Think like a

Data Scientist

A workbook for driving collaboration between the business and data science teams.

2

Key Business InitiativesA Business Initiative is a cross-functional plan or program that is typically 9 to 12 months in duration, with well-defined financial or business metrics, that supports the entity’s business objectives.

•Sense of urgency (9-12 Months)

• Important to the business

•Compelling ROI

•Clear business ownership

•Strong IT leadership

•Bounty of potential data sources

•Analytics friendly

Examples of business initiatives:The Walt Disney Company might have as key business initiatives to “leverage the MagicBand to increase guest satisfaction by 15%” or “leverage the MagicBand increase attendance at Class B attractions by 10%.”

Whereas Business Objectives describe what an organization expects to accomplish over the next 1 to 3 years. The Business Objectives for The Walt Disney Company might include introduce the MagicBand, successfully launch the “Pirates of the Caribbean: Dead Men Tell No Tales” and “Star Wars Episode VIII” movies, and successfully launch the new Disney World “Avatar: Flight of Passage” attraction.

3

STEP 1 | Identify Business Initiatives

•

•

•

•

•

•

•

List info sources:

•

•

•

•

•

EXAMPLE: Increase same-store sales

EXAMPLE: Blog.chipotle.com

EXAMPLE: Analyst reviews

Chipotle: Store Manager Persona

Overview: Typically 2 to 4 years of chipotle experience; promoted through the ranks to store manager within same store

Quote: “I love chipotle, so working here only seemed natural”

Chipotle stakeholder: Store managerChipotle business initiative: Increase same store sales

Key Decisions Supporting Business Questions Pain PointsStaffing • What’s the likely store traffic throughout the day?

• When and for how long are the local events?Lacks real-time model that allows him to update changing demand variables

Inventory • Given store traffic estimates, how much additional inventory will I at what points in the day?

• What local events that could impact inventory?• What are the types of local events?

Doesn’t have insights into extra inventory at neighboring stores that might fill temporary inventory gaps

Customer Satisfaction

• What’s the best comp to give unsatisfied customer?• How important is that customer to me?• What is the cost of the different comp options?

Doesn’t know “how valuable” that particular customer is to the Chipotle

Production • Do I need to shift staff to produce more materials given the store traffic and time of day?

• Which products are most popular for local events?

Doesn’t have insights into local events that could dramatically impact production needs

4

Key Business StakeholdersBusiness Stakeholders are those business users or business functions (Sales, Marketing, Finance, Logistics, Customer Services, etc.) that either impact or are impacted by the targeted business initiative. It is from this base of users that we will collaborate to identify and prioritize the key business decisions necessary to support the targeted business initiative.

We want to develop personas for each of the business stakeholders to better understand their work characteristics and job characteristics. Understanding this helps to capture the decisions and questions that these stakeholders must address with respect to the targeted key business initiative.

A persona is a one to two page “day in the life” description that makes the key business stakeholder “come to life” for the data science and user experience (UEX) development teams. Personas are useful in understanding the goals, tasks, key decisions, and pain points of the key business stakeholders. The persona helps the data science team to identify the most appropriate data sources and analytic techniques to support the decisions that the business users are trying to make and the questions that they are trying to answer. Personas are created for each type of business stakeholder affected by the given business initiative.

Chipotle: Store manager persona

5

STEP 2 | Identify Key Business Stakeholders

EXAMPLE: Store Management – responsible

Identify those business stakeholders who either impact or are impacted by the targeted business initiative. Write a brief description of each.

•

•

•

•

•

•

•

•

•

•

•

•

•

•

for day-to-day store operation

6

Business EntitiesBusiness Entities are the physical entities (e.g., customers, products, employees, students, patients, parolees, trucks, wind turbines, jet engines) around which organizations seek to uncover or quantify analytic insights. For Chipotle, the key business entities could include customers, employees, products, and suppliers.

Examples of business entities include:

•Customers

•Patients

•Students

•Employees

•Stores

•Products

•Medication

•Trucks

•Wind turbines

7

STEP 3 | Identify Business EntitiesSTEP 3 | Identify Business Entities

Identify the business entities around which we want to uncover or quantify analytic insights such as behavioral characteristics, tendencies, propensities, interests, associations and affiliations. Write a brief description of each.

