A Data Collection Plan defines how evidence will be gathered before data collection begins.

It answers:

  • What will we measure?
  • Why?
  • How?
  • Where?
  • When?
  • Who collects it?

The objective is to prevent teams from collecting large amounts of data that cannot answer the problem-solving question.

Start with the decision

Before choosing the data, define what the team needs to understand.

Examples include:

  • establish a baseline;
  • compare shifts;
  • test a suspected cause;
  • measure a defect rate;
  • verify an improvement.

DMAIC uses data throughout Define, Measure, Analyze, Improve, and Control.

The collection plan should support a specific decision inside that logic.

Define each measure operationally

For every variable or attribute, define:

  • name;
  • unit;
  • inclusion rule;
  • exclusion rule;
  • calculation.

Operational Definition ensures different people interpret the measure consistently.

A term such as “delay” is too vague unless the start and stop conditions are defined.

Identify the data source

Possible sources include:

  • machine historian;
  • ERP;
  • manual observation;
  • inspection record;
  • work order;
  • customer system.

The team should understand whether the source reflects the process accurately.

Existing data is convenient but not automatically suitable.

Define where and when to collect

Sampling should represent the conditions relevant to the question.

Consider:

  • shift;
  • product;
  • machine;
  • day;
  • process stage;
  • supplier lot.

Problem Stratification helps identify meaningful categories that may need to be captured during collection.

If the team forgets to record the machine or shift, later stratification may be impossible.

Define sampling logic

The plan should state:

  • sample size;
  • frequency;
  • random or systematic selection;
  • observation duration.

The appropriate design depends on:

  • process frequency;
  • variation;
  • risk;
  • analysis method.

Avoid choosing sample size only because it is convenient.

Verify the measurement system

Measurement System Analysis helps determine whether the measurement process is capable of supporting the intended analysis.

Collecting more data does not correct a poor measurement system.

Protect data integrity

Plan how data will be:

  • recorded;
  • stored;
  • checked;
  • version controlled.

Manual collection forms should minimize:

  • ambiguous fields;
  • missing entries;
  • transcription.

Automated data still needs validation.

Pilot the plan

Run a small pilot before full collection.

Ask:

  • Can collectors apply the definitions?
  • Are fields missing?
  • Is the frequency practical?
  • Does the data answer the intended question?

Small Tests of Change provides useful thinking for testing the collection process before scaling it.

Common mistakes

Collecting data before defining the decision, using vague measures, ignoring stratification fields, selecting a sample for convenience, trusting existing system data without validation, and discovering after weeks of collection that the measurement cannot answer the question are common mistakes.

Practical sequence

  1. define the decision or question.
  2. identify required measures.
  3. create operational definitions.
  4. identify the source.
  5. define stratification fields.
  6. define sampling and frequency.
  7. verify the measurement system.
  8. design the recording method.
  9. pilot the collection plan.
  10. collect only after the plan is usable.

The practical lesson

A Data Collection Plan creates discipline before data collection.

Good analysis begins by collecting the right evidence in a way that can be trusted.

This topic also connects with DPMO. Use that method when the improvement requires the related operating or management discipline.