Course : Business Problem

Solving, Insights and Storytelling

Edit Master
Lecture text styles Patterns
On : Analysing
and Storytelling

2
 Session Agendas

In this session we will discuss and

revise important concepts from
‘Analysing Patterns and Storytelling’ and
resolve doubts related to the module.

3
 4

Five Patterns of Insights

Once we have analysed the data, there are 5 ways in which we can communicate
the insights:

1. Unknown Result
- The number itself is being unknown and of significance is of
use Eg: Zimbabwe has a population of this much

1. Surprising Extreme
- Extreme values that were ideally not expected
- Largest and Smallest values
Eg: Aam Aadmi party has the highest number of criminal cases against
them

upgrad.com
 5

3. Surprising Comparison
- The comparison is unusual or surprising
Eg: Lucknow’s mortality rate is five times higher than the national mortality rate

4. Significant Outliers
- Unusual form of extreme; where the result is far more than what you
would normally expect
Eg: Mumbai’s call drop rate is five times higher than any city nearby

5. Abnormal Distribution
- An unusual pattern in a distribution that we did not expect
Eg: the number of students scoring 50,60,70 etc is much higher than students
scoring 49,59,69 etc (Teacher could be rounding off the scores)

upgrad.com
 6

Analysis Approach

The five patterns are sufficient in categorising the insights, now we need a a few
techniques to analyse your data so that you can apply the five patterns to them
and
start generating insights.

These analyses can be categorised into two types: exploratory data analysis and
hypothesis-driven analysis.

Furthermore, these analysis techniques utilise 2 methods of data manipulation

in order to extract insights - deriving new columns and summarising rows.

upgrad.com
 7

Deriving new columns:

- statistical

Statistical models often take a sample of the original data and infer from it the behaviour of the
entire population

1. Metadata Lookup: The metadata is essentially an additional dataset or a data-sheet that

is available with you that provides information on the original data.

Eg: The population of a country in a dataset could be made available in a separate sheet
and you can perform a VLOOKUP to create that column in the original dataset.

2. Calculations: You can perform a variety of calculations using the numeric columns in
your dataset.

Eg: If we have a dataset on suicide information, we can calculate the Suicide Rate (%)
from the Suicides and the Population columns.

upgrad.com
 8
3. Binning: This process essentially bins a given numeric column to specific categories.

Eg: We can convert the Suicide Rate into specific bins and categorised them as High, Medium or Low.

4. Business-Specific Metrics: This part would be specific to your domain and hence the metrics or
KPIs that you might be using would be a useful additional column.

Deriving new columns:

- machine learning

In machine learning models you run algorithms on a set of predefined data called "train data" to
formulate the model and then run it again on another set of data called "test data" to test the model's
accuracy and precision.

1. Classification
2. Clustering
3. Time Series Analysis
4. Feature Extraction
5. Sentiment Analysis

upgrad.com
 9

Summarising rows:
1. Summarising the Numeric Columns Eg:

2 Grouping Eg:

upgrad.com
 10

3. Pivot table Eg:

4. Machine Learning
- Apply a model and the model itself gives us some additional information

Eg: Predict the sales of one product on the basis of sales of three other products and
calculate their importance.

upgrad.com
 11

5. Use relationships between the columns like Correlation, etc. Eg:

upgrad.com
 12

The Analysis Approach Flow

Derive new
columns

Summarise the
Rows

Apply the five patterns of

insights

Check if it leads to something that

is useful and interesting.

upgrad.com
 13

Effective Communication: The Pyramid Principle

The pyramid principle is very effective in communicating insights, in simple terms it

means putting the outcome first and the supporting statements behind it.

upgrad.com
 14

The Pyramid Principle is crucial in business communication because of 2 specific reasons:

1.You save time: You get to the point immediately thereby not wasting people’s time. This
is crucial because executives have a limited time-span to devote to anyone and hence it is
important that you feed them information that they need to know. If required, do drill down
to the finer details.

