DATA4100 – Week 2: Statistical Methods & Visualisation

DATA4100 · Week 2

Statistical Methods for Summarising & Visualising Data

Turning retail data into decisions — hands-on with Power BI and Tableau using the Superstore dataset.

Kaplan Business School Australia

Today's Workshop

Learning Outcomes & Agenda

Four things you will be able to do by the end of this session

1

Ask the Right Questions

Frame business problems as analytical questions using the Superstore case as context.

2

Choose the Right Chart

Use the chart-selection framework to match your data and question to the correct visualisation type.

3

Build Charts in Power BI

Create bar, stacked, pie, treemap, line, and scatter chart visuals step by step from a CSV file.

4

Interpret & Communicate

Read chart outputs critically and explain what each visualisation means for the business.

Dataset

Superstore Sales — a fictional US retailer selling Furniture, Office Supplies, and Technology across four regions. Download from the Week 2 reading list on MyKBS → GitHub → Datasets → Superstore Dataset.csv

Context

The Superstore Business Case

What data do we have and what are we trying to find?

Your Role

You are a Business Analyst at Superstore. Your mandate is to identify weaknesses and opportunities to enhance growth and profitability. Every chart you build should lead to a decision.

Dataset Columns

ORDER DETAILS

order_idorder_dateship_dateship_mode

CUSTOMER & PRODUCT

customersegmentmanufactoryproduct_name

CATEGORISATION

categorysubcategory

GEOGRAPHY

regioncitystatezip

FINANCIALS

salesprofitquantitydiscountprofit_margin

Preliminary Exploration in Excel

Open the CSV in Excel first — scan the rows and columns before jumping into visualisation tools.
How many rows (orders) are there? What date range does the data cover?
Are there any blank cells, unusual values, or obvious data quality issues?
Which fields are numbers? Which are categories? Which are dates?

Why This Matters

A Business Analyst who understands the data structure before visualising it asks better questions and catches errors early. Never skip this step.

Key Data Fact

The dataset spans January 2019 – December 2022 and contains 9,994 order rows across three product categories and four US regions.

Business Questions

Five Questions Worth Answering

Each question maps to a statistical method and a specific chart type

1

How do sales and order volumes vary across the four regions — Central, East, South, West?

Method: Descriptive Statistics → Bar Chart / Map

2

Which product subcategory performed best in terms of total sales?

Method: Descriptive Statistics → Bar Chart (sorted descending)

3

What are the top-performing products in Furniture, Office Supplies, and Technology?

Method: Descriptive Statistics → Treemap / Clustered Bar

4

Has sales performance improved over time — and what might the next 12 months look like?

Method: Time Series & Forecasting → Line Chart + Forecast

5

Does offering higher discounts actually increase profit — or does it destroy it?

Method: Correlation → Scatter Plot (Discount vs. Profit)

Activity

In groups, add three more questions of your own. For each one, decide: which statistical method applies, and which chart type would you use?

Overview

Six Statistical Methods, Six Visual Approaches

Every method has a natural chart type — and a set of fields that suit it

01

Descriptive Statistics

Bar, Column, Stacked Bar, Pie, Donut, Treemap, Table

sales · profit · quantity · profit_margin per segment / region / category

02

Time Series & Forecasting

Line Chart (continuous date axis), Forecast overlay

order_date + sales · profit · quantity

03

Correlation

Scatter Plot, Correlation Matrix, Heatmap

discount vs. profit · sales vs. quantity

04

Geographical Methods

Choropleth Map, Bubble Map, Filled Map

city · state · region + any financial metric

05

Regression

Scatter Plot with Trend Line (line of best fit)

discount → profit · quantity → sales

06

Distributions

Histogram, Density Plot, Box Plot

sales · profit · discount (shape of the data)

Key Insight

Choosing the wrong chart doesn't just look bad — it misleads the audience. The next slide gives you a framework to always choose correctly.

Framework

Which Chart Do I Need?

Start with your purpose — the chart type follows from what you want to show

Quick Answers

Nationalities at campus?

Pie or Bar Chart
→ Composition (parts of whole)

Attendance vs. grades?

Scatter Plot
→ Relationship between two numeric vars

Distribution of final grades?