•

•

•

•

•

•

•

EXAMPLE: Stores, products, competitors•

•

•

•

•

•

•

EXAMPLE: Stores, products, competitors

8

Business DecisionDecisions are defined as a conclusion or resolution reached after consideration or analysis that leads to action (where the action may actually be a “take no action” decision). For Chipotle, “Determine how much product to have on supply for the day” is an example of a business decision.

What decisions do the business stakeholders need to make about the business entities in support of the targeted business initiative? What data insight would support those decisions? These help to form the basis for generating an actionable analytics recommendation that can accelerate a targeted key business initiative.

Capturing and validating these decisions is critical to the “Thinking like a Data Scientist” process. Leading organizations like Uber and Netflix are disruptive because they build a business model that seeks to simplify their targeted customers’ key “decisions.” For example, one of the customer decisions Uber addresses is “How do I easily get from Point A to Point B?” Another example, one of the customer decisions Netflix addresses is “What content (movie, TV show) can I easily watch tonight?”

9

STEP 4 | Capture Business Decisions

Document business stakeholder key decisions and write brief descriptions.

•

•

•

•

•

•

•

EXAMPLE: How much to stock? When

should I order? How much staffing is

required?

10

Brainstorm Business QuestionsQuestions are sentences worded or expressed so as to elicit information, but not necessarily to elicit an action or drive a decision. For Chipotle, questions could include “What are our top selling products on weekends?” or “What days of the week are the slowest?” or “What combinations of ingredients are most popular?”

What happened?Descriptive analytics

•What were revenues last week?

•How many customers visited the store during last Sunday’s Farmer’s Market?

•What products sold more than the week before?

•How many employees did we hire last month?

What is likely to happen?Predictive analytics

•What will be revenues next week?

•How many customers will visit the store during next Sunday’s Farmer’s Market?

•Which products will likely sell the most next week?

•How many new employees will we need to hire next month?

What should we do?Prescriptive analytics

•Run BOGOF Burrito promotion Wednesday 7–9pm to attract 55–75 more college students.

•Add 2 more workers 11:00am–2:00pm and 5:00pm–9:00pm on May 5.

• Increase chicken inventory next week by 15%.

• Increase hiring pipeline next month by 20 candidates.

11

STEP 5 | Create Predictive Questions

(What happened?)

(What is likely to happen?)

•

•

•

•

•

•

•

•

EXAMPLE: How many customers will

visit this week?

EXAMPLE: How many customers will

visit next week?

Descriptive questions Predictive questionsThe purpose of this exercise is to convert Descriptive questions about what happened into Predictive questions about what is likely to happen.

12

Predictive Variables and Metrics“Data science is about identifying those variables and metrics that might be better predictors of performance”.

The Predictive Variables and Metrics Exercise is a technique for leveraging the predictive questions captured in Step 5 to brainstorm the variables and metrics that might be better predictors of performance.

For example, the descriptive question “How many customers visited the store last week?” in Step 5 was transformed to the predictive question of “How many customers do we think will visit the store next week?”

Then the Predictive Questions Exercise reframes the predictive question as such:

What data might you need in order to predict... how many customers do we think will visit the store next week?

Then brainstorm and list the data sources you might want to have in order to answer that predictive question.

13

STEP 6 | Identify Predictive Variables and Metrics

••

•

•

•

•

•

•

•

•

Predictive variable and metrics

EXAMPLE: Store location

EXAMPLE: Current economic

conditions

EXAMPLE: Seasonality

What data might you need in order to predict... how many customers do we think will visit the store next week?

14

Creating ScoresScores are a dynamic rating or grade standardized to aid in comparisons, performance tracking and decision-making. Scores help to predict likelihood of certain actions or outcomes (e.g., FICO score measures the likelihood of a borrower repaying their loan). Scores are actionable, analytic-based measures that support the decisions your organization is trying to make, and guide the outcomes the organization is trying to predict.

The purpose of the “Score” technique is to look for groupings of strategic noun dimensions and attributes that can be combined to create a more predictive and actionable score.

These scores are critical components of our process by supporting the decisions that we are trying to make, and/or what actions or outcomes we are trying to predict with respect to our targeted business initiative.

