Week 2: Data Driven Decision Making and Data Storytelling

Agenda

D³M

• Why? cognitive biases

• How? develop a process

Data storytelling

• Why? it works

• Intro on how to tell a data story

D³M

Fundamental problem

 

 

People that run businesses, government, or other organizations are striving to reach some goal and need to adapt to changing conditions

Data analysis as part of the solution

Instead of relying on experience or gut instinct alone, businesses can use careful data analysis to help guide their decisions (e.g., google, amazon)

Objective data analysis is less susceptible to psychological biases

However, data analysis alone is rarely the solution

Cogitive or psychological biases

Predictable mental errors that arise from our limited ability to process information objectively

• Examples: confirmation bias, anchoring, gambler’s fallacy

Strong positive and negative emotions can influence decisions (Bucurean, 2018)

Behavioral economics studies why people behave the way they do

Confirmation bias is the tendency to search for, interpret, focus on and remember information in a way that confirms one’s preconceptions.

The anchoring bias, or focalism, is the tendency to rely too heavily—to “anchor”—on one trait or piece of information when making decisions (usually the first piece of information acquired on that subject).

Gambler’s fallacy, the tendency to think that future probabilities are altered by past events, when in reality they are unchanged

Empirical evidence:

source: https://blog.happyfox.com/data-driven-decision-making-using-data-to-fuel-growth/

D3M relies on facts in the data: Need to ensure quality and fidelity of the data

Choice are tested: The decision to go with a certain marketing strategy, color scheme, or phrase is based on data analysis demonstrating that the choice is better than the alternatives at achieving the objective

Inspired by experience but not completely reliant on it. There is context and information not always measured in data

D3m is risk aversing because decisions are made based on evidence and not emotions or gut feeling

Directive vs corrective because the organization is being proactive and making informed decisions rather than reacting to negative outcomes. This ad campaign didn’t work so we should scrap it and start another one.

D³M is expanding rapidly

Figure 1. Adoption of Data-Driven Decision-Making in US Manufacturing (Brynjolfsson & McElheran, 2016, AER)

D³M is a process

D³M process overview

1. Define an objective

2. Establish a hypothesis

3. Identify data need, build data process, collect data

4. Analyze the data

5. Communicate insights

Adapted from https://www.smartsheet.com/data-driven-decision-making-management

1. Define question or objective

What is the business or operational question you are trying to answer?

• Do we sell our product now or hold out for a higher price?

• Do we contract for feed or buy on the spot market?

The USFS spends a lot of money on wildfire suppression. Their objective is to spend as little as possible while protecting life and property. They want to identify which fires are the most expensive and understand why in hope that they can spend less in the future.

USFS is interested in suppression expenditures after a large fire occurs.

They use a regression model to study what factors explain variation in the suppression costs of past fires, then use that model to predict the suppression cost of a single fire. The predicted cost is compared to the actual cost.

You Do It: Define question or objective

In Pairs:

Identify a business that you might be interested in starting some day or any organization that you might want to work for. The business should be related to agriculture and/or natural resources.

1. Briefly describe this business (give the name if it already exists)

2. Define a question or objective that business might face. Try to be specific.

Include responses to both 1 and 2 in your iClicker response (140 characters).

2. Establish hypothesis

What do you think the answer is and why?

You should be able to describe the mechanics of the system and explain your hypothesis

Wildfire management is complex. Some factors that drive resource use (and cost) are determined by human decisions while others are biophysical. Some factors can be anticipated while others cannot. Hypothesis: there are decisions about resource use that lead to higher-cost fires.

You Do It: Establish hypothesis

In Your Same Pairs:

Based on your business question or objective, what do you think the business would need to do to achieve this goal, and why?

Write a short response in the iClicker field (140 characters).

3. Data

What data do you need to answer your question?

How will you assemble the data? Do you need equipment?

What resources do you need?

Will the data be reliable?

Collect data in a way that it can be analyzed - not sticky notes.

The USFS collects data on expenditures. Expenditures tend to be tracked well because someone has to pay those bills. It also collects data on factors that are expected to influence fire behavior and management strategy:

• vegetation, topography, weather
• values at risk like homes

You Do It: Data

In Your Same Pairs:

What data do you need to answer your question?

