Chi Tylana and Frazier McDonald analyzed retail company Taylor & Swift's customer data using Teradata sandbox systems. They segmented customers by their browsing and buying behaviors across channels. They discovered that customers who shopped across multiple channels had higher purchase values. The investigators identified opportunities to increase store visits and sales by recovering abandoned online purchases and offering in-store pickup. They proposed and tested these ideas, finding significant revenue increases from driving more omni-channel shopping.
1. HOW WE DID THE The Case of the
INVESTIGATIONS Retail Turnaround
2. Prelude – Case of the Retail Turnaround
This video appears on www.youtube.com. You can find it by
searching using keywords:
“BSI Teradata Case Retail Turnaround.”
This accompanying deck is designed to answer questions about
the Teradata and partner technologies shown in the story. For
best effect, run it in Powerpoint animation mode.
This is our second BSI episode for the Retail Industry. For
another episode, see “BSI Teradata Case of the Retail Tweeters.”
There are also many other episodes available that showcase the
use of business analytics to solve real-world problems.
2
3. Note from the Investigators
Hi everyone,
We’re the brains behind the scenes and wanted to answer your
questions about “how did we do our analytics to help Taylor &
Swift with Omni-Channel Retailing?”
This write-up will give you an idea of our clients’ architecture
and some details of the BI screens.
Take a look, and if you still have questions, send them to us at
the www.bsi-teradata.com FB page!
Yours truly,
Chi Tylana and
Frazier McDonald
BSI Investigators
3
4. Scene Synopsis
Scene 1: Taylor & Swift has a
problem
Scene 2: BSI Investigators Chi
Tylana and Frazier McDonald
analyze the data
Scene 3: Chi and Frazier show
Five Omni-Channel Retailing
ideas to Taylor & Swift execs
Scene 4: Omni-Channel
Retailing experimental results
4
6. Taylor & Swift has a problem with overall revenues
dropping. Their COO Mark Woolfolk calls a meeting with
his VP of Digital Stores, Becky Swenson, plus two
investigators from BSI Teradata he’s brought in to
provide a fresh set of ideas for turning things around.
6
7. Senior Leadership at
Mark Woolfolk is the Chief Operating Officer
for T&S. He knows operating results have
been just so-so, stagnant, and has a hunch
they need some new ideas to get people into
the stores, which will improve financial
results.
Becky Swenson is the VP for the Digital
Store. Her results are definitely better than
the physical stores but she cannot
compete against the pure Webs on price
alone (T&S cost structure issue), and
wants to help with possible synergies, but
isn’t sure what to do.
7
9. Key Performance Indicator: Same Store Results
Revenue by month per store is dropping
Seasonal Spring Back to Holiday
Promotions: Fling School Season
9
10. Becky sees better results for # of visits on
the digital channels (WebStore, Mobile)
However, this is misleading – number of purchases/visits is up
but size of purchase/market basket has dropped.
10
11. Overall Results – 210 Stores Plus Digital
Average store revenue continues to decline, while digital
channel (Web and mobile) sales are flat. Problem!
11
12. The Job for BSI
1. Analyze customer segments based on behavior for visiting
and buying
2. Because multi-channel visitors purchase more, figure out
how to use insights from the digital channels to drive more
people into stores
3. Use Taylor & Swift’s investments in “active” near-real-time
technology and sandboxing for fast data discovery
4. Come back with some recommendations for turning the
financial results around
12
14. Chi and Frazier load T&S data into Teradata
and Aster sandbox systems
• Sandbox systems are great for discovery of trends
> They load 18 months of purchase data from all channels, plus
web click data into their sandbox systems
> They use Tableau to do quick visualizations
• They begin by segmenting customers by browse vs. buy
channels
> Some people stick to one channel (e.g., browse on the Web, and
buy on the Web)
> Others switch channels (e.g., browse on mobile or Web, then go
to the store to buy)
> A simple Venn diagram can show the relative numbers of people
• The focus of the work will be on those who are on the digital
channels – can we get them into the stores, too?
For more technical information about Sandbox technologies
and Agile Analytics, click here
14
15. Venn Diagram - Browse/Buy Analytics
Behavior Across Channels
15
16. Geospatial Analytics
Geocoding customer street/city addresses provides customer
“dots” on the page. Chi then uses Teradata geospatial
capabilities to find only those customers within a 20-mile drive
of a physical store.
16
17. Customer Value Depends on (# of Visits)
Multiplied by (Average Market Basket $ Size)
This is the “proof” that multi-channel visitors are more
valuable.
17
18. Brainstorming: How To Get More People to
Become Multi-channel Shoppers?
Idea #1:
CLICK AND
COLLECT
If customer is
near a store and
all items in the
market basket
are in stock,
offer local store
pickup option
18
19. Frazier is an expert at doing Data Discovery
using Aster Analytics
• Aster Data (now Teradata Aster) was acquired by Teradata in
2011 and is used by numerous customers to analyze “non-
traditional” data that doesn’t fit nicely into traditional
relational tables and rows
• Graph pattern matching is an example that we show in this
episode
> Specifically, the page-by-page views that a customer looks at
and which items are put in a market basket is of high interest
• Teradata Aster and Tableau can help you visualize all
patterns
• For more information about click here
• For more information about , click here
19
21. Teradata Aster – nPath Analysis
• The highlighted path shows one shopper who put Labels in
the shopping cart, then Envelopes, then an Office Machine,
and finally an Electronics item.
• These digital pathways provide more information than
traditional POS (point-of-sale) information from the store
system: not just WHAT you bought, but IN WHAT ORDER.
• Aster can also be used to monitor non-purchase behavior.
• Specifically: Bail-out Analytics can be quite useful to see
where people “X-out” of sessions before purchasing. This can
be helpful in redesigning Web sites to decrease confusion
and increase conversions, and in making decisions about
whether shipping charges are a problem area, etc.