Histogram
→ Distribution of one numeric var

Remember

This framework applies every time you build a new chart. Bookmark this slide.

Statistical Method 01

Descriptive Statistics: Mean, Median & Mode

Measures of location — where is the centre of the data?

x̄

Mean (Average)

Sum all values ÷ count of values. Sensitive to extreme outliers — a single huge sale can pull it up.

M

Median

Middle value in an ordered list. If n is even, average the two middle values. Robust to outliers.

Mo

Mode

The most frequently occurring value. Useful for categories (e.g. most common ship mode).

Worked Example — 16 Arrival Times (minutes)

Data: 9 14 11 10 10 8 10 9 14 8 15 12 9 9 7 9
Ordered: 7 8 8 9 9 9 9 9 10 10 10 11 12 14 14 15

Mean = 164 ÷ 16 = 10.25 min
Median = (9 + 10) ÷ 2 = 9.5 min (n is even)
Mode = 9 (appears 5 times)

Excel / Power BI Commands

Statistic	Excel Formula	Power BI Aggregation
Mean	`=AVERAGE(A1:A100)`	Average (default)
Median	`=MEDIAN(A1:A100)`	Median (from dropdown)
Mode	`=MODE(A1:A100)`	Not built-in — use DAX

Bridge to Power BI

We just calculated these by hand. Now we'll build a table visual in Power BI to verify — dragging segment to Rows and Average of profit_margin to Values. The table should show: Consumer 0.11 · Corporate 0.12 · Home Office 0.14.

Which should you use?

For skewed data (like sales, which has expensive outliers), median is more representative than mean. Power BI lets you switch aggregation with one click.

Power BI — Step 0

Loading the Superstore Dataset into Power BI

Do this once at the start of every Power BI session

1

Open Power BI Desktop. On the splash screen, click "Get data from another source →"

2

In the Get Data dialog, select Text/CSV from the list on the right. Click Connect.

3

Navigate to your Downloads folder, select Superstore Dataset.csv. The preview shows the first 200 rows. Confirm the delimiter is Comma and the file origin is UTF-8.

4

Click Load (not Transform Data). Your fields will now appear in the Data pane on the right side of the screen.

5

Click the Table icon (second icon on the left panel) to view the raw data and confirm it loaded correctly before building any charts.

Common Issue

If order_date shows as Text rather than Date, Power BI won't recognise it as a date field. You'll fix this in the Time Series section of this workshop.

What You Should See After Loading

After Loading

All column names appear in the Data pane (right side). You are now ready to build visuals by dragging fields onto the canvas.

Power BI — Chart 01

Bar Chart: Sales by Subcategory

Comparison — which product subcategory generates the most revenue?

Business Question

"Which subcategory should Superstore prioritise for marketing investment?"

1

Click on a blank area of the canvas. In the Visualizations pane (right), select the Clustered Column Chart icon (first row, second icon).

2

From the Data pane, drag subcategory to the X-axis field well.

3

Drag sales to the Y-axis field well. Power BI will automatically use Sum of sales.

4

Click the Focus mode icon (square with arrow, top-right of the chart frame) for a full-screen view. Click Back to report to return.

Interpretation

Phones and Chairs lead in sales (~$330K each). Fasteners and Labels are at the bottom (<$15K). Sort the axis descending to make the ranking instantly readable: click the chart → the "..." menu → Sort descending.

Visualizations Config

Clustered Column Chart

X-axis

subcategory ×

Y-axis

Sum of sales ×

Legend

Add data fields here

Output: Sum of Sales by Subcategory

Power BI — Charts 02 & 03

Clustered & Stacked Bar Charts

Comparing multiple measures, and showing part-to-whole breakdowns

Clustered Bar — Quantity & Profit by Category

"Does the category with most orders also generate the most profit?"

1

Select Clustered Column Chart

2

Drag category → X-axis

3

Drag quantity and profit → Y-axis (both)

4

Two differently-coloured bars now appear per category. Click Focus mode.

Read the chart

Technology has high profit despite fewer orders than Office Supplies — higher-value items. Furniture has many orders but low profit margins.

Stacked Bar — Orders by Category & Subcategory

"What's the product mix breakdown within each category?"

