The gap between retailers that can act on their data and those that cannot is widening, and most organisations are only beginning to understand why. A variety of teams, from merchandising, to finance, to marketing, are working from different definitions of their best customer, and AI initiatives are stalling before they reach production.

The cause is industry-wide: data estates that have grown fragmented, ungoverned, and expensive to maintain over decades of departments solving their own problems independently. In one major retailer CI&T worked with, that fragmentation ran so deep that a business user accidentally entered a 9-digit product identifier into a unit cost field. Without any validation checks, the product was recorded as costing millions rather than pounds, significantly distorting financial reporting before anyone caught the mistake.

This article examines how retailers got here, what it is actually costing them, and why the case for fixing it has never been stronger.

How reporting became so complex

Large retailers do not set out to build separate reporting structures. The fragmentation builds up over the years, department by department, driven by two very human problems.

"The first is a lack of trust,” says Abhay Bagai, Head of Data & AI at CI&T. “A finance team starts using a particular reporting tool, and then when marketing wants to use it too, they don't trust what the finance team has built. So they create their own visualisations for their own purposes.
“The second is the bottleneck. If someone needs certain numbers to run their business but it will take weeks for the central IT team to build the report, they'll just do it themselves in Excel, or a tool closest to hand."

Multiplied across every department and every year, the result is dozens of tools, thousands of reports, and no single view of the business. A major retailer CI&T worked with was operating more than 20,000 reports across multiple business functions. Only around 23% were actively used, yet the organisation was still paying to maintain and license all of them.

What fragmented data is actually costing you

The direct costs are straightforward. Every reporting tool carries a licensing fee, and organisations running four or five parallel platforms pay for all of them. Each also requires specialist skills to maintain, pulling talent budgets across multiple niche technologies instead of building concentrated expertise.

The indirect costs are harder to quantify but arguably more damaging. When departments work with disparate datasets and different definitions, they can actively undermine each other.

"Different departments end up with a different view of what the truth is," says Bagai. "A finance team might have a completely different picture of what a high-value customer looks like compared with the marketing team. Often, marketing’s primary goal will be to acquire more customers and/or to grow basket size. Meanwhile, finance will say we don't want that customer because, in the long term, they cost us money due to increasing return costs for those customers. It just creates a mess."

Shadow IT compounds things further. Business teams that export data from one system, manipulate it in a spreadsheet, and upload it to another tool introduce error risk at every step.

Why AI is only as good as what sits beneath it

Cleaning up that kind of error is fixable. What is harder to fix is an entire data estate that has accumulated the same structural weaknesses across thousands of reports. That is the foundation on which many retailers are finding themselves now.

Melissa Minkow, Global Director of Retail Strategy and Insights at CI&T, says: "Data is the key building block for operational excellence, improvement to the customer experience, and future AI enablement. There is no point in having data if it is not actually accurate and un-usable."

The pressure extends beyond current AI applications. Minkow points to emerging requirements around digital product passports and supply chain transparency as areas where data accuracy and traceability will become non-negotiable. Retailers under board pressure to have an answer on agentic commerce face the same challenge. When asked to rate the quality of their core product data out of ten, many retailers C&IT encounter say three or four. Their data foundations are not ready, and many find themselves behind on multiple fronts simultaneously.

"The shiny thing on the agenda is fine as an aspiration," says Bagai. "But if you don't address the underlying data challenges, you are never going to get there. The businesses that have invested in the more foundational work, the less exciting but necessary stuff, are the ones that will activate those experiences quickly. We will see a much bigger gap appear between the companies that made those smarter investments and the ones that have just been chasing the next shiny thing."

The challenge now is not to move faster towards the next capability, but to build the infrastructure that makes scaling possible in the first place.

Why deferring is no longer an option

Boards are questioning the returns on previous technology investments, and an organisation running multiple overlapping data platforms is carrying a cost that could be redirected towards growth at a moment when every pound counts. For many retailers, that commercial pressure is now converging with a hard deadline: legacy reporting platforms are reaching end of life, and the question of migration has moved from strategic consideration to operational necessity.

The reason it was deferred for so long is understandable. Rationalisation projects - an AI-driven process that analyses and collapses thousands of duplicate legacy reports into a single, AI-ready data foundation - were historically slow and expensive, designed to take in a few hundred reports. At ten or twenty times that scale, the effort felt disproportionate to the return. What has changed is that it no longer takes that long.

Starting with manual analysis of a legacy tech stack where there is lots of reliance on key SME's, CI&T deploys agents that analyse the existing tech stack rapidly and share insights on its actual usage. The outputs are objective in a way that interviews cannot be: rather than asking a business user whether their report matters and receiving an answer shaped by habit, the analysis reads the metadata directly.

At the major fashion retailer, AI analysis of the 27,000-report estate revealed that a large proportion had not been accessed in six months, and identified 40% more scope than two previous partners had found. A programme of that scale would typically take over twelve months to deliver. CI&T did it in five, precisely because the approach started with evidence rather than assumption.

The retailers who move first will be hardest to catch

AI is reshaping how consumers discover, evaluate, and purchase products, and the retailers best positioned to capitalise on that shift are not necessarily the ones with the most ambitious AI strategies. They are the ones who invested in getting their data right first.

"In a really tough consumer market, the companies that are going to win are those that know which bets to make, investing in data foundations,” says Bagai. "The retailers who will define the next decade are already scrutinising their data strategy and acting on it."

---

Want to learn more about how CI&T can help you harness your business's potential? Get in touch!