Visualize and analyze the data · Topic 6 of 6
Patterns, Trends and AI Insights
PL-300 analysis: use Analyze to explain changes, group, bin and cluster data, AI visuals, reference lines, forecasting, anomaly detection and Copilot summaries.
- 5 min
- Analyze offers Explain the increase or decrease, and Find where this distribution is different.
- Group combines selected values. Bin makes equal ranges. Clustering finds similar items in a scatter chart.
- Key influencers shows what drives a metric. Decomposition tree breaks a measure down, with AI splits.
What it is
Beyond showing numbers, Power BI helps you explain them. The Analyze feature finds reasons for a change, grouping, binning and clustering organise data, AI visuals find drivers and segments, the Analytics pane adds reference lines, error bars and forecasts, and anomaly detection flags outliers. Copilot can summarise the semantic model in plain language.
How it works
The Analyze feature
Right click a data point in a line or column chart and choose Analyze. There are two options.
- Explain the increase or Explain the decrease looks at the change between this point and the previous one and shows the factors that contributed most, in a chart you can add to the report.
- Find where this distribution is different compares two parts of a distribution and shows which fields differ.
Analyze needs a field on the axis that is a date or a category with a measure in values. It works on the fields in the model, so a well modeled dataset gives better explanations.
Grouping, binning and clustering
| Tool | Use |
|---|---|
| Grouping | Combine categories you select into one group, such as several small countries into "Other". Right click and choose Group data |
| Binning | Put numeric or date values into equal sized ranges, such as ages in 10 year bins. Create from New group, with bin size or number of bins |
| Clustering | Find similar items automatically in a scatter chart. Choose Automatically find clusters and set the number of clusters |
Each creates a new field in the model that you can use as a legend, an axis or a slicer. Binning supports histograms.
AI visuals
| Visual | What it does |
|---|---|
| Key influencers | Shows which factors most affect a metric, for example what drives customer churn, and compares segments |
| Decomposition tree | Lets you break a measure down by fields in any order, and can pick the next field automatically with AI splits (high value or low value) |
| Q&A | Answers a typed question with a visual |
| Smart narrative | Writes a text summary that updates with the page |
The Key influencers and Decomposition tree visuals work with categorical and numeric fields, and need enough rows to find a reliable pattern.
Reference lines, error bars and forecasting
Open the Analytics pane for a visual. Available lines depend on the visual.
| Feature | Use |
|---|---|
| Constant line, Min, Max, Average, Median, Percentile line | A reference line on the axis. The constant line is a fixed target |
| Trend line | A fitted line through the data |
| Forecast | Predicts future points on a line chart with a continuous date axis. You set the forecast length, confidence interval and seasonality, and can ignore recent points |
| Error bars | Shows an upper and lower bound for each value, with fixed or field based limits |
| Symmetry shading | Highlights where a scatter chart is above or below the diagonal |
A forecast gives a range, not a promise, and needs enough history. Seasonality is detected automatically when you leave it unset, or you set the number of points per cycle.
Detect outliers and anomalies
On a line chart with a date axis, the Analytics pane has Find anomalies. Power BI marks points that fall outside the expected range, with a sensitivity setting where a higher value marks more points. For each anomaly, it can suggest possible explanations by looking at other fields. You can also find outliers with a scatter chart, a percentile line, or a top N filter, and with column profiling in Power Query.
Copilot to summarise the model
In the Copilot pane, ask Copilot to "summarize the model". The reply describes the tables, fields and what the semantic model contains, which helps an author learn an unfamiliar model before deciding what to analyse. Copilot needs the paid capacity and tenant setting described earlier.
A worked scenario
Revenue fell in March. The analyst right clicks the March point on the line chart, chooses Analyze, Explain the decrease, and finds that one product category in one region caused most of the drop. She adds the explanation chart to the report. A scatter chart of customers by spend and visits uses clusters to find four segments. A line chart gets a forecast for six months and Find anomalies, which flags a week with an unusual spike.
What to remember for the exam
- Analyze offers Explain the increase or decrease, and Find where this distribution is different.
- Group combines selected values. Bin makes equal ranges. Clustering finds similar items in a scatter chart.
- Key influencers shows what drives a metric. Decomposition tree breaks a measure down, with AI splits.
- The Analytics pane has reference lines, trend line, forecast, error bars and Find anomalies.
- Forecast and anomaly detection need a line chart with a continuous date axis.
"Why did the figure change?" points to Analyze. "What drives this metric?" points to Key influencers.
How the exam asks about this
- "Explain why sales dropped in March." Analyze, Explain the decrease.
- "Combine small categories into one." Group data.
- "Find customer segments automatically." Clusters in a scatter chart.
- "Predict the next quarter's sales on a line chart." Forecast.
Common mistakes
Using a pie chart to show a trend.
Expecting forecast without a date axis.
Using clustering on a non scatter visual.
Mixing up Key influencers and Decomposition tree.
Check yourself
Which feature shows what contributed to a change between two points?
Analyze, Explain the increase or decrease.
Where do you create a forecast?
In the Analytics pane of a line chart.
Which visual shows what drives a metric?
Key influencers.
Topic 6 of 6 in Visualize and analyze the data
Test yourself on Visualize and analyze the data
The full Microsoft PL-300 question bank has 657 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by Microsoft. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.