The Five Columns That Matter

Most data strategies fail for the same reason cities go bankrupt: they try to govern everything.

Somewhere, a well-meaning data owner is responsible for 4,000 attributes. Each has a definition, lineage and quality score. The effort is impressive. The economics are less clear.

Finance offers a useful clue.

Every accounting system in the world starts in exactly the same place: account, date, detail, amount and debit/credit.

Five columns.

That is the remarkable part.

From those five columns emerge financial statements, management reports, regulatory filings and audit opinions. Entire finance functions are built upon them.

Yet when organisations think about data, they often assume importance is proportional to quantity. The more data they have, the more data they feel compelled to govern.

The result is predictable. Governance expands faster than value.

The mistake is thinking governance is about completeness. It is not. It is about economics.

Every control, definition and quality check carries a cost. The relevant question is simple: does governing this data improve decisions enough to justify the investment?

Data should be governed in proportion to the economic value it creates.

The Same Data, Wearing Different Clothes

Imagine walking through a large company. A CFO is reviewing a sales dashboard. A salesperson is working through a list of prospects. A customer is browsing a website, deciding whether to buy. These appear to be different activities. Different people. Different tools. Different products. But there is a good chance

The Unit of Execution

As organisations grow, getting things done often becomes harder, not easier. The problem is rarely a shortage of talent. It is that talented people become separated by reporting lines, functional boundaries, and competing priorities. Work slows as coordination expands. The most effective teams organise around an outcome rather than an

The Data Quality Paradox

Data quality has an unusual characteristic: success removes the evidence of the problem. At first, the issues are obvious. Reports disagree. Fields are empty. Three labels describe the same thing. The defects announce themselves. Then a good team goes to work. Definitions become consistent. Pipelines become reliable. Errors disappear. Complaints

The Cost of Learning Too Slowly

Imagine two companies making the same mistake. The first discovers it after three weeks. The second discovers it after three years. The mistake is identical. Yet one survives and the other may not. Why? Because mistakes have a peculiar property. They grow while nobody is looking at them. A product

Read more

The Same Data, Wearing Different Clothes

The Unit of Execution

The Data Quality Paradox

The Cost of Learning Too Slowly