Data collection plan and check sheet

Start with the question

A data collection plan begins with the question the data have to answer. "Is the reject rate different between shifts?" tells you to record the shift on every result. "Let's collect some data on rejects" tells you nothing, and the data that come back usually cannot answer the question that turns up later.

Operational definitions

An operational definition says exactly what counts and how it is measured, clearly enough that two people would record the same thing. "Scratch" is not an operational definition. "Any mark on surface A visible at arm's length under the bench light, that a fingernail catches on" is. Without one, a change in the data may only be a change in who was recording.

Data types

• Continuous (variable) data are measured on a scale: diameter, time, force.
• Discrete data are counts: defects per unit, calls per hour.
• Attribute data are categories: pass or fail, defect type. Measurements and counts are quantitative data; categories are qualitative. Many texts, including the ASQ Green Belt Body of Knowledge, group discrete and attribute data together, so expect either usage.

The type decides which charts and tests you can use later. It also decides how much data you need: attribute data need far larger samples than continuous data to detect the same change. If something can be measured on a scale, it is usually worth measuring rather than judging pass or fail.

Stratify at the time

Record the things you might want to split the data by when you collect it: shift, machine, operator, material lot. It costs a column on the sheet. Trying to reconstruct which lot a part came from three weeks later is usually impossible. The plan check flags measures with nothing to stratify by.

The check sheet

A check sheet is the simplest data collection form there is: categories down the side, periods across the top, a mark each time something happens. It shows patterns as it fills (all the marks on Wednesday, or in one row) and it produces counts ready for a Pareto chart. The one here totals itself; print it if you would rather collect on paper.

The location check sheet

A tally by category says what went wrong; a location check sheet, also called a defect concentration diagram, says where. Draw a simple outline of the item, split into zones (a rectangle in rows and columns, or a circle in rings and sectors for round parts), pick a defect type and tap the zone each time one is found. Defects that pile up in one place point to a cause at that place: a fixture contact point, a sharp edge, a handling spot, one nozzle of a spray head.

The tool counts by zone and by type, and tests whether the spread could be even with a chi-square goodness-of-fit test: each zone's expected count is the total shared out by area (equal zones in a rectangle; outer rings of a circle are larger). Treat the p value as approximate when the expected counts are below 5.

On the exam

Data collection plans (III.B.1), qualitative and quantitative data (III.B.2) and data collection techniques (III.B.3) are all in the 2022 CSSYB Body of Knowledge: expect to classify data by type and to recognize what a data collection plan contains. Check sheets are one of the basic quality tools in CQE V.A.5, CQPA II.B and CMDA V.A, and tally or check sheets are a data collection method in CQE VI.A.3. Expect to pick the check sheet that fits a question: a tally sheet for "how often", a location check sheet for "where".

Other tools and templates