Scores are very important constructs in the world of data science, and can help to cement the business stakeholders’ buy-in to the data science process.

A familiar example of scoring is the FICO score, which combines multiple questions and dimensions about a loan applicant’s finance history to create a single score that lenders use to predict a borrower’s ability to repay a loan.

In this exercise, group the variables and metrics from Step 6 into common subject areas that might be the basis for creating scores that support the key decisions from Step 4.

15

STEP 7 | Create Actionable Scores

•

•

•

•

•

•

•

•

•

•

•

•

•

•

•

•

•

•

•

•

•

•

Potential scores

EXAMPLE: Economic potential

EXAMPLE: Local vitality

STEP 8: Map scores to decisions

Chipotle business initiative: Increase same store sales

Map potential scores (and supporting metrics) to the recommendations necessary to support key business decisions

Decis ions

S taffing

Inventory

P romotions

E xample R ecommendations

• How many people to staff?

• What special skills do I need?

• What are the staffing times?

• How much food inventory?

• How much utensils inventory?

• What local events to sponsor?

• What coupons to offer?

• At which local events to offer?

S cores/ M etrics

E conomic P otential• Local demographics• Local house values• Local economy• Local unemployment rate

Local Vitality• Miles from high school• Miles from mall• Miles from business park• # Local sporting events• # Local entertainment

16

Putting Analytics Into ActionFacilitate the development of a compelling and actionable user experience by starting with a simple “Recommendations Worksheet.”

•The Recommendations Worksheet ties the decisions that our business stakeholders need to make to the predictive analytics or scores that the data science team is going to need to build.

•The “Recommendations Worksheet” captures the potential recommendations that could be delivered to the busi-ness users (or consumers) in support of those decisions.

•Finally, the worksheet captures the potential scores (and the supporting variables and metrics) that can be used to power the recommendations.

Building upon the relationship between identifying variables (step 6) to creating scores (step 7) based upon the key decisions (step 4), now we want to identify the specific analytic recommendations (prescriptive analytics) to make the variables and scores actionable.

•First, list the decisions captured in Step 4.

•Second, brainstorm the recommendations that one could deliver to the Business Stakeholder in support of that decision.

•Finally, brainstorm the analytic scores - and the supporting variables and metrics - that could be used to drive the recommendations.

Step 8: Map scores to decisions

STEP 8: Map scores to decisions

Chipotle business initiative: Increase same store sales

Map potential scores (and supporting metrics) to the recommendations necessary to support key business decisions

Decis ions

S taffing

Inventory

P romotions

E xample R ecommendations

• How many people to staff?

• What special skills do I need?

• What are the staffing times?

• How much food inventory?

• How much utensils inventory?

• What local events to sponsor?

• What coupons to offer?

• At which local events to offer?

S cores/ M etrics

E conomic P otential• Local demographics• Local house values• Local economy• Local unemployment rate

Local Vitality• Miles from high school• Miles from mall• Miles from business park• # Local sporting events• # Local entertainment

17

STEP 8 | Map Scores to Decisions

Stakeholder:

Business initiative:

Business decisions Potential recommendations Scores/metrics

•

•

•

•

Score

Score

Score

••

18

Additional Worksheets

19

20

Identify Data SourcesThroughout this exercise, participants are going to uncover all sorts of new data sources that “might” provide value with respect to the targeted business initiative and the key business decisions. Data sources are likely to come from:

Historical operational and transaction systems data (ERP, Financials, HR, Supply Chain, Sales Force Automation, Marketing, etc.) for which data is captured, but likely not on readily accessible platforms.

Internal unstructured data sources including email conversations, consumer comments, clinical studies, research papers, and notes from employee and customer interactions.

External data sources including social media, newsfeeds, weather, traffic, economics, research papers, white papers, and public domain data from government and college institutions.

