WELCOME SAVING MILLIONS WITH WORKFORCE OPTIMIZATION APPLIED - - PowerPoint PPT Presentation

WELCOME

SAVING MILLIONS WITH WORKFORCE OPTIMIZATION

APPLIED ARTIFICIAL INTELLIGENCE WITH: BELLE TIRE + DOMO + RXA

Don Barnes III, Ben Schein, Jason Harper

3

BEN’S BIO: EVOLUTION OF A DATA DORK

Philosophy, Politics and Economics (no data) 6 years at database software start up: training, coding, selling (and cleaning the bathrooms) MBA 5 years breaking things in finance at Target (push the envelope of what is given to me) 5 years building analytic solutions at scale (stop breaking things and start building a bigger envelope)

• Target.com Business Intelligence Analytics and Testing (BIAT): 2013-2015
• Enterprise Data BI and Analytics (EDABI) Center of Excellence: 2015-2018

Domo: VP, Data Curiosity (June 2018 - present): working as product and data culture evangelist based in the office of the Chief Technology Officer

4

A Leading applied artificial intelligence company from Ann Arbor, MI A LEADING Domo Implementation Partner

JASON HARPER

Founder & Chief Executive Officer

5

6

Jason Harper Founder & CEO Jon Prantner Co-Founder & CAO Jon Kuznicki COO Tom Stanek President

RXA LEADERSHIP TEAM

Jeff Doak CTO Heather Reed CXO

Come say ‘hi’ and get free swag (DELIVERED*) at our virtual booth! www.RXA.io/booth

* While supplies last :)

7

Since 1922 120 Locations 2500 Employees

DON BARNES, III

President & Chief Tire Guy

8

BEN

Driving the cultural change required to make AI and machine learning initiatives successful

HONOR THE HUMAN INTELLIGENCE

• Empower users to engage with the

detailed output of machine learning/AI algorithms

• Ask for feedback based on the

human experience and business context

• Help humans understand their

continued role in improving the business

• Avoid creating a “black box”

11

AVOID THE BLACK BOX

• Humans hate a ”black box” so make your work accessible
• Show your work so that they understand what you are doing
• Do not over summarize
• Allow for human inputs and tuning
• Include flexibility to handle different business contexts (new

customers, new lines of business, etc)

12

JASON

WHAT REALLY IS AN AI SOLUTION?

14

WHAT REALLY IS AN AI SOLUTION?

Everything from a simple customer churn forecast to an autonomous pizza delivery vehicle. It could be as fancy as you like, but it really boils down to using data to make better decisions, faster.

15

IDENTIFYING WHEN YOU’RE READY FOR AI

One of the most frequent reasons for not diving into AI that we hear today is: “we’re just not ready yet” If you are watching this video today, you are ready for AI. Get started by identifying your business

• bjectives and mapping them to KPIs that we

can measure. There’s a proven framework for this…

16

HOW TO IMPLEMENT AND DRIVE ADOPTION OF AN AI SOLUTION

17

DRIVE ADOPTION : ANALYTICS PLAN

Download here: www.Rxa.Io/domo

18

WHAT SKILL SETS DOES MY TEAM NEED FOR AI

The dirty little secret of data science is that it’s very much an art. When you are building out your teams it’s important to recognize that successful teams balance technical skills with creative ability. And the glue that will hold your team together is trust.

19

DATA SCIENTIST JOB DESCRIPTION

The ideal candidate's favorite words are learning, data, scale, and agility. You will leverage your strong collaboration skills and ability to extract valuable insights from highly complex data sets to ask the right questions and find the right answers. Responsibilities Analyze raw data: assessing quality, cleansing, structuring for downstream processing Design accurate and scalable prediction algorithms Collaborate with engineering team to bring analytical prototypes to production Generate actionable insights for business improvements Qualifications Bachelor's degree or equivalent experience in quantitative field (Statistics, Mathematics, Computer Science, Engineering, etc.) At least 4 years' of experience in quantitative analytics or data modeling Deep understanding of predictive modeling, machine-learning, clustering and classification techniques, and algorithms Fluency in a programming language (Python, C, C++, Java, SQL) Familiarity with Big Data frameworks and visualization tools (Cassandra, Hadoop, Spark, Tableau) Areas of Desired Expertise Include Machine Learning Deep Learning Data Mining Customer Scoring Data Analysis Predictive Analytics Predictive Modeling NLP Text Mining R Artificial Neural Networks (ANN) Market Mix Modelling (MMM) Statistical Data Mining Python SQL

DOWNLOAD HERE: WWW.RXA.IO/DOMO

20

WHAT DO I NEED TO DO TO MAINTAIN AN AI SOLUTION

The skill set shifts a bit. Obviously, you need to drive adoption throughout the AI deployment Lifecycle, and that comes with trust. Domo provides the systems to monitor when data inflows and outflows break and can be set up to monitor when algorithms need adjustments and re-scoring… and of course we are tracking end users use and looking for changes there to alert us if there are adoption issues.

