed Bath & Beyond Chapter 11 filing, April 23, 2023, as reported by CNBC, NPR and Forbes.

6.    The Information, “What’s Behind Headlines About Amazon’s ‘Just Walk Out’ Feature,” April 2024; Amazon statement disputing the human-reviewer characterisation.

7.    HyperFinity, survey of 200 retail decision-makers, “Many retailers feel pressure to adopt AI but only a few have defined strategies,” Retail Technology Innovation Hub, June 16, 2026.

8.    Gartner, Forecast: Enterprise IT Spending for the Retail Market, Worldwide, 2023–2029, 3Q25 Update, November 2025. Proprietary; no public link.

The prestige beauty retailer case draws on direct observation across retail programs in North America, Europe and Asia-Pacific, and is unattributed by agreement.