You are viewing a public preview
Sign in to save your progress, do assignments, join discussions and get feedback.
Trend lines, correlation, forecasting, trend and seasonality analysis; clustering and customer segmentation; parameters and their use with calculated fields; filter, highlight and URL actions; identifying patterns, outliers and group differences, and interpreting business findings.
Learning outcomes
The lecture slides for this lesson are below.
Practise advanced charts, calculated fields, table calculations and LOD expressions on the practice data; explore trends, forecasts, clusters, parameters and interactive actions, and write a short interpretation of what you find. (About 24 hours of independent study.)
By the end of this lesson you can:
Key idea: Exploration tools help you find structure in data, but a pattern is only useful once you can explain it in business terms and know how certain it is.
| Tool | What it does | Typical use |
|---|---|---|
| Set | A custom field defining a subset; every member is either In (True) or Out (False) | Compare top customers with the rest |
| Group | Combines members of one dimension into a higher-level category | Merge provinces into areas |
| Bins | Divides a measure into equal-width intervals, creating a new dimension | Histogram of order values |
| Cluster | Groups marks with similar values using k-means | Customer segmentation |
Sets are created by right-clicking a dimension (Create → Set). You can pick members manually, define a condition (for example SUM(Profit) > 0) or use the Top tab (top N by a measure). Condition and Top sets update when the data change. Put the set on Color to see In/Out, or use Show Members in Set.
Groups collect members with the same characteristics into a higher-level category within a single dimension (paperclip icon or Create → Group).
Bins are created from a measure (Create → Bins). Tableau suggests a bin size; you can change it. The result is a new dimension such as Sales (bin).
Clusters use the k-means algorithm to divide marks into k groups. Drag Cluster from the Analytics pane onto a view; Tableau uses the measures in the view as variables (you can add others), chooses the number of clusters automatically or uses the number you set, and you can drag the result to the Data pane to save it as a group. Describe Clusters summarises each cluster.
A parameter is a workbook variable (a number, date or string) whose value the viewer can change through a parameter control. A parameter does nothing on its own: it must be referenced by a calculated field, a filter (for example Top N by a parameter), a set, a reference line or a bin size. Show Parameter displays the control.
Actions (Worksheet → Actions or Dashboard → Actions) respond to a hover, a selection or a menu click:
Look for patterns (trends, seasonality, clusters), outliers (marks far from the rest) and group differences (one segment behaves differently). For each, ask: Is it large enough to matter? Could a data issue explain it? Is there another factor behind it? Then state the finding with its limits and a business implication.
The sales director wants one chart that can switch between Sales, Profit and Quantity by Region, so she can see whether the ranking of regions changes with the metric. Use the course data.
Choose Measure, data type String, allowable values List: Sales, Profit, Quantity.CASE [Choose Measure]
WHEN "Sales" THEN SUM([Sales])
WHEN "Profit" THEN SUM([Profit])
WHEN "Quantity" THEN SUM([Quantity])
END
Every branch is aggregated, so the calculation is valid.
3. Build the view. Region on Rows, the calculated field on Columns, sorted descending. Right-click the parameter and choose Show Parameter.
4. Dynamic title. Insert the parameter into the title: Regions ranked by <Choose Measure>.
5. Add context. Put Category on Color to see which category drives each region's bar.
6. Interpret. Switch the parameter. If a region ranks high on Sales but lower on Profit, its sales may depend on discounts or low-margin categories; that is a hypothesis to test next with Profit Ratio by Discount, not a conclusion.
Flashcards · Data Exploration and Interpretation
10 wordsA custom field that defines a subset of data; each member is In (True) or Out (False).
Equal-width intervals created from a measure, producing a new dimension.
Groups of marks with similar values found with the k-means algorithm.
An algorithm that splits data into k groups by minimising the distance of points to their group centre.
A fitted model (linear, exponential, etc.) showing the overall direction of data.
Projected future values from exponential smoothing, with trend, seasonality and a prediction interval.
A pattern that repeats over a fixed period, such as every year.
A workbook variable the viewer can change; it must be used in a calculation, filter, set or line.
An interaction where selecting marks in one view filters other views.
An interaction that emphasises related marks and dims the rest.
FIXED LOD from Lab 4B) into bands of 500 and describe the distribution.