The Data Iceberg

For most organisations, data management looks like an iceberg.

The few critical data elements an organisation manages are catalogued, trusted, and easy to reach. The rest is usually a mess, slapped together in a “miscellaneous” table. Out of sight and out of mind.

Just like an iceberg, only a fraction of that data is visible. What remains beneath the surface is dark data: data that is collected and stored but never used.

Dark data represents wasted potential. Data assets only generate value when they are put to use. To actually use data, you must be able to find it, trust it, and classify it. Otherwise, your assets turn into liabilities.

All Content, No Context

At school I just didn’t get Shakespeare. Macbeth wasn’t written to be read. It was written to be performed, for a seventeenth-century audience. The words survived. A lot of the context didn’t. I couldn’t get value from Macbeth because I didn’t understand the jokes, the cultural references, or even the vocabulary. It was only once I was equipped with translations, definitions and contextual notes that I began to get it.

Just like a modern audience reading Shakespeare, nothing can be understood in isolation from its context. This is increasingly important in a world of misleading statistics, clickbait headlines, and short-form video.

Everyone now complains, “that was taken out of context!”

But AI Will Solve This, Right?

AI certainly makes data more findable. But surfacing data isn’t the same as understanding it or trusting it.

AI will likely only accelerate the gap between data and its context. Just as text separated Shakespeare’s words from the world it was created in, modern technologies have continued to create new mediums that widen this gap.

A joke within the running community is that if you didn’t record it on Strava it didn’t happen. Well, the same can be said for AI. For as far as its concerned if the input data doesn’t exist, then it simply doesn’t exist. This is why AI experts stress “context is key”, that is context that is readable to a machine.

Metadata: The Data We Forgot To Collect

It is ironic that in the age of “Big Data”, metadata is the data we forgot to collect.
Weak metadata management is one of the key findings from Robinson Ryan’s 2026 State of the Data Management Industry report: only 16% of organisations said most of their data was organised with clear definitions, none claimed to have catalogued everything, and two-thirds had minimal or no ability to trace their data’s lineage.
At its most fundamental, metadata is the who, what, when, where, and how of your data. It lets organisations process, maintain, integrate, secure, audit, and govern their data. Without it, an organisation cannot truly understand its own data, systems, or workflows.

Imagine you ask your AI assistant a question. It responds with an answer, but without metadata it can’t tell you where that answer came from. It could be the master record or a draft from a graduate ten years ago. Neither you nor the AI would know.
Failing to cite your sources reliably wouldn’t pass a high-school assessment, so why should data driven organisations have lower standards?

Technique, Not Tech

The answer isn’t to reach for new technology. It’s to reach into the data management professional’s toolbox and find the right technique for the job. Three do most of the work: cataloguing, lineage, and classification.

Cataloguing answers what: what the data is and what it means.

Lineage answers where, when, and how: where it came from and where it lives, how it was produced and transformed on its journey, when it was captured, whether it’s still current, and when it expires.

Classification answers who: who owns the data and who is allowed to use it.

Mature metadata management is the difference between an iceberg and a data asset.