Non-Sufficient Funds and FraudAnalysis & Recommendations
May 12, 2009
Core Growth Strategy
Overview of redbox Billing & Controls
• New customers can only rent 3 DVDs on their first rental.
• Active Customers may not rent
more than 5 DVDs at a time. • Redbox “deactivates” a
credit/debit card (blocking the ability to rent) after 90 days of trying to collect for an amount of $30 or more.
• Redbox maintains a blacklist of
customer names that are blocked from renting on any credit card. Names are added to the blacklist after 5 cards with the same name are deactivated. Names that are too common (i.e. Maria Garcia) cannot go on the blacklist. The blacklist also includes common gift cards.
Controls
Extra Nights Declines (END)
• Customer returns the DVD after 2 or more nights.
• When redbox charges the customers’ card for the extra nights, the charge is declined.
Passive Buy Declines (PBD)
• Customer keeps the DVD for 25 nights.
• When redbox charges the customers card for the “passive buy”, the charge is declined.
Uncollected Charges
• The current billing process has been in place since March of 2009
• When a customer swipes their credit card, redbox immediately verifies and charges $1/DVD for the first night rental.
• After the DVD is returned, redbox charges $1 per night for each extra night (called Extra Night Charges).
• If the DVD is not returned after 25 nights, redbox charges $24 and the customer owns the DVD (called a Passive Buy).
Current Billing
3
Potential Financial Impact in Year 1
If redbox is able to implement all of the changes identified to date, the potential annual EBITDA impact is $2.6-$2.8 million
$600K - $800K
$2 million
$2.6-$2.8 million
Opportunity Est. Yr. 1 EBITDA Impact
New Controls to reduce Passive Buy Declines
New Controls to prevent Extra Night Declines
Total
4
Definition of NSF/Fraud
For redbox, Non-sufficient Funds (NSF)/Fraud consists of uncollected
charges due to credit card declines for:
Extra Nights($1/night)
disks were returned
1
Passive Buys($24/disk)
disks were kept by customer
2
2008 Uncollected Charges
Gross Charges ( $M)
% of Declined
Transactions
5
How Big is the Opportunity to reduce Uncollected Charges?
The total uncollected charges from declines on extra nights and passive buys is expected to be ~$61.5M in 2009
• 2009 is estimated as 8 % of planned revenue, this assumes that the 2009 rate of uncollected charges is the same as the 2nd half 2008
• Extra Night Declines (END) include declined charges on extra nights for DVDs that were returned to redbox
• Passive Buy Declines (PBD) include declined charges on unreturned DVDs, and therefore include some component of intentional theft
Extra Night Declines
Passive Buy Declines
$32.8
$61.5
Estimated Uncollected Charges ($M)
Assumptions
6
2008 Trends
Losses from uncollected charges have declined as a percent of total revenue throughout 2008
• The collection period was extended to 90 days in April ‘08
• Stricter Blacklist policies were implemented in Q2
• Weekly Blacklist updates were implemented in Q3, resulting in a sharper decrease in PBDs in Q4
• It is too early for a comparative measure for Q1 2009, since more than 60% of the collection period is remaining at the time of analysis
ObservationsEstimated Uncollected Charges Due to END/PBD
Extra Night Declines
Passive Buy Declines
Declined Transactions as % of Total Revenue
9.7% 9.3%8.7%
7.4%
7
Passive Buy Declines: June ‘08 Snapshot# of Customers (1,000’s)
31,000 customers had a PBD result from a June ’08
rental. These 31K are associated
with 70K credit cards.
6,000 had a PBD on one of their
other credit cards prior to 6/08, so
are excluded from the “new”
offender analysis.
5,000 were “1 and done”. The June PBD was their first and
only transaction with redbox on any credit card.
9,000 had paid rentals prior to
the June PBD, but never rented again on any
credit card after the PBD in June.
11,000 continued to rent from
redbox on one or more credit cards
after their first PBD in June.
A B C D E
Customers with Passive Buy Declines on June 2008 Rentals
$120K declined charges in 6/08
$216K declined charges in 6/08
$264K declined charges in 6/08
8
Passive Buy Declines: June ‘08 Snapshot
Total net future revenue = $390
Revenue ($1,000s)
Post-June 2008 Revenue Stream
11,000 continued to rent from
redbox on one or more credit cards
after their first PBD in June.
