How Retailers Are Using MDM to Power AI and Personalisation

Everyone is being sold AI right now. Demand forecasting tools. Personalisation engines. Intelligent replenishment. Automated product descriptions. The retail technology market is awash with vendors promising that their AI solution will transform your customer experience, optimise your margins and put you ahead of the competition.

Most of it works… eventually. But here’s the thing that doesn’t make it into the sales pitch: AI is only as good as the data it runs on. And in most retail businesses, that data is nowhere near ready.

This isn’t a reason to dismiss AI. It’s a reason to understand what actually needs to happen before AI delivers what it promises. And at the heart of that is master data.

The gap between AI ambition and AI reality in retail

Gartner estimates that through 2025 and beyond, 80% of organisations seeking to scale digital business will fail because they don’t take a modern approach to data governance. That statistic gets quoted a lot. It’s worth sitting with for a moment.

Four in five businesses investing in digital transformation — including AI — will not achieve what they set out to achieve. And the primary reason isn’t the technology. It’s the data underpinning it.

We see this directly in retail. A business invests in an AI-powered personalisation platform. It connects to their CRM, their ecommerce platform, their loyalty scheme. The vendor runs the first campaign. The results are underwhelming. The personalisation doesn’t feel personal. The recommendations are generic. The forecasts are off.

The diagnosis, almost every time, is the same: the customer data is full of duplicates, the product data doesn’t match across systems, and nobody is quite sure which version of the truth the AI is learning from. Garbage in, garbage out — it’s one of the oldest principles in computing, and AI hasn’t changed it. If anything, AI amplifies it. A poor recommendation engine running on clean data will produce mediocre results. The same engine running on inconsistent data produces results that actively damage customer trust.

What AI actually needs from your master data

To understand why retail master data management is the foundation for AI, it helps to be specific about what AI use cases actually require from your data.

Personalisation depends on a unified customer record. Your AI personalisation engine needs to know that the customer who bought online last Tuesday is the same person who returned a product in-store last month and who called your contact centre the week before. If those three interactions are stored against three different customer records — which is common in retail businesses that have grown through acquisition or channel expansion — the AI sees three different people and personalises for none of them properly.

Demand forecasting depends on accurate, consistent product data. Forecasting models learn from historical sales data. If the same product has been recorded under different material codes at different points in time, or if product attributes like unit of measure or product hierarchy have been inconsistently applied, the model is learning from a corrupted history. The forecast it produces will reflect the inconsistency in the data, not the actual demand pattern.

AI-driven product recommendations depend on clean product taxonomy and attributes. For an AI to recommend “customers who bought this also bought that”, it needs to understand what products actually are — their category, their attributes, their relationships to other products. If your product master data is incomplete or inconsistently classified, the recommendation logic has nothing solid to work from.

Automated supplier management — using AI to flag anomalies in supplier lead times, pricing or quality — depends on a clean, complete vendor master. If your supplier data is fragmented across different systems with different naming conventions, the AI cannot build the baseline it needs to spot when something is wrong.

In every case, the AI use case is real and the technology is capable. The limiting factor is the master data it runs on.

How retail MDM creates the foundation for AI

Retail master data management is the practice of governing the three core data domains that every retail business depends on — product, customer and supplier — so that the data in those domains is accurate, complete, consistent and trusted across every system in the business.

When that foundation is in place, AI stops being a disappointment and starts delivering.

A unified customer master means your personalisation engine sees one complete view of each customer across every channel. Every interaction, every purchase, every preference — attributed to the right person and available to inform the next recommendation or offer.

A governed product master means your forecasting model learns from clean, consistent historical data. Product hierarchies are applied correctly. Attributes are complete. The data the AI trains on actually reflects reality.

A complete vendor master means your supplier AI tools have the baseline they need to monitor performance, flag anomalies and automate routine supplier communications without producing false positives from data inconsistencies.

This is why WaterWipes, when they implemented Maextro and built their data governance foundation on SAP BTP, specifically designed it with future AI and automation use cases in mind. They weren’t building a data project. They were building the infrastructure that would make every future digital initiative — including AI — work properly. The governance foundation came first. The AI ambitions were built on top of it.

That sequencing is what separates retailers who get genuine, lasting value from AI investment from those who keep finding that the technology underperforms.

The personalisation problem is a data problem

Let’s get specific about personalisation, because it’s where the gap between expectation and reality is most visible to customers — and therefore most costly.

Personalisation at scale in retail requires three things to work in concert. You need to know who the customer is across every channel. You need to know what products are available and how they relate to each other. And you need to know the history of how this customer has interacted with those products.

The first two are master data problems. The third is a transaction data problem that depends on the first two being solved.

Retailers who have invested heavily in personalisation technology but haven’t governed their customer and product master data typically end up in the same place: campaigns that feel generic, recommendation widgets that surface irrelevant products, loyalty schemes that don’t recognise customers across channels. These aren’t failures of the personalisation technology. They’re failures of the data foundation underneath it.

Fixing the data foundation — establishing a single, governed customer record and a clean, attributed product master — is the unlock. Once that’s in place, the personalisation technology can do what it was designed to do.

AI readiness is a process, not a purchase

One of the most important mindset shifts in retail right now is moving from treating AI readiness as something you buy to treating it as something you build.

You cannot purchase your way to AI readiness. You can buy a personalisation platform, a forecasting tool, a recommendation engine. But you cannot buy clean, consistent, governed master data. That has to be built through a disciplined approach to data management — defining ownership, establishing governance processes, implementing the right tooling, and maintaining data quality as an ongoing operational discipline rather than a one-off project.

The retailers making the most genuine progress with AI are the ones who recognised this early. They invested in getting their master data right before they invested in the AI tools that would consume it. The sequencing felt slow at the time. In practice it was the fastest route to AI that actually works.

What this looks like in practice

The typical journey for a retailer building an AI-ready data foundation looks something like this.

The starting point is an honest audit of the current state. Where are the duplicates in your customer data? How complete is your product attribute data? Are your supplier records consistent across your ERP and procurement systems? This doesn’t need to be an exhaustive exercise — even a targeted assessment of your highest-priority data domains will surface the most significant gaps quickly.

From there, the focus is on establishing governance for the domains that matter most to your AI use cases. If personalisation is the priority, customer master data governance comes first. If demand forecasting is the goal, product data. If supplier AI tools are on the roadmap, vendor master.

Governance means defining who owns each data domain, what the quality standards are, how new records are created and validated, and how ongoing quality is monitored and maintained. Tooling enforces those standards and automates the routine parts of the process so that governance doesn’t become a bottleneck.

Once the foundation is stable, AI tools layered on top of it perform as advertised. Personalisation becomes genuinely personal. Forecasts become reliable enough to act on. Supplier anomalies get flagged before they become supply chain problems. The AI investment that felt like it should be working but wasn’t starts delivering what was promised in the first place.

The bottom line

AI is not going to wait for retailers who aren’t ready. The competitive pressure to deploy personalisation, forecasting and automation tools is real and it is only going to increase.

But deploying those tools on top of inconsistent, ungoverned master data is not a shortcut. It’s a longer route to the same destination — via a lot of disappointing results, wasted budget and eroded internal confidence in the technology.

The retailers who will get the most from AI are the ones who treat retail master data management as the foundation it actually is. Not a prerequisite to be endured before the interesting work begins, but the investment that makes every other digital initiative work better.

If you want to understand how Maextro provides the retail master data management foundation that makes AI and personalisation work in practice, we’d be glad to show you what that looks like for a business like yours.