Analyser reference: terms, charts and settings

• Observation / N: an observation is one recorded value. N is the number analysed after cleaning and date filtering. Check unexpected differences from your source records.
• Mean: add the values and divide by how many there are. This is the average. Extreme values can pull it up or down.
• Median: the middle value when values are sorted; with an even number, the average of the middle two. It is less affected by extremes than the mean.
• Standard deviation (SD): a measure of spread around the mean, in the same units as the observations. A small SD does not, by itself, prove stability or predictability. Summary-bar SD covers the selected series; Stage Summary SD covers that stage.
• Stage / change point: a stage is a period assigned its own average by the analysis. A change point separates stages. These are estimates from the available observations—not proof of permanent system change. “Baseline” labels the first stage; it does not certify stability.
• Change %: the relative difference between a stage average and the previous one. For a percentage measure, this is not the same as a percentage-point difference. It is undefined when the previous average is zero.
• Confidence score / Conf %: the method compares a segment’s cumulative pattern with randomly shuffled versions of its values. Higher scores indicate stronger evidence against that shuffled-order comparison. They are not probabilities that a change is real or an intervention worked. Repeated testing can produce false alarms.
• Common-cause and special-cause variation: routine variation within a stable process, versus signals suggesting something outside that routine pattern. One detected stage does not establish that all variation is common-cause.

Step Change Analysis: red points show observations; the blue line shows stage averages, both on the left-hand scale. The green CUSUM uses the right-hand scale. Stage Summary lists dates, Mean, SD and Change %. Individual final-stage confidence scores are not shown.

CUSUM: a cumulative sum—a running total of each observation’s difference from the whole-series mean. It rises for values above that mean and falls for values below it. Steeper movement means a larger difference per observation, not statistical significance. The separate CUSUM tab shows this line alone. A bend is not automatically a detected change, and “rising” does not necessarily mean “improving”.

Run Chart: observations in time order with a median line. This tool marks runs of six or more observations on the same side of the median, ignoring values exactly on it. It does not apply every published run-chart rule. Marked runs prompt investigation; they do not identify causes.

X-mR: the upper chart shows individual observations (X). The lower shows moving ranges (mR): absolute differences between successive values. There is no moving range for the first observation.

• UNPL / LNPL: upper and lower natural process limits for individual values. Here they are the whole-series mean plus or minus 2.66 times the average moving range.
• URL: upper range limit for mR, calculated as 3.267 times the average moving range.

A point beyond a limit prompts investigation—not automatic blame or adjustment. These limits use the entire selected series, not separate stages. Shifts, seasonality and links between successive values can affect them. Limits are not targets, acceptable-performance standards or guaranteed 99.7% coverage.

• Turn Length: controls when the algorithm stops testing short segments for further splits. It does not guarantee a minimum number of observations in every returned stage.
• Loops: the number of random shuffles for each comparison. More loops reduce simulation variability but take longer. Small differences on rerunning can occur, especially near a threshold; extra loops do not fix unsuitable data.
• Conf %: the minimum comparison score required for a split. Higher thresholds generally report fewer splits. The headline chart label shows this selected threshold. Individual final-stage scores are not shown because a score from an earlier split is not a separate test of the final stage. The setting is not a guarantee about cause or permanence.
• Show 3SD Limits: adds each stage mean plus or minus three stage SDs. These bands describe spread—not confidence intervals for the mean, targets or the moving-range-based X-mR limits.
• Start / End Date: limits the observations analysed. Changing the period changes the reference statistics and may change the result.

Record your settings and period. Do not keep adjusting them to obtain a preferred answer. Too little data, missing records, seasonal patterns, gradual trends or related successive observations may require a different approach.