• Frazier takes several looks at pathways – an area that Aster
calls “nPath” because there can be 1, 2, … n steps on the
way to purchase.
21
23. How to Reduce Dropouts on the Shipping
Page
28% of customers who initiate the purchase sequence after
shopping are dropping out at the Shipping Charges page
Idea #2: Coupons for In-Store
Pickup vs. Shipping Charges
If the products are all
in stock, then offering
a modest amount of
money ($5) to
customers to drive to
pick up the items
might drive them into
the stores
23
24. Teradata Aster Analytics
Endpoint: Bail Outs When Out of Stock – Split Shipments
T&S loses more customers if they make it past the Shipping
Charge page, but then find that the order will be split because
24 some items are not in stock.
25. Dropped Demand Recovery
• Frazier finds that another 48% of the customers bail out
when they find that something in the market basket isn’t going
going to be shipped because Taylor & Swift is out of stock.
• Frazier could also also analyze whether they come back – after 1
day, after 3 days, after a week.
• If neither of these happen, then we have “Dropped Demand” and
can assume we lost the sale (to competition) or the customer is
going to wait longer
• If we act quickly, we might be able to recover the Dropped
Demand, which leads to
Idea #3: send an email when the local store is back in stock
• They could come to the store to buy, or buy on the digital store – in
either case, we get the sale
25
26. Teradata Aster Analytics
Discovery: If “First in Basket” ships first, Purchase is salvaged
Deeper discovery – who does NOT bail out despite a split
shipment? Answer: in many cases, if the First in Basket
26
makes it into the First Shipment.
27. Teradata Aster Analytics
First in Basket Items are Very Important
• Frazier’s final discovery in this story is that sometimes with a
split shipment, customers still go to the Purchase page
> A study of those customers illuminates a new discovery – that if
the item they put first in their basket makes it into the first
shipment, then they proceed
• As a consequence, it’s important for Taylor & Swift to pay
close attention to all First in Basket items since those are the
“drivers” for purchases
• Chi suggests
Idea #4: they use First in Basket visuals in store circulars
• And Frazier comes up with
Idea #5: adjust “safety stock” levels
(at the digital store as well as physical stores) to ensure that
it’s likely that these leading products are always in stock
27
28. SCENE 3:
CHI AND FRAZIER SHOW THEIR
FIVE OMNI-CHANNEL RETAILING
IDEAS TO TAYLOR & SWIFT EXECS
29. Mark likes the “Recover Dropped Demand”
Send Emails when back in stock at stores
29
30. The Emails can be personalized and also feature
other browsed-but-not-bought products
Clock countdown
feature may help
30
31. The Email Campaign can be run automatically using
Aprimo Relationship Manager with Real-Time Messaging
• Taylor & Swift bought Teradata’s Aprimo Relationship
Manager tool two years ago to help design and execute
marketing campaigns
For more information about click here
• It’s not difficult to add “events” with workflows to describe
what to do when Taylor & Swift notices various activities by
customers
• In the case of Dropped Demand, Chi and Becky set up a
workflow to automatically detect when out-of-stock order
bailouts occur by Web-only customers who live near stores.
When the item is back in stock, an email goes out
automatically using the Real-Time Messaging module.
31
32. Workflow for Driving the Automatic E-Mails
Click to see the sequence of events that Aprimo will automatically
monitor – driving emails for Dropped Demand items.
32
33. Mark also wants to try Click and Collect
(featuring the First in Basket item)
33
35. Eight Weeks Later, Experimental Results Are In
Experiment 1: Click and Collect
Experiment 2: Recover Dropped Demand
Experiments were tried at 20 of T&S’s 210 stores
35
37. Financial Impact – Click and Collect
• The Cascade visual shows:
> the number of Web sessions
> what number of offers were made for store pickups (when every
item is in stock)
> the number of pick up offers accepted (so items were held)
> the number of actual pickups
• This campaign drove 59,000 people into the stores that
otherwise probably would not have gone there
• They bought what they ordered
• But we also measured incremental (impulse) purchases,
which was $32.08
• An additional $1.9M revenue
• Scaling up from 20 stores in 8 weeks to 210 stores annually,
this could be $120M of added revenue
37
39. Financial Impact – Dropped Demand Recovery
• The Cascade visual shows:
> the number of dropped demand sessions
> the number of emails sent when back in stock for that item
> the number of pick up offers accepted (so items were held)
> the number of favorable responses to re-order
> which channel they used – digital or in-store
The results were split 50-50, with half the people re-ordering on
the digital channels and half going to stores
But a key finding was that those who went back to the digital
channels ordered an incremental $12 of merchandise beyond the
dropped demand items, whereas in-store purchases made $22 of
additional purchase. Total incremental revenue of $835K on top of
the $2.2M in dropped demand merchandise – total $3.0M
Scaling up nationwide, annually, this could yield $189M of revenue
to Taylor & Swift
39
40. Mark and Becky are happy with the BSI
analytics and experiments…
40
41. Omni-Channel Retailing is a Very Hot Topic
For more information
• PODCAST: “Trending in Retail Consumer Insight”
• Best in Class: Cabelas
> RIS News article “Why Cabelas Has Emerged as the Top Omni-
Channel Retailer”
> Baylor Business School, Prof. Jeff Tanner, “Decoding Path to
Purchase”
• Best in Class: DSW -
41
42. Thanks for viewing these slides
And thanks to our Teradata divisions and Partners for making it all possible!
42
Editor's Notes
48% of customers: Further analysis (not shown) can be done to see whether people then buy a substitute that is in stock as a substitute (e.g., they want a printer but the brand they prefer is not available in their price range, so they pick a different brand)