1

Click the Stacked Bar Chart icon (horizontal bars, first column)

2

Drag category → Y-axis

3

Drag Count of sales → X-axis

4

Drag subcategory → Legend. Each colour = one subcategory.

Common Mistake

Don't add too many legend items. If you have more than 8 subcategories in one stack, the colours become impossible to distinguish. Use a filter or aggregate smaller categories into "Other".

Power BI — Chart 04

Treemap: Top Customers by Profit

Who are the highest-value customers? Use filters to focus on the top tier.

Business Question

"Which customers should Superstore focus retention efforts on?"

1

In the Visualizations pane, select the Treemap icon (looks like nested squares).

2

Drag customer → Category field well.

3

Drag profit → Values field well. Each customer gets a rectangle — bigger = more profit.

4

Add a filter: In the Filters pane, find Sum of profit. Change the filter type to "is greater than", type 1000, and click Apply filter. This removes low-value customers from the view.

Interpretation

Tamara Chand is the largest rectangle — the single highest-profit customer. The filter at >$1,000 reveals roughly 20–30 customers worth targeting for loyalty programs.

Visualizations Config

Treemap

Category

customer ×

Details

Add data fields here

Values

Sum of profit ×

FILTER — Filters on this visual

Sum of profit
is greater than → 1000
Click "Apply filter"

Output: Top Customers by Profit (filtered)

Power BI — Charts 05 & 06

Tables & Matrices

Precise numerical summaries — great for verifying calculations and cross-tabulating categories

Table — Average Profit Margin by Segment

"Which customer segment is most profitable per sale?"

1

Select the Table icon in the Visualizations pane.

2

Drag segment → Columns (or the single Rows field)

3

Drag profit_margin → Values. Click the dropdown arrow next to it and choose Average.

Segment	Avg Profit Margin
Consumer	0.11
Corporate	0.12
Home Office	0.14 ← highest
Total	0.12

Matrix — Avg Sales by Category × Subcategory

"Which specific subcategory has the highest average transaction value?"

1

Select the Matrix icon (looks like a grid with +/- expand arrows).

2

Drag category → Rows

3

Drag subcategory → Columns

4

Drag sales → Values, set to Average. Use the horizontal slider at the bottom of the matrix to scroll and see all columns.

Category	Chairs	Copiers	Phones
Furniture	503.86	—	—
Office Supplies	—	—	—
Technology	—	2,198.94	—

Blank Cells

Blank cells in a matrix don't mean zero — they mean that subcategory doesn't belong to that category. This is correct behaviour.

Power BI — Charts 07 & 08

Pie Chart & Donut Chart

Parts-to-whole — use only when you have fewer than 6 categories and the total is meaningful

Pie Chart — Sales Count by Region

"Is there a dominant region in terms of order volume?"

1

Select the Pie Chart icon.

2

Drag region → Legend

3

Drag sales → Values, set to Count. Power BI calculates percentages automatically.

Donut Chart — Total Profit by Segment

"Which customer segment generates the most total profit?"

1

Select the Donut Chart icon (next to pie chart).

2

Drag segment → Legend

3

Drag profit → Values, set to Sum. A donut works the same as a pie but leaves room for a KPI label in the centre.

Common Mistake

Don't use pie/donut charts when slices have similar sizes — humans are bad at comparing angles. Use a bar chart instead when differences are subtle.

Statistical Method 02

Time Series Analysis

Tracking how a numerical measure changes at regular time intervals

Definition

A time series is a set of observations of a quantitative variable taken at equally-spaced points in time — daily, monthly, quarterly, yearly.

What to Look For

Trend — Is the overall direction going up, down, or flat over the period?
Seasonality — Does a pattern repeat at regular intervals (e.g., sales spike every November/December)?
Outliers — Are there sudden spikes or drops that need explaining?
Structural breaks — Did something change the baseline? (e.g. COVID-19 in March 2020)

Superstore Context

"The dataset runs January 2019 – December 2022. Our line chart will show whether median sales are growing, and whether there are seasonal patterns worth forecasting."

Median Monthly Sales — Jan 2019 to Jul 2022

Power BI — Time Series Setup

Preparing Date Data for Time Series Charts

order_date loads as Text by default — you must convert it and create monthly bins before plotting

Part A — Change the Column Data Type

1

Click the Table view icon on the left sidebar (looks like a grid).

