Data and Control

Key figures only become controllable through decisions

Data only becomes a control instrument when it is derived from clear goals, collected according to consistent rules, interpreted in the right context, and linked to accountable decisions.

The topic area 'Data and Control' therefore not only deals with analytics or reporting. It organizes the entire chain:

• business question,
• Goal and hypothesis,
• Measurement and event,
• Data source and data quality,
• Segment and comparison framework,
• Reporting and interpretation,
• Decision and action,
• re-examination.

A dashboard is a form of representation in this order. It is not the control logic itself.

The central problem

Digital systems continuously generate data. Websites, campaigns, CRM, newsletters, shops, service processes, and technical platforms provide their own key figures. This diversity gives the impression that decisions can always be justified more precisely.

In practice, recurring breaks still occur.

Goals are not measurably translated

A formulation like “more visibility”, “better leads”, or “higher digital impact” sets a direction. It does not yet define how progress will be recognized and what decision will be derived from it.

Systems use different definitions

A lead, a conversion, a session, or an active user can mean different things depending on the platform, timeframe, and technical logic. If these differences are not documented, reports appear comparable even though they measure different things.

Data quality remains invisible

Duplicate events, missing consents, changed forms, test accesses, incomplete campaign parameters, and system changes can influence key figures. A cleanly designed chart does not automatically show this uncertainty.

Key figures are reported but not controlled

A monthly report can be created and read regularly without thresholds, audit tasks, or responsibilities being defined. Then it describes the past without improving future decisions.

Attribution is confused with causality

Attribution models mathematically distribute shares of a conversion to touchpoints. They do not fully answer why a decision was made. Especially in longer B2B processes, personal contacts, internal coordination, offline influences, and unobservable research remain partly outside the measurement.

Experiments begin without a hypothesis

A test only generates reliable learning if it has been defined beforehand what change is expected, which target metric is relevant, and which conditions can influence the result. Variants without a clear hypothesis produce activity, but not necessarily insight.

What digital control means

Digital control is the organized connection of data and decisions.

It answers five questions:

1. What impact should be achieved?
2. What signals indicate progress, quality, or risk?
3. How robust are these signals?
4. Which decision follows from which finding?
5. How is it checked whether the measure actually had an effect?

Control does not mean centrally controlling every development. It creates a reliable framework in which those responsible can identify deviations, investigate causes, and adjust priorities.

Central terms

Metric

An observable or calculated value. A metric can be important without already being a KPI.

KPI

A deliberately selected metric used to assess a central goal, process, or risk. The selection does not turn a key figure into objective proof. It defines its control function.

Target value

An expected or target value at a defined point in time or under specific conditions. Target values require a professional justification and a baseline value.

Baseline

The documented initial state against which later changes are classified. Without a stable baseline, improvements and deteriorations can be incorrectly assessed.

Event

A defined action or system state that is captured. Events require unique names, triggers, parameters, exclusions, and meanings.

Data quality

The suitability of data for the intended purpose. This includes, among other things, completeness, correctness, consistency, timeliness, and context.

Segment

A technically justified subset of data, for example by user group, device, region, channel, content type, or process phase. Segments prevent relevant differences in average values from disappearing.

Reporting

Regular or ad-hoc preparation of data and findings. Reporting informs. Control only arises through interpretation, responsibility, and decision-making.

Attribution

Exemplary assignment of effect or conversion shares to contact points. Attribution is a perspective on the observable decision path, not a complete reconstruction of cause and effect.

Experiment

Planned change to test a hypothesis. An experiment requires a defined target variable, comparison logic, sufficient data quality, and documented conditions.

The control loop

Typical misconceptions

„We need a better dashboard first.“

A dashboard can make confusing data more understandable. It cannot fix missing target definitions, conflicting events, and unclear responsibilities.

„The more data, the more reliable the decision.“

Additional data can reduce uncertainty. They can also create new contradictions, spurious correlations, and room for interpretation. Quality and relevance are more important than mere quantity.

“A KPI is objective.”

A KPI is based on a selection. This selection can be meaningful and robust, but it remains tied to a goal, definition, period, and data source.

"Attribution shows which channel generated the conversion."

Attribution distributes observable impact according to a model. It does not replace a causality check and does not map every influence on a decision.

“A positive test value proves a lasting improvement.”

A result is only valid for the tested conditions. Seasonality, sample, technical changes, target group mix, and follow-up effects must be considered.

"Data protection prevents meaningful measurement."

Unclear or excessive data collection complicates measurement and trust. Purpose limitation, data minimization, and documented processes can improve the technical quality of a measurement system.

Connection to other topics

• Digital Strategy Defines goals, priorities, and dependencies. Check data to see if the assumptions made hold true.
• Positioning and Impact determines which perception and decision should be supported. Measurement must not shorten this effect to reach.
• Websites and digital systems generate events and process data. Their architecture determines what can be reliably observed.
• Visibility and Search deliver findability and interaction signals. These must be linked to quality, decision, and business outcome.
• Content and Expertise require metrics for usage, orientation, and maintenance, without reducing professional value to click numbers.
• AI and automation increase the need for quality control, logging, and responsible limit values.
• Organization and Transformation Anchor roles, approvals, and learning loops in operations.

Framework of action

1. Create decision inventory

Collect recurring decisions in marketing, sales, website, content, and digital processes. Mark which of these are based on reliable data, assumptions, or mere habit today.

2. Document measurement concept

Assign a goal, hypothesis, metric, event, data source, quality rule, and responsible role to each prioritized decision.

3. Limit KPI set

Separate key performance indicators from diagnostic values. A small set of KPIs increases clarity, as long as supplementary data remains available for root cause analysis.

4. Maintain data quality

Define test cases, check intervals, change logs, and escalation paths. Data quality, according to ISO 8000, requires not only technical rules but also roles and documented responsibility.

5. Align reports with decisions

Every report should indicate which question it answers, what uncertainty exists, and what action or check follows from a deviation.

6. Embed experiments in learning loops

Tests are derived from hypotheses, checked technically and factually, documented and evaluated according to their scope. A test is not an end in itself.

What companies should not do

Companies should not treat reporting as a monthly mandatory production. A report without a recipient, decision, and reaction ties up resources without generating control.

You should also not claim uniform accuracy across all systems. Website analytics, CRM, advertising platforms, and sales data have different definitions, gaps, and time references.

The subsequent selection of suitable key figures is equally problematic. Anyone who decides which value is to be considered a success only after an action has been taken weakens the significance of the evaluation.

Consequences for companies

A robust data and control system does not create complete certainty. It makes uncertainty visible and manageable.

It connects:

• few relevant goals,
• clear definitions,
• fit-for-purpose data,
• documented quality,
• traceable interpretation,
• clear responsibility,
• controlled improvement.

This turns reporting from a retrospective view into a learning management layer of the digital system.

Subject-matter connection

Merge KPI, analysis, and control logic

A viable control system connects business goals, measurement concept, data quality, reporting, and decision-making processes. Structured discovery creates transparency about existing data, gaps, definitions, and responsibilities before additional dashboards or tools are introduced.

Structurally check data basis and control logic

Author

About Göke Frerichs

Göke Frerichs has been combining digital strategy, communication, technology, and implementation since 1999. As a digital strategist and Smart Digital Creative, he supports owner-managed B2B companies in developing clear and reliable digital systems from individual measures. His perspective is based on many years of consulting and implementation experience in the DACH region and North America.

More about Göke Frerichs