List the data sources that come to mind in the iClicker field.

Think about: How will you collect and assemble these data? Do they already exist?

4. Analyze data

Design an analysis strategy to answer your question

Model different scenarios

What can be changed to better achieve your objective?

They use a regression model to study what factors explain variation in the suppression costs of past fires, then use that model to predict the suppression cost of a single fire. The predicted cost is compared to the actual cost. Fires that exceed their expected costs can be interrogated more closely.

5. Communicate insight, implement, restart

Use the data to tell a story - develop visuals that show the audience the consequence of the current decision and how a different one could change the outcome

Articulate the benefits of the change

Implement the change and restart the process

Develop a graphic to show which fires lie above the regression line of cost on X. Then, look for patterns across the fires that cost more than expected. Find what they have in common.

You Do It: Analyze and Communicate

In Your Same Pairs:

1. How would you analyze these data to answer your question? What software would you use? What statistics would you construct?

2. How would you communicate your findings to other people in the business? What visualizations might be useful?

D³M across scales

People make decisions all day every day - some are big and consequential and some are small.

Data can help make better decisions across the spectrum, and in some cases, may be the only determinant.

• Would you trust an algorithm to make decisions based on data without human input?

• How much would you rely on data for strategic decisions?

How about a thermostat at home or cruise control in your car

Data storytelling

Importance of storytelling

Storytelling is fundamental to human existence. We connect, share, teach and learn through stories.

Statistics provide evidence but we need stories to truly communicate the insights to the audience

Our brains are predisposed to remember stories

Let’s look at an example: The minimalist version

source: https://medium.com/@ccr20/post-9-the-functional-art-1f0ee13b225e

Let’s look at an example: With pizzazz

source: https://timharford.com/books/datadetective/

3 elements of a data story

Source: Dykes, B. (2019). Effective Data Storytelling: How to Drive Change with Data, Narrative and Visuals (1st edition). Wiley.

• Explain: narrative and data help explain what is happening. Context and commentary guide the audience through the meaning of what is being shared.

• Enlighten: a picture is worth a thousand words or numbers. Graphs and charts help you see trends and patterns. We will discuss some of the theory and best practices for making effective graphics. You want to make it as simple as possible for the audience to see your insight.

• Engage: narrative and visuals help engage the audience. We spend billions on visual media entertainment. The best movies may have good effects but definitely have a good story.

Data story ≠ data visualization

While clear and appealing data visualizations are part of a data story, they are not data stories by themselves.

Data stories bring your analytical insights to life

Six elements of a data story

1. Data foundation

2. Main point

3. Explanatory focus (vs exploratory)

4. Linear sequence

5. Dramatic elements

6. Visual anchors

1. Story is grounded in data - something real that is observable

2. The story needs to have a main point. Have something to say.

3. Your goal is to explain some insight. What is going on? Who are the actors? What are the mechanisms? This relates to what we will discuss about causal inference.

4. Your story will consist of a series of supporting points or information that are tied together to build to some larger point. The supporting information should build: a) consumer prefs are shifting to favor Kale, b) we offer more Kale products and sales increase.

5. Dramatic elements include characters, setting, plots. These elements help humanize the data story and make it relatable. Set some context: sales are $200k. Is that good or bad? It depends on previous sales. Be selective about the information - only include information that helps to advance the narrative.

6. Visual anchors are the charts and graphs or other visuals you rely on to tell the story. We will spend considerable time on developing effective visuals to help tell your story.

Review Question 1

Which of the following is NOT one of the 5 steps in the D³M process?

1. Define objective

2. Analyze data

3. Apply intuition

4. Communicate insights

Review Question 2

Which of the following is one of the reasons it is important to use data when making (important) decisions?

1. Gut instincts are often correct

2. Data analysis is less susceptible to biases

3. Data analysis is more susceptible to biases

4. Annecdotal evidence is more convincing when communicating insights

References

Brynjolfsson, Erik, and Kristina McElheran. “The rapid adoption of data-driven decision-making.” American Economic Review 106.5 (2016): 133-39.

Dykes, B. (2019). Effective Data Storytelling: How to Drive Change with Data, Narrative and Visuals (1st edition). Wiley.

Harford, T. (2022). The data detective: Ten easy rules to make sense of statistics. Penguin.