2.Your message is concise: Similar to the newspaper headlines, the structure of the
insight should be such that you are able to pass on the most useful information in the
most economical way possible and the pyramid principle helps you achieve just that.

upgrad.com
 15

Effective Communication: Logical Flows and Usage of Words

If the supporting arguments are ordered in a logical structure, people tend to remember more.
Thus the six essential flows needed for communication are:

● TIME
● PLACE
● THREE ASPECTS OF THE SAME THING
● BENEFITS
● SCALE
● COUNTER-ARGUMENTS

The usage of words is pretty critical in communicating the insights. The three pillars of using
words effectively are:

● Being concise
● Using active voice
● Using a positive tone to deliver the message.

upgrad.com
 16

Checklist - Business Communication

Make sure that you have the following information with you before you send out any form of
business communication:

1. IS THIS MESSAGE RELEVANT?

2. IS THERE A NON-OBVIOUS INSIGHT?
3. WHAT IS THE ONE LINE SUMMARY?
4. WHAT ARE THE SUPPORTING STATEMENTS?
5. DO THE STATEMENTS FOLLOW A LOGICAL FLOW?
6. ARE THE WORDS CONCISE?

upgrad.com
 17

Data Storytelling
Data storytelling is the practice of building a narrative around a set of data and its
accompanying visualizations to help convey the meaning of that data in a powerful and
compelling fashion.

Why Data Storytelling?

1. Statistics show that only 5% people remember numbers, whereas 63% people
remember the story.
2. Storytelling is more persuasive than only statistics.

The four main elements of storytelling are:

● Message
● Visuals
● Structured flow
● Narrative.
upgrad.com
 18

Power of Data Visualization

Different forms of the same data is given below, try to come up with your own insights from
each of them.

upgrad.com
 19

Type of Variables
1. Qualitative Variables

N ominal: Used for labelling variables without any scale. There is no overlap or
order. Eg: Gender of a person

O rdinal: They are basically nominal variables with an underlying

order. Eg: Food ratings

1. Quantitative Variables

I ntervals: Numeric measures where the underlying order and the difference is known.
Eg: Temperature

Ratios: Numeric measures which are basically interval scale variables along with
an absolute zero allowing several forms of statistical analyses to be performed.
Eg: Height

upgrad.com
 20

Visualisation of Quantitative Data

1. Scatterplot

Scatter plot enables you to analyse the data in the following ways:

● You can study how two numerical variables are related

● You can study the way in which the data is distributed within a specific interval by using
the dot density.

Eg:

upgrad.com
 21

2. Line Chart

Line chart has the following properties:

● You measure an interval over a period of time.

● Useful for time series analyses like finding trends and showing the relationships
between them.
Eg:

upgrad.com
 22

3. Histograms and Frequency Polygons

Histograms are useful for visualising the distribution of numerical data and make
inferences from them.

Eg:

upgrad.com
 23

Visualisation of Qualitative Variables

1. Pie Chart

In the chart, each category would be represented by a different colour. The sector
covered by that category would represent the percentage share it occupies with respect
to the data that is being analysed.

Eg:

upgrad.com
 24

1. Bar Chart and Stacked Bar Chart

● The rectangles in a bar chart represent the number of observations belonging to

each category. Useful when the number of categories is too high for a pie chart.

● When we further divide the rectangles in a bar graph into specific categories, the
visualization is referred to as a stacked bar chart. This is useful when the given bar
chart is unable to offer insight at a sub-category level.

Eg:

upgrad.com
 25

Visual Design Principles and Storyboarding

The two visual principle ideas are the concepts of :

● Trade-off between Accuracy and Precision: This is done to reduce the complexity of the
figures and numbers that are being presented and thereby reduce the amount of
attention needed on the part of the audience to grasp the information.

● Drawing attention: This can be done in several ways - by highlighting the significant
figures, using different fonts, boldening the key values, etc.

The concept of storyboarding has the following advantages:

● Assimilates the information and clarifies the logic of the hypothesis

● Identify gaps in Analysis
● Avoids Redundant Work

upgrad.com
 26

Any Queries?

Thank
You!

upgrad.com