The Home Depot: Driving Category-Specific Business Decisions with Alteryx

Proprietary & Confidential to The Home Depot

Driving Category-Specific Business Decisions with Alteryx

Charles ColemanThe Home Depot

Alteryx RoadshowSeptember 15, 2015

2 Proprietary & Confidential to The Home Depot

Agenda

About us

What is Assortment Planning?

Initial state – chaos and ugly analytics

The change – advanced data processing

The end result – automated workbooks, profit

3 Proprietary & Confidential to The Home Depot

Let’s Get This Out of the Way…

“Sooooo, do you get a discount?”

4 Proprietary & Confidential to The Home Depot

Hah, You’re Funny

NO!!!

We work on a ridiculous scale.

2,000 stores + corporate office = 367,000 employees in the US alone.

Put another way, we employ 0.12% of the United States.

5 Proprietary & Confidential to The Home Depot

Who We Are, How We Grow

So, we sell tools … The old way: 1979 – 2000’s – Expand!! The new way: optimize sales $/ft2

6 Proprietary & Confidential to The Home Depot

What is Assortment Planning?

Assortment: everything the store sells

Assortment Planning: telling every store what to sell

“Where should I sell this hammer (if anywhere)?”

7 Proprietary & Confidential to The Home Depot

Some Depot Stats

2000 stores 160,000 SKUs 100,000 SKUs in one store’s assortment 575 categories, each with multiple strategies Transactional data down to the SKU-register-second

level Multiple price points per SKU

… That’s a lot.

8 Proprietary & Confidential to The Home Depot

Before Alteryx

Clustering process before: Rudimentary, regional Largely Pro/DIY Little execution, little change

Size changes our granularity? Small categories = sophisticated Tableau workbooks

Large categories = slow, clunky, unusable workbooks

9 Proprietary & Confidential to The Home Depot

Clustering Before – Regional

10 Proprietary & Confidential to The Home Depot

Clustering Before – Pro/DIY/Both

11 Proprietary & Confidential to The Home Depot

The Initial Problem

How to get 20 users able to clean, join, and use their data in Tableau while using business machines?

12 Proprietary & Confidential to The Home Depot

At First, It Did Not Go Well…

Too much data

Changing categories

Awful computers and IT wars

Everyone wanted individual workbooks

13 Proprietary & Confidential to The Home Depot

My Alteryx Sales Pitch

“So this one process now takes between hours and a few days.”

14 Proprietary & Confidential to The Home Depot

We Needed a Quick Win…

Bad data outputs from our systems = 10x records Solution = remove bad data Before: several hours After: ~3 minutes

15 Proprietary & Confidential to The Home Depot

Developing a Clustering Data Process

Data sources: Transactional data – TeraData and proprietary tool Product attributes – TeraData Store information

Two different databases One proprietary tool

Clustering data – output from proprietary tool Need to combine all and output a .tde file quickly

16 Proprietary & Confidential to The Home Depot

Success?

When we ran data through Alteryx, it: Saved hours to days of processing time

Kept business users out of databases

Required support 16 hrs/day to run ETL

17 Proprietary & Confidential to The Home Depot

That’s Nice, But…

What would make this awesome?

Automation.

18 Proprietary & Confidential to The Home Depot

How to Automate This?

Only one Designer license – time to learn to code!

Materials: Old, crappy desktop machine Awesome (umm… borrowed) laptop running under a desk Shared drive Management buy-in Team members’ patience

19 Proprietary & Confidential to The Home Depot

User’s Process

Put category sales, product files through clustering

Save clustering files onto our shared drive

[Have a Coke and a smile]

Receive complete Tableau Desktop file

20 Proprietary & Confidential to The Home Depot

The End Result

A machine under my desk - running 24/7 – that uses scripts to: Pulls users’ files down from a shared drive Uses an Alteryx workflow to clean data, assign a decision tree

to it, and produce a .tde output Change Tableau data source to new extract Use decision tree to modify category charts and graphs Create “fake” PowerPoint slides for presentations Send each user his/her workbook along with an email

All of this in 9 minutes!

21 Proprietary & Confidential to The Home Depot

What We Actually Built

In total: mass-customized, highly analytical Tableau workbooks, nearly on-demand

22 Proprietary & Confidential to The Home Depot

A Quick, Early View: Alteryx

Data inputs

Clean and join, clean and join Tableau

extractoutputApply decision tree(Now mostly DB

connections)

23 Proprietary & Confidential to The Home Depot

A Quick View: Tableau – Change Clusters

24 Proprietary & Confidential to The Home Depot

A Quick View: Tableau – Identify, Analyze Outliers

25 Proprietary & Confidential to The Home Depot

A Quick View: Tableau - Presentation

“We should sell this hammer here.”

26 Proprietary & Confidential to The Home Depot

Results: Metrics and Gains

Time savings per iteration: 98.5% (~80 hours to 1 hour)

[A dramatic increase in] revenue per category

More, better outlier analysis: Sales clustering outliers – how do assortment effects boost

or kill sales? Cannibalization Assortment, line structure, competitor gaps

Demographic outliers – is there more potential

Changed the way we hire and train!

27 Proprietary & Confidential to The Home Depot

That was then… What since?

Two extra weeks per user per review gave us time for: More categories; broader, faster impact

More intense localization – by customer

More granularity on substitutable items

Exception reporting by SKU, store, season, market, category, region

Better integration of online data to predict store sales

28 Proprietary & Confidential to The Home Depot

Questions?