Potential Chipotle Data Sources

•Point of Sales Transactions

•Market Baskets

•Product Master

•Store Demographics

•Competitive Stores Sales

•Store Manager Notes

•Employee Demographics

•Store Manager Demographics

•Consumer Comments

•Weather

•Traffic Patterns

•Yelp

•Zillow/Realtor.com

•Twitter/Facebook/Instagram

•Twellow/Twellowhood

•Zip Code Demographics

•EventBrite

•MaxPreps

•Mobile App

•…

21

Identify Data Sources

•

•

•

•

•

•

•

•

•

Client key business initiative:

Data Source Increase

Increase Shopping Bag

Revenue

Increase # Corporate

Events

Increase Promotional Improve NPI

Point of Sales Transactions 4 4 4 4 4

Market Baskets 4 4 2 4 4

Store Demographics (Zip Code) 3 3 1 3 3

Local Competitive Stores 2 2 2 2 2

Store Manager Demographics 1 1 3 1 1

Consumer Comments 3 3 3 3 2

Social Media 2 1 1 3 3

Weather 3 1 1 1 1

Local Events 4 2 1 2 1

3 1 1 2 1

Zillow 1 2 2 2 2

22

Assess Data ValueOnce a wide range and variety of data sources have been identified throughout the “Thinking like a Data Scientist” process, it is now time to determine the value of the different data sources vis-à-vis the business value that data source brings in support of the key business decisions. From a process perspective:

•List the data sources as row headers in the first column

•List the key business decisions as column headers in the first row (Note: this will likely require a couple of pages)

As a group, assess the business value of each data source with respect to the decision. I recommend using Harvey balls (scale of 0 to 4) because it produces a very easy to read chart. The Harvey ball font can be found at: http://www.ambor.com/public/hb/harveyballs.html

Data value assessment

23

Assess Data Value

Client key business initiative:

Use Case#1

Use Case#3

Use Case#2

Use Case#4

Use Case#5

Data Source

Use cases are clusters of decisions around a common subject area.

Data Source Ease of

Acquiring Cleanliness Accuracy Granularity Cost

Point of Sales Transactions 4 4 4 4 4

Market Baskets 4 4 4 4 4

Store Demographics (Zip Codes) 4 4 4 4 4

Competitive Stores Sales 2 2 2 2 1

Store Manager Demographics 4 4 4 4 4

Consumer Comments 2 2 1 2 3

Social Media 1 1 1 1 1

Weather 3 3 2 2 2

Local Events 1 2 2 2 1

2 2 1 3 2

Zillow 1 1 2 2 1

24

Assess Implementation FeasibilityNext, we want to determine the implementation feasibility of the data sources over the next 9 to 12 months. There are a number of feasibility that need to be considered. We have jumpstarted the process by pre-identifying some of the more common impediments:

•List the data sources as row headers in the first column

•List the key business decisions as column headers in the first row. (Note: this will likely require a couple of pages)

Have the IT group assess the implementation feasibility of each data source over the next 9 to 12 months. I recommend using Harvey balls (scale of 0 to 4) because it produces a very easy to read chart. The Harvey ball font can be found at: http://www.ambor.com/public/hb/harveyballs.html

Implementation feasibility assessment

25

Implementation Feasibility Assessment

Data Source Ease ofAcquiring Cleanlness Accuracy Granularity Cost

26

Prioritization MatrixThe Prioritization Matrix is a vehicle for driving IT and business stakeholder alignment around the top priority use cases. The Prioritization Matrix provides a framework around which the business users can debate and decide the business value (vertical axis) of each use case in relation to the other use cases. Simultaneously, the IT stakeholders can debate and decide the implementation feasibility (over the next 9 to 12 months) of each use case in relation to the other use cases.

Each use case is placed on a separate Post-it note.

The workshop participants debate and decide the placement of each use case on the prioritization matrix based upon business value and implementation feasibility (over the next 9 to 12 months) vis-à-vis the placement of the otheruse cases.

Prioritization matrixUse cases

Increase Shopping Basket size

Increase Corporate events

Improve New Product Introductions

Increase Special Events (birthdays, parties)

Hi

Hi

Lo Implementation Feasibility

Bus

ines

s V

alue

Increase same store sales

C

A F

B

D

E

A.A

B.B

C.C

D.D

E.E

F.F

27

Prioritization MatrixB

usin

ess

Implementation Feasibility

Lo Hi

Hi

Use CaseUse Case #1

Use Case #2

Use Case #3

Use Case #4

Use Case #5

Use Case #6

A

B

C

D

E

F

Business Initiative:

Copyright © 12/2016 Dell Inc. or its subsidiaries. All Rights Reserved.