Edcen Data All articles
Data Governance

The Answer Engine: Why UK Data Catalogues Need to Start With Questions, Not Tables

Edcen Data
The Answer Engine: Why UK Data Catalogues Need to Start With Questions, Not Tables

The catalogue as shelfware

A familiar pattern has emerged in UK mid-market organisations. A data catalogue is procured, populated and launched. It lists tables, columns, owners and lineage. It may even win praise in a steering committee. Six months later, analysts are still fielding the same requests by email, finance keeps a private spreadsheet of certified figures, and compliance maintains its own map of evidence sources. The catalogue exists, but it is not where work happens.

The problem is rarely the software. It is the design brief. Too many catalogues are built as inventories of the data estate rather than directories for the people who must make decisions, meet obligations and answer questions. In 2026, with budgets under scrutiny and AI pilots demanding trustworthy inputs, that distinction matters. A catalogue that documents ten thousand tables but does not shorten the path to a trusted answer is an expensive record of what exists, not a working asset.

Why table-first inventories stall

Technical metadata is necessary but not sufficient. Lineage diagrams help engineers assess change impact. Ownership fields help governance teams assign accountability. Yet a finance director does not wake up needing a table name. They need to know which revenue definition is sanctioned, whether a regional variance is genuine, and who can approve a restatement. An operations manager needs to know which supplier data is fit for a service-level review. A compliance officer needs to locate the evidence behind a regulatory return without commissioning a hunt.

When the catalogue answers only technical questions, people route around it. Shadow spreadsheets reappear. The catalogue becomes a compliance artefact maintained for audit rather than a tool used in the flow of work. The remedy is not more metadata. It is a different organising principle.

Three anchors: questions, obligations, decision rights

A useful catalogue should be indexed by the things people actually need to do. The first anchor is the recurring business question. Rather than starting with the dataset, start with the twenty or thirty questions that recur across finance, operations and commercial teams. For each, record the trusted data sources, the certified metric, the known caveats, the business owner and the route to request access or challenge a definition. The catalogue then becomes an answer engine, not a card index.

The second anchor is the regulatory obligation. UK organisations face a range of duties, from UK GDPR records and subject-access handling to sector rules such as FCA Consumer Duty or operational resilience requirements. A catalogue designed around obligations should show, for each duty, the evidence sources, the control owner, the review cycle and the reports in which the evidence appears. That makes compliance less dependent on individual memory and more resilient to staff turnover.

The third anchor is decision rights. A catalogue entry that names an owner but not the limits of their authority is only half-governed. Each certified metric or critical dataset should show who can change the definition, who approves access, who signs off a restatement and who is accountable when quality falls short. These rights should be visible in business language. When decision rights are explicit, the catalogue starts to govern behaviour rather than simply describe it.

A 2026 review checklist

A practical review should be run jointly by finance, operations and compliance, not left to the data team alone. The following checks are a starting point.

  1. List the twenty most common business questions in the organisation. For each, confirm whether the catalogue provides a trusted route to an answer within three clicks or fewer.
  2. Select five regulatory obligations that matter to the business. For each, confirm that compliance can locate the evidence source, control owner and review date without raising a request.
  3. Review every certified metric. Confirm that it has a named business owner, a definition in plain English and a named change approver.
  4. Test ownership. Ask a sample of owners to confirm their responsibilities within the last ninety days. If confirmations are stale, the catalogue is drifting.
  5. Check that lineage is presented in business terms. A change to a source system should show which reports, obligations and decisions are affected, not only which tables are downstream.
  6. Review search terms. If users must know technical names to find anything, the business layer is underdeveloped.
  7. Establish a time-to-answer baseline. Sample five real questions, measure how long they take to resolve today, including the human chasing, and repeat after the catalogue is redesigned.
  8. Retire inactive entries. A catalogue that preserves every historical table without marking status becomes a museum. Mark deprecated assets clearly and remove them from default search results.

Measuring what matters: time-to-answer

The most useful metric for a catalogue is not the number of assets documented. It is the time it takes to reach a trusted answer. Time-to-answer should be measured against real questions drawn from finance close, operational performance reviews, regulatory reporting and customer queries. Include the elapsed time from question to answer, not just the time spent in the tool. If a user finds a dataset in two minutes but then spends two days confirming whether it is the right one, the catalogue has not solved the problem.

A quarterly sample of five to ten questions is enough to establish a trend. Pair it with measures such as the proportion of questions answered without escalation, the reuse rate of certified metrics and the number of stale ownership confirmations. These measures connect the catalogue to business performance. They also give the board a clearer answer when asked whether the investment is working.

Governance that earns its keep

A business-facing catalogue still needs maintenance. Assign stewardship for the business glossary. Link entries to access requests and change control so that governance happens in the same place people look for answers. Use automated metadata collection to reduce manual effort, but invest human attention in the business layer: definitions, caveats, obligations and decision rights. Start with a manageable set of questions rather than attempting to catalogue the entire estate at once. Mid-market organisations do not need a large central team to make this work; they need clear ownership and a ruthless focus on use.

The catalogue should be a directory, not a document store. In 2026, UK businesses need evidence that their data investments are improving decisions and reducing risk. The test is simple. Can a finance, operations or compliance professional use the catalogue to answer a real question or meet a real obligation faster than before? If not, the catalogue needs rebuilding around the work, not the tables.

All Articles

Related Articles

The Kilowatt Ledger: Bringing Energy into UK Analytics Governance

The Kilowatt Ledger: Bringing Energy into UK Analytics Governance

The Evidence File: What UK Boards Should See Before an AI Pilot Touches Customer Records

The Evidence File: What UK Boards Should See Before an AI Pilot Touches Customer Records

Walking-Away Rights: A Board-Level Test for Analytics Vendor Portability

Walking-Away Rights: A Board-Level Test for Analytics Vendor Portability