2

Click on the order_date column header to select it.

3

In the Column tools tab at the top, find the Data type dropdown (currently says "Text"). Change it to Date.

4

The column will now show dates formatted as Wednesday, 2 January 2019 etc. Power BI can now use this as a time axis.

Part B — Create Monthly Bins Column

5

With order_date still selected, go to Column tools → Groups icon (looks like a binned histogram icon).

6

In the Groups dialog: Name it order_date (bins), Group type = Bin, Bin size = 1 Month. Click OK.

7

A new column order_date (bins) appears in your data. Each row now shows January 2019, February 2019 etc. Use this for the line chart X axis.

Before vs. After

order_date (original)	order_date (bins)
Wednesday, 2 January 2019	January 2019
Thursday, 3 January 2019	January 2019
Friday, 4 January 2019	January 2019
Sunday, 6 January 2019	January 2019
Monday, 11 February 2019	February 2019
Tuesday, 12 February 2019	February 2019

Why Monthly Bins?

Plotting every individual day produces a noisy, cluttered line. Grouping into months reveals the trend and seasonality clearly — the signal without the noise.

Groups Dialog Settings

Name: order_date (bins)
Field: order_date
Group type: Bin
Bin type: Size of bins
Bin size: 1 → Months

Power BI — Chart 09

Line Chart: Median Sales Over Time

Using the monthly bins column to visualise the time series

Business Question

"Is Superstore's sales performance improving over time? Are there seasonal patterns?"

1

Back in Report view, click a blank canvas area. Select the Line Chart icon in Visualizations.

2

Drag order_date (bins) (the new column you just created) → X-axis.

3

Drag sales → Y-axis. Click the dropdown arrow next to it and change the aggregation from Sum to Median.

4

Click Focus mode to view full-screen. The X axis should show Jan 2019 → Jul 2022.

Read the chart

Notice recurring spikes in Q3–Q4 each year — the holiday season drives higher-value orders. The overall median is relatively flat (~$50–60), suggesting growth is in volume not ticket size. Hover over any point to see the exact month and value.

Visualizations Config

Line Chart

X-axis

order_date (bins) ×

Y-axis

Median of sales ×

Secondary y-axis

Add data fields here

Legend

Add data fields here

Output: Median Sales over Time

Statistical Method 03

Correlation: Does Discounting Hurt Profit?

Measuring the strength and direction of the linear relationship between two variables

What Correlation Tells You

Value	Meaning	Chart Look
r = +1.0	Perfect positive relationship	All points on a rising line
r = +0.5 to +0.9	Moderate to strong positive	Cloud of points rising left→right
r = 0	No linear relationship	Random scatter, no slope
r = −0.5 to −0.9	Moderate to strong negative	Cloud of points falling left→right
r = −1.0	Perfect negative relationship	All points on a falling line

Drawbacks of Correlation

Measures strength, not slope — high correlation doesn't tell you how much profit changes per unit of discount.
Linear only — a curved (nonlinear) relationship can look like r ≈ 0.
Two variables only — can't capture how three or more variables interact.
Correlation ≠ Causation — discounts don't cause low profit; other factors matter too.

Discount vs. Profit — Scatter Plot

Key Finding

Clear negative correlation: higher discounts = lower (or negative) profit. Orders with ≥50% discount regularly lose money. Superstore's discount strategy needs review.

Power BI — Chart 10

Scatter Plot: Discount vs. Profit

Visualising the correlation between two continuous variables

Business Question

"Does increasing the discount rate cause Superstore to lose money?"

1

Select the Scatter Chart icon in the Visualizations pane.

2

Drag discount → X-axis

3

Drag profit → Y-axis

4

Important: Click the dropdown arrow next to each field and choose "Don't summarize". This shows each individual order as a dot — not just the totals.

5

Optional: Drag category → Legend to colour-code by product category. Drag sales → Size to make larger orders appear as bigger dots.

Why "Don't Summarize"?

If you leave the default "Sum", Power BI collapses all orders into one dot — you'd see nothing useful. "Don't Summarize" keeps every row as its own data point.