21

CASE STUDY: WORKFORCE OPTIMIZATION

22

DON

2017 Labor Costs “Felt” Too High.

24

“THIS PROMOTION IS ALWAYS THE BEST OF THE YEAR...I HAVE TO MAKE SURE I HAVE LOTS OF PEOPLE WORKING.”

25

THE PROBLEM:

Staffing issues

• n the clock and underutilized
• or-

short staffed resulting in longer wait times and lost revenue

26

THE SOLUTION:

WE PARTNERED WITH RXA & DOMO TO DELIVER THEIR LABOR OPTIMIZATION PLATFORM

27

STEP 1

28

The focus was on deploying the pre-built solution, but customizing it to exactly fit our business. Using the Workforce Optimization algorithms as the foundation, we built a plan to

• ptimize our labor that was

focused exactly on our needs.

ANALYTICS PLAN HIGHLIGHTS

29

BUSINESS INITIATIVE

Reduce Bottom Line Expenses

BUSINESS PROBLEM

Employees are on the clock and underutilized

BUSINESS MEASURE

Hourly labor utilization

BUSINESS OUTCOME

Reduction in labor expense

1 2 3 4

ASSESS

30

• Store Attributes

○

Date Opened

○

# of Bays

○

# of Customers Served

○

# of each employee type

○

Part Delivery Schedule

• Historic Timeclock data

○

Employees on clock by job code

• n a 30 minute basis
• Historic Order Track data

○

Employees actively working on an RO by job code on a 30 minute basis

○

Work Order by job code on a 30 minute basis

• Historic Weather Data and Weather Forecasts

ASSESS

31

• Store Attributes

○

Date Opened

○

# of Bays

○

# of Customers Served

○

# of each employee type

○

Part Delivery Schedule

• Historic Timeclock data

○

Employees on clock by job code

• n a 30 minute basis
• Historic Order Track data

○

Employees actively working on an RO by job code on a 30 minute basis

○

Work Order by job code on a 30 minute basis

• Historic Weather Data and Weather Forecasts

OTHER FACTORS.

32

INITIAL ASSESSMENT CUSTOMERS WAITING & UNNECESSARY LABOR COSTS

33 Optimal Staffing Opportunity

46%

46% of hourly shifts are not properly

• staffed. Longer customer wait times,
• r staff standing around.

Over Staffing

29%

29% of hourly shifts are overstaffed, directly resulting in unnecessary labor expenses.

MODELING

34

Distributed Random Forest (DRF) is a powerful classification and regression tool. When given a set

• f data, DRF generates a forest of classification or

regression trees, rather than a single classification

• r regression tree.

Sample Model Description:

• 500 Trees in Each Forest
• Maximum Depth of 13
• Minimum Depth 7
• Maximum Leaves 33
• Staff On The Clock
• MSR (Mean Square Residuals) 92%
• Amount of Variance Explained 65%
• Staff Actively Working
• MSR (Mean Square Residuals) 11%
• Amount of Variance Explained 95%

IMPLEMENTATION

35

It’s CRITICAL to meet your consumers where they are. Understand their process, and hand feed them the information they need to make their decisions, when they need it.

Via Domo Dashboards à Pushed right to your manager’s inboxà

ADOPTION & MAINTENANCE

36

DRIVING ADOPTION AND MAINTAINING

Once we had adoption and results, we monitored use through Domo. RXA to uses Domo to alert the data science team when the predicted schedules are falling

• ut of tolerance and the models

need to be adjusted outside the automated process.

37

RESULTS

38

STORE MANAGER FEEDBACK

39

“I never have extra bodies or am short staffed.” “I’m trying to follow it as much as I can.” “I think it’s spot on.” “We haven’t been running behind.”

BUSINESS

40

14% reduction in hourly labor costs Increased Same Store Sales 2.1% Maintained Net Promoter Score of 75 Increased Same Store EBITDA by 8.4%

THANK YOU :)

41

THANK YOU