The reality of data production : why data is never a standardized product

Turning data into business value is often harder than it seems. One reason is simple: data is not a standardized product. Data producers and data consumers have different objectives, which often creates friction. Here are four realities of data production that explain why.

1. Data is purpose-built

Unlike manufactured products, data is created to serve a specific operational purpose. Collection frequency, attributes, precision, retention and quality controls are all optimized for the producer’s primary use case.

As a result, it is rather difficult to retarget or recycle data for another use case, because it is not natively optimized for it. A dataset that perfectly answers one client’s needs may be completely unsuitable for another, and it often needs to be enriched with external sources.

2. Data is non-fungible

In economics, an asset is fungible when one unit can be replaced by another equivalent unit. A MICHELIN tire can be replaced by a Bridgestone or Continental tire that meets the same specifications. Data, however, is non-fungible.

Each dataset has unique characteristics and attributes, and entire machine learning and AI models are built on top of it. Change the data source and you need to change everything, even if the data seems very similar. Think of your email account: switching provider is technically possible, but the information, contacts, habits and services linked to it make the transition complicated.

For data providers, winning a customer can be a long and difficult process. But once trust is established and the data is integrated into the customer’s operations, the relationship can become very long-lasting.

3. Data is partial and fragmented

Every data producer sees only part of the world. An automaker sees its vehicles, a telecom provider its subscribers, an insurer its policyholders, a local government its territory and a fleet manager its customers. Each producer observes reality from a limited perspective and collects data with its own biases.

Take traffic data for example, covering several European countries does not mean you are representative in every one of them, and being representative somewhere does not guarantee geographical coverage. Our role is to transform raw data into actionable recommendations, which becomes difficult when one key piece of information is missing. Some business questions can be answered confidently, while others require additional data sources to fill the gaps.

4. Data sharing is a trust challenge

Data producers are rarely neutral actors: they compete for customers, market share, influence, innovation and profit. Yet to build meaningful datasets, with coverage, representativeness and compatible characteristics, they would need to put their data together. Organizations must weigh competition risks, intellectual property protection, confidentiality requirements and legal and regulatory constraints.

Every producer owns a piece of the puzzle, and each fear that sharing it will help competitors more than itself. The challenge is therefore not simply “Can we share the data?” but “Can we trust each other enough to share it?“

Looking to find out which data can really answer your business questions? Contact our team to understand how our solutions can help you.

Ready to get started ?