From having data to using data

Many organisations have more data than ever, yet a straightforward management question can still take weeks to answer. Departments use different definitions, reports contradict one another and dashboards mostly show what has already happened. The problem is rarely a shortage of charts. What is missing is a connection between data, decisions and day-to-day action.

Data-driven working means that relevant information contributes to a concrete choice at the right moment. Experience and professional judgement do not disappear. Data makes assumptions visible, helps compare alternatives and allows the organisation to learn from the effect of a decision.

Start with the decision, not the dashboard

A dashboard without a decision question quickly becomes a collection of indicators. Reverse the sequence. First define which decision needs to be better or faster. Who makes it, how often, what courses of action are available and when is more information needed?

A municipal team may want to determine which cases need additional attention. A service provider may want an earlier warning when lead times increase. Only once the desired action is clear can teams determine which data, timeliness and quality are required. This prevents months of work on information that is interesting but changes no behaviour.

Why initiatives often stall

• Fragmented accountability: IT manages systems, analysts produce reports and the business expects someone else to safeguard quality.
• Competing definitions: terms such as customer, incident, waiting time or outcome mean different things across departments.
• Invisible lineage: users do not know the source of a figure, which transformations were applied or how current it is.
• Too many indicators: dozens of KPIs compete for attention without making clear when someone should act.
• No learning cycle: a decision is taken, but no one later examines whether it had the expected effect.

Data governance that helps the operation

Data governance need not become a central gate that delays every change. It should create clarity where collaboration would otherwise break down.

A shared language

For critical terms, record the definition, owner, source and calculation. Start with terms used in management information or public accountability. A compact, maintained glossary is more useful than a large catalogue no one keeps current.

Visible ownership

A data owner is accountable for meaning and use; a data steward supports daily quality and coordination. IT provides reliable processing and availability. These roles work only when they have authority, time and a clear escalation route.

Measurable data quality

'The data must be good' is not a testable requirement. Define which dimensions matter for each use: completeness, accuracy, timeliness, consistency or uniqueness. A monthly report can tolerate different thresholds from an operational alert that should trigger action within minutes.

Transparency about uncertainty

No dataset is completely neutral. Selections, missing values and definitions affect the outcome. Show ranges, exceptions and known limitations alongside the figure. This increases trust because users understand what they can and cannot conclude.

Turn every data product into a working process

A useful data product needs more than an owner and a technical platform. It needs a repeatable cycle:

1. formulate the decision question and desired effect;
2. select the minimum data required;
3. discuss insights with people who know the context;
4. record the chosen action and assumptions;
5. measure the effect and adjust the question, process or data.

The final step is what makes an organisation data-driven. Without feedback, data mainly produces a more convincing opinion; with feedback, it creates learning.

The human side: from blame to inquiry

Employees will not surface problems when figures are used only to assign blame. Leaders must set the example. They ask about definitions and uncertainty, welcome information that contradicts expectations and distinguish a signal from a judgement.

Invest in data literacy that fits each role. Not everyone needs to code analyses. A manager should be able to challenge a chart; a process owner should recognise quality problems; an analyst should understand which decision an analysis is intended to support.

A practical first 100 days

• Choose one valuable decision. Select an issue with a visible owner, sufficient data and room to act.
• Map the chain. Follow core data from source and definition through to reporting and decision.
• Make quality visible. Measure a few relevant dimensions and address the largest root cause structurally.
• Change the meeting. Centre the decision question, deviations, actions and evaluation instead of reviewing every number.
• Repeat and scale. Record what worked and apply the same pattern to the next issue.

The foundation for responsible AI

Data-driven working is also a prerequisite for reliable AI. A model cannot make weaknesses in definitions, sources and processes disappear; it can reproduce them at scale. Organisations that know where data comes from, who is accountable and how quality is measured are better able to assess and monitor AI.

The principle is simple: do not begin by collecting more data; organise better decisions. Continuous Connect helps organisations make that connection — from definitions and ownership to data quality, management information and a functioning learning cycle.

← Back to all blogs