Home / Lean Six Sigma / SPC & process capability
Find which family holds the variation — within the part, part to part or over time — and whether the data sequence hides clusters, mixtures, trends or oscillation. Use it at the start of an improvement project, with a few parts measured at several points and times: you get which family dominates (what to investigate first) and, on the run chart, whether the data order is random.
Pre-sizing and estimating tool: results do not replace design, calculations or reports signed by a licensed professional, nor verification against the applicable codes and standards.
How to read: each panel is a level of the outermost factor (for example, an hour); inside it, the groups are the parts and the connected points are the positions. Compare three distances: between panels (variation over time), between the groups of a panel (part to part) and between the points of one group (within the part). The largest distance points to the family to tackle first.
Share of the total variation (sum of squares) explained by each factor, from the outermost to the innermost; the rest is repetition within the same combination.
How to read: each bar is the share of total variation coming from one family (factor); "repetition" is what is left within the same combination. The longest bar is where to look for the cause: if it is time, look at wear, warm-up and adjustments; if it is the part, raw material and fixturing; if it is position, machine geometry.
The numbers behind the plotted points: N, mean, standard deviation, minimum and maximum.
Key: N = number of measurements at the level; Mean, StDev, Minimum and Maximum = summary of those measurements. In the families table, DF = degrees of freedom (levels − 1), SS = sum of squares (variation assigned to the family) and % of total = the same share as the bar chart.
Points in collection order with the median. Shaded band = longest run about the median; thick line = longest run up or down.
How to read: horizontal axis = collection order; vertical axis = value (or subgroup mean); line = median. A random process crosses the median often and goes up and down without pattern. A long shaded band = many points in a row on one side (level shift); a long thick line = continuous rise or fall (trend).
Bars below the α line indicate a non-random pattern. Log scale.
How to read: one bar per test (clustering, mixtures, trend, oscillation), with the P-value written next to it. The dashed line is the chosen α. Green bar, right of the line = no signal; red bar, left of the line = the pattern of that test is real and worth investigating.
Runs about the median (clustering and mixtures) and up or down (trends and oscillation), with normal-approximation P-values.
Take this page’s results to another calculator with the fields already filled in.
The calculation report gathers this page's inputs, results, equations, assumptions and references, with fields for project, engineer and revision. The full report also includes the charts and interpretations, in color. In the print dialog, choose Save as PDF.
A chart that shows at once how a measurement changes within the part, from part to part and over time. Each point is the mean of a factor combination, and lines join the means of the same group. It helps decide which family of variation to investigate before running experiments.
Measure a few consecutive parts (3 to 5) at several positions (ends and center, for example) and repeat at several times (hourly, by shift) until you capture at least 80% of the normal process variation. Record one row per measurement with the position, part, time and value.
The page splits the total variation among the factors, from the outermost (hour, shift) to the innermost (position). In the shaft example, the hour-to-hour drift holds most of it: the problem is temporal, typical of tool wear or warm-up, rather than clamping or material.
No: it is a visual, descriptive tool with no P-value. Use it to choose where to look and then confirm with ANOVA or an experiment. The next-step button takes the same data to ANOVA.
Whether the order of the points looks random. It counts runs of points on the same side of the median (too few = clustering, too many = mixtures) and runs that only rise or only fall (too few = trend, too many = oscillation), with a P-value for each pattern.
The run chart uses no control limits; it detects order patterns even when all points are inside the limits. The control chart shows whether the process is stable relative to its natural variation. The two complement each other.
When several parts are measured at the same time (for example, 5 per hour), each subgroup becomes one point with its mean or median, and the tests look at the sequence of times. In the crankshaft example, 125 measurements form 25 subgroups of 5.
Important notice: Indus Metrics tools produce estimates and pre-sizing results from the data you enter and from published engineering methods. They do not replace design, calculations or reports signed by a licensed professional engineer (PE), nor verification against the applicable codes, standards and regulations, such as OSHA rules and the ASME codes. Material and fuel properties vary: check manufacturer data or your own process analyses before deciding. By using this site you agree to the Terms of Use.
© 2026 Indus Metrics. Found an error? Write to contato@seudominio.com.