Visualizations Config

Scatter Chart

X-axis

discount — Don't summarize ×

Y-axis

profit — Don't summarize ×

Legend (optional)

category ×

Size (optional)

sales ×

What to Look For

The cluster of red dots in the bottom-right confirms it: high-discount orders consistently generate negative profit. This is an actionable finding — recommend a maximum discount threshold.

Going Further

In Tableau, you can add a Trend Line via Analytics pane → Trend Line → Linear. This adds a line of best fit and shows the r² value.

Tableau

Time Series & Forecasting in Tableau

Tableau's built-in exponential smoothing forecast extends your line chart into the future

Part A — Build the Time Series

1

Open the Superstore sample workbook (Start screen → Sample Workbooks → Superstore), or connect to the CSV via the left pane.

2

Drag Order Date → Columns. Click the field pill and change to Continuous → Month (scroll down in the menu for the second "Month" option).

3

Drag Segment and SUM(Sales) → Rows.

4

Drag Segment → the Colour box under Marks. You now have three separate coloured time-series lines (Consumer, Corporate, Home Office).

Part B — Add the Forecast

5

Click the Analytics tab on the left panel (next to the Data tab).

6

Drag Forecast from the Model section and drop it onto the chart. A shaded extension appears beyond the last data point.

7

Right-click the forecast area → Forecast → Describe Forecast to see the model details (seasonal periods, trend, confidence intervals).

Forecast Chart — Sales by Segment

What the Shaded Area Means

The shaded confidence band = the 95% range of likely values. A wider band = more uncertainty. If the band is very wide, the forecast is unreliable — don't over-commit to it.

Reference Card

Common Mistakes — One per Chart Type

These are the errors that cost marks in assessments and credibility in professional reports

Chart Type	Common Mistake	Why It Matters	Fix
Bar / Column	Y-axis doesn't start at zero	Small differences appear huge — visually misleading	Always start bar charts at zero. Use a line chart if you need to zoom in.
Pie / Donut	More than 6 slices	Tiny slices are unreadable; angles are hard to compare	Aggregate small categories into "Other" or switch to a bar chart.
Stacked Bar	Too many legend colours	Viewers can't distinguish 12+ colours reliably	Limit to 6–8 segments or group smaller ones.
Line Chart	Connecting non-continuous data	Suggests a trend where there may be none	Use a line chart only for time series or naturally ordered data.
Scatter Plot	Forgetting "Don't Summarize"	You get one dot instead of thousands — no pattern visible	Always change aggregation to "Don't summarize" for both axes.
Treemap	Showing all records unfiltered	Hundreds of tiny equal boxes convey nothing	Apply a filter (e.g., profit > $1,000) to focus on the meaningful tier.
Forecast	Over-trusting a wide confidence interval	Wide band = high uncertainty; decisions based on it can fail	Report the confidence interval alongside the forecast line. Note limitations.

Summary

The Complete Picture: Question → Method → Chart → Tool

Use this as a reference card for your dashboard assessment

Business Question	Statistical Method	Chart Type	Tool
"Which subcategory has the highest sales?"	Descriptive Statistics	Bar Chart (sorted)	Power BI
"How do quantity and profit compare by category?"	Descriptive Statistics	Clustered Column Chart	Power BI
"What is the product mix within each category?"	Descriptive Statistics	Stacked Bar Chart	Power BI
"Who are the top-profit customers?"	Descriptive Statistics	Treemap + Filter	Power BI
"Which segment has the highest profit margin?"	Descriptive Statistics	Table / Matrix	Power BI
"What share of orders comes from each region?"	Descriptive Statistics	Pie / Donut Chart	Power BI
"Is sales performance growing? Any seasonality?"	Time Series	Line Chart (monthly bins)	Power BI / Tableau
"What will sales look like in the next 12 months?"	Forecasting	Line Chart + Forecast Band	Tableau
"Does discounting destroy profit?"	Correlation	Scatter Plot	Power BI / Tableau
"Which states/cities are high-growth opportunities?"	Geographical	Filled Map / Choropleth	Power BI / Tableau

Dashboard Activity

Pick three questions from this table. Build charts that answer them in Power BI. Give your dashboard a title that tells a story (e.g. "Superstore Profitability — What's Working and What Isn't"). Add slicers for Region and Year so your audience can explore.