Individual Value / Dot Plot: See Every Data Point Before You Trust the Average

The Individual Value / Dot Plot tool does one thing well. You upload your data and it draws every single observation as a dot. Nearby values share a bin and stack vertically. You see the shape of your data instead of a single summary number.

That matters in the Analyse phase of RDMAICS. By the time you reach Analyse, you usually have a measurement set from the Measure phase. The temptation is to reduce it to an average and move on. An average hides a lot. Two very different processes can share the same mean. A dot plot shows you which one you actually have.

Reach for this tool early in Analyse, before you build control charts or run comparison tests. It is a sanity check on shape. It answers three questions quickly. Where is the centre. How wide is the spread. Is anything sitting apart from the rest.

This guide uses the sample data built into the tool. The case is fictional. It is included so you can follow every step without supplying your own worksheet first.

The scenario is a baseline measurement taken across a clinic network. Twelve weekly observations were recorded, labelled Week 1 through Week 12. Each observation carries the same group value of All clinics and the same series value of Baseline. The time or sequence column runs 1 to 12.

The measurement values are 18, 17, 19, 18, 20, 17, 18, 19, 18, 17, 19 and 18. There are no notes attached to any row. That is the entire input. Twelve rows, one measurement each, nothing missing.

The worksheet layout is simple and you should copy it. One row per observation. One column that holds the number you want plotted. The tool marks that column with an asterisk, which tells you it is the required field.

The remaining columns are optional but useful. Observation or ID gives each dot a readable label. Group or stratum lets you split the plot later if you suspect more than one population. Series or scenario separates a baseline from a post-change set. Time or sequence records the order. Notes carry anything a reader would need to interpret a specific point.

In the sample data, the group and series columns are constant. Every row says All clinics and Baseline. That is deliberate. It gives you a single undivided distribution to read first. Once you understand the whole, stratifying becomes a decision rather than a reflex.

A practical note on preparation. Clean your numbers before upload. Blank cells, text entries and stray units reduce the count of usable observations. The tool reports N so you can check that the number it used matches the number you expected.

Run the tool and the output arrives in four parts. The plot itself, an executive interpretation, a statistical results table, and a short set of findings and next actions.

The plot shows the measurement axis running across values labelled 17.38, 17.98, 18.57 and 19.63. The stack axis runs from 0 to 8. The footer under the chart reads n=12, median=18.000, and 4 display bins. The median is marked on the plot so you can locate the centre without reading the table.

The statistics table gives you the numbers. N is 12. The mean is 18.1667. The median is 18. The standard deviation is 0.9374. The minimum is 17 and the maximum is 20. Tukey outliers number 0, which means no value falls outside the 1.5 times interquartile range fences.

Key findings summarise this in plain terms. Distribution spread is 3. No flagged outliers. The observed range is 3. The decision ready conclusion states that no Tukey outliers were identified in the supplied values. The tool also offers an AI review that explains the result in narrative form if you want a second reading.

The first decision is whether to keep investigating this data set as a single population. The mean of 18.1667 and the median of 18 sit close together. The range is 3. Nothing is flagged as an outlier. That pattern is consistent with one population rather than two distinct groups overlaid.

The second decision is about spread. A standard deviation of 0.9374 across twelve observations tells you how tight the baseline is. If your specification or target demands tighter variation than that, you now have a measured starting point to argue from. If the spread is acceptable, your improvement effort may belong on the centre instead.

The third decision is where to look next. The tool itself proposes two actions. Inspect the dot pattern for clusters or multiple populations. Stratify by a meaningful process factor if clusters are present. In this case the plot uses four display bins across a range of 3, so there is limited scope for dramatic clustering. That is still useful information. It tells you the baseline is unremarkable and that a change later on should be easy to see against it.

Keep the output as a record. When you reach the Improve phase and collect a second set of measurements, you can plot it as a new series and compare the two distributions side by side using the same method.

The most common mistake is treating the dot plot as a time series. The tool states this explicitly in the chart description. The plot reveals clusters, gaps and unusual values without implying time order. Even though the sample data carries a sequence from 1 to 12, the horizontal axis is measurement, not week. If you want to see drift over time, you need a run chart or a control chart.

The second mistake is reading too much into bin boundaries. The sample output uses four display bins. Bins are a drawing convenience. Shift the bin width and the visual shape shifts with it. Anchor your conclusions in the statistics table, which is calculated from the values themselves, not from the bins.

The third mistake is over interpreting a small sample. Twelve observations is enough to see obvious structure. It is not enough to be confident about the tails of a distribution. Zero flagged outliers in twelve points does not guarantee that outliers never occur.

The fourth mistake is stratifying without reason. Splitting the sample data by group would produce a single group, All clinics, which tells you nothing new. Split only when you have a process reason to believe the groups behave differently.

The Individual Value / Dot Plot tool will not tell you why your data looks the way it does. It shows you the distribution of twelve numbers. The cause of the spread of 3 sits in the process, in the measurement system or in the people doing the work. The tool cannot see any of that.

It will not tell you whether your process is stable or in control. There is no control limit calculation here and no test for special cause signals over time. A process that is drifting steadily upward can still produce a tidy dot plot with no flagged outliers. Use a control chart for stability.

It will not compare groups for you statistically. The plot can display stratified dots, but it does not run a hypothesis test and it does not produce a p value or a confidence interval. If you need to claim that two groups differ, you need a comparison test, not this plot.

It will not validate your measurement system, judge your data against a specification, or declare any value acceptable. The Tukey rule flags values outside 1.5 times the interquartile range. That is a statistical convention, not a quality standard. In the sample data it returned zero. A zero means nothing met that mathematical condition. It does not mean the process is good.

Finally, it will not fix a poor data set. If rows are missing, if the measurement column mixes units, or if the twelve numbers came from four different recording methods, the plot will still render. The output will simply describe bad data accurately. Judgement stays with you.