Master data poverty is a disease - and most companies don’t know they have it

Many companies buy a Ferrari. But all they have is a field to drive it on.

This image stayed with me after a conversation with Jonas Rashedi on the podcast MY DATA IS BETTER THAN YOURS – and it captures a point I have been discussing with clients for years, but have rarely been able to express it precisely.

Right now, every company is chasing AI use cases. New models, new tools, new promises. But AI needs three things to actually work: Context. Structure. A place where both can live.

Without that, you have a Ferrari – and no asphalt.

‍

The Real Problem Is Not AI. It Is Master Data Poverty.

I have been talking to companies about their data and technology strategies for more than 25 years. And one thing has not fundamentally changed during that time: most companies underestimate the true state of their master data.

Master data lives in spreadsheets. Coordination happens by phone calls, hallway diplomacy, and endless email chains. Product information is spread across six different systems, and no one knows exactly which one is telling the truth.

I call this master data poverty. It is not an IT weakness. It is a strategic risk. And it is the real reason why AI projects fail, not because the model is wrong, but because the foundation is missing.

‍

What a Modern DAM Needs to Deliver and What It Is Not

A Digital Asset Management system is not an image archive. It may have been ten years ago. Today, it is the foundation of content value creation.

What exactly do I mean by that?

The most impressive example I know is OTTO. The company evaluates more than 23 million assets and images using multimodal analysis across several language models. The cost per image is less than one cent. This is not a coincidence, nor is it the result of a particularly clever AI model. It is the result of having the right foundation in place: 67 million relationships, 16 billion attributes, clean master data, and a system that brings all relevant information together in one place and makes it accessible to AI.

Of course, OTTO could build the technical solution itself. What OTTO cannot build on its own is everything that comes before and after it. Before a classifier can do its job, it needs a structured data foundation and a flexible data model that manages 23 million assets, 67 million relationships, and 16 billion attributes in a way that allows an event trigger to know precisely which asset needs to be classified and when. Afterwards, the scores must be written back as structured metadata, immediately and across all channels, without a batch job and without a polling loop.

That is not an API call. It is an infrastructure decision.

What makes this example so compelling to me is that it shows what becomes possible when you build the infrastructure first. Not when you buy the tool. When you build the infrastructure.

And this is not an isolated case. Beiersdorf has been a customer for twelve years, not because it has no other choice, but because the foundation works. Weidmüller, with annual revenue of around one billion euros, needed five months for implementation and now uses the system for all of its product data worldwide. Siemens completed a global rollout covering more than one million products in nine months. IFM operates 48 of its own online stores and integrates data in real time using event stream technology.

What these companies have in common: none of them started with the AI project. They started with the data foundation.

The pattern is the same everywhere: the companies that successfully scale AI in production are the ones that built their data foundation first

‍

Where the Journey Is Headed

We are in the middle of a paradigm shift. DAM is becoming a knowledge graph, a system that not only manages assets, but also connects knowledge, maps relationships, and provides AI agents with a structured foundation on which they can work.

This isn't a vision. It's already happening with our customers today.

The companies investing now in clean master data, clear structures, and a foundation capable of supporting AI will no longer have to explain in three years why their AI projects are delivering results. The others will still be stuck in the field.

‍

Listen to the Conversation

Jonas Rashedi and I discussed all of this in this episode of MY DATA IS BETTER THAN YOURS, openly, concretely, and with real figures from real projects.

‍

👉 Listen to the podcast in German now

Apple Podcasts

Spotify

YouTube

We talk about the OTTO case, Tesa and Siemens, the build versus buy question, and where DAM is headed as a category. It is a 55-minute conversation that is well worth your time.

‍

‍

‍