What is stratification?
Stratification is the practice of splitting data into meaningful groups and comparing them. It is one of the seven basic quality tools because averages and totals hide where problems come from. A defect rate of 5% might be 2% on two lines and 11% on the third. Stratification makes that visible and points the investigation in the right direction.
Typical stratification factors
- Equipment: machine, line, mold cavity, fixture, test station
- People: operator, shift, crew, inspector
- Materials: supplier, lot, batch, raw material grade
- Time: hour, day of week, month, season
- Product: part number, variant, customer, region
Stratification works with every other basic tool: stratified Pareto charts, histograms per machine, and groups on a scatter diagram.
Stratification example
The example records leak test failures for 36 batches across three lines, two shifts and two suppliers. Stratified by line, line B stands out. Stacking by shift shows why: most of line B's failures come from the night shift. Switching the grouping to supplier shows a much smaller difference, so the investigation should start with line B's night shift: setup, staffing and handover.
Tips
- Plan stratification before collecting data. You cannot split by a factor you did not record.
- Compare like with like. If groups differ in volume, compare rates, not totals.
- Go one level deeper on the worst group, then use a fishbone diagram to explain it.
How to use this Stratification tool
- Collect data with its context. For each observation, record the value plus the factors that might matter, such as line, shift and supplier.
- Paste it into the table. Paste a block from Excel: the value in the first column and up to three factors next to it. Name the factors.
- Choose counts or measurements. Counts are summed per group; measurements are compared by spread and mean.
- Stratify by one factor. Pick the factor to group by, and optionally a second to stack or colour by.
- Switch factors. Look at the same data by each factor in turn. The one that separates the groups most is your lead.
- Act and export. Investigate the stand-out group with root cause analysis and export the results.
Frequently asked questions
What is stratification in quality control?
Stratification means separating data into groups (strata) by a factor such as machine, shift, operator, supplier, material lot or location, and comparing the groups. A problem that looks spread out overall often turns out to come mostly from one group.
Why is stratification one of the 7 basic quality tools?
Because combined data hides causes. Two machines with different averages can produce a wide, flat histogram; one bad supplier can make a whole process look unstable. Stratifying turns "the process is bad" into "this machine on night shift is the problem".
Which factors should I stratify by?
Use the factors that could plausibly influence the result: the 6Ms of a fishbone diagram are a good checklist (people, machines, methods, materials, measurement, environment), plus time (shift, day, week). Record these factors when you collect the data, or you will not be able to stratify later.
What is the difference between counts and measurements mode?
Use counts when each row is a number of defects, failures or events to be added up per group; the tool shows bars with each group's total and share. Use measurements when each row is a measured value; the tool shows the spread of each group with box plots, points and means.
How do I combine two factors?
Choose one factor to stratify by and a second to stack (counts) or colour (measurements). The tool also reports the single largest combination, for example "Line B × Night", which is where to look first.
Can I export the stratified results?
Yes. Export the chart, the summary by group and your data to PDF, PNG, Excel or CSV. It is free with no sign-up.