E
EBITDA Impact of Customer Termination
Although this group of customers had net positive revenue following the first PBD in June ‘08, termination of these customers and all of their associated credit cards immediately after the first PBD would have had a net positive impact on EBITDA
EBITDA Loss on Future Paid Rentals:
EBITDA Gain on Preventing Unpaid Rentals:
RevenueEBITDA MarginEBITDA Impact
RevenueCOGS + OP ExpEBITDA Impact
$(635,000)18%
$(114,300)
$245,00072%
$176,400
Net EBITDA Impact $62,100
Note: In practice, the potential EBITDA gain is less than in this illustration because termination would occur after 90 days of collection on first PBD, so future gains would exclude rentals during this 90 day period
$264K declined charges in June ‘08
9
PBD Recommendations
The total potential EBITDA impact of the recommended PBD solution is estimated to be ~$600,000-$800,000 per year
1. Change blacklist processing from blocking names, to blocking name/billing zip code combinations.
2. Convert blacklist from a manual weekly procedure to an automated, nightly procedure; incorporating all of the manual changes suggested in “Short Term Solutions”.
3. Build new charge validation logic to try to collect the $24 due on the previous PBD immediately when a customer with a PBD visits a kiosk. Allow rental if customer covers PBD + current rentals, otherwise decline at kiosk.
4. Maintain blacklist on kiosk CPU for local processing (preventing off-line fraud)
5. Continue analysis of the potential value of extending the collections period beyond 90 days
Short Term Solution Longer Term Solution
1. Add names to the blacklist as soon as 90 day collection period is ended on 1st PBD ($24 in charges) on only 1 credit card.
2. A high number of customers should be added to the black list under the current policy, but are not because their name is shared by other good customers. For these customers, we need to deactivate all of the credit cards that share their name and billing zip code.
3. Work with field ops to change connectivity hardware at high-risk kiosks to prevent customers from committing off-line fraud.
Extra Night Declines: June ‘08 Snapshot
100,00050,00025,000 75,000# of June END Rental Visits
37% of the total END charges from June 2008 rentals were from rental visits with a basket size of 3 or more disks
Avg. Declined Charges per Visit ($)
Opportunity:
Develop a risk profile to identify and limit the basket size for
high risk customers
Extra Night Declines from June 2008 Rentals
1 disk taken
3 disks taken
2 disks taken
4 disks taken
$51,000
5 disks taken
$29,000
$214,000
$265,000
$207,000
June ’08 END
Total # END Visits
# Credit Cards w/ END
Total END Charges
89,000
73,000
$767,000
$25
$20
$15
$10
$5
11
END Recommendations
The total potential EBITDA impact of the recommended END solution is estimated to be over $2M per year
1. Implement a credit risk matrix that assigns a risk score to each redbox account (credit card). Update rental systems software to restrict basket size per credit card based on the risk score.
Note: Analysis to date has identified one specific risk matrix / basket size limit scenario that would have had a $2M EBITDA impact if implemented in 2008. Further optimization is possible with continued analysis.
2. Continue analysis to determine if there is a benefit of implementing the risk matrix in (1) at a customer, rather than credit card level.
Short Term Solution Longer Term Solution
1. Continue scenario analysis to determine whether immediately implementing more stringent basket size limits at specific high risk kiosks, or for all customers could have a positive EBITDA impact.
Appendix
13
EBITDA Impact of Preventing Uncollectibles
Example Business Model
Product Cost $.60Labor Cost $.15Overhead Cost $.05
Sell Price $1.00
EBITDA Margin 20%
1 unit is stolen
1 unit is purchased
EBITDA Impact of Fraud vs. Sales
+$.20
-$.80
Customer A
Customer B
Customer C
Units Stolen1
2
1
Fraud EBITDA-$.80
-$1.60
-$.80
Units Purchased5
2
4
EBITDA$1.00
$.40
$.80
Customer EBITDA$.20
-$1.20
$.0
-$1.00Even though the net revenue from these 3 customers is +7, the net EBITDA impact is -1, and the company would want to implement
policies to prevent these transactions from occurring
Passive Buy Declines
14
From arevenue perspective, the paid charges associated with customers that owe redbox for at least one passive buy appears to exceed the unpaid charges they continue to generate
2008 Paid vs. Unpaid ChargesCustomers with at least one PBD in 2008
Note: Unpaid charges excludes the first PBD
Customer Rollup Methodology - 1
one customer
ZIP Code
s
Names
Emails
Names
Emails
ZIP Codes
Cardholder name formats differ across accounts
Emails are manually entered and could be incorrect or not truly
exist
ZIPs are manually entered but most customers believe they
need to be correct for the transaction to be successful
Redbox customers often have more than one credit or debit card they use for transactions. To date, there is no standard procedure on matching multiple accounts to a unique customer. This is a proposed methodology for determining how many accounts each redbox customer actually represents.
There are three main variables that play a part in the account matching process: names, emails, and ZIP codes.
“Dirty” data must be cleansed before processing can begin; matching exact account names is often difficult because different credit and debit cards use various card holder naming conventions; for example, a MasterCard transaction looks like “Jane Smith”, whereas a Visa transaction may look like “Smith Jane”.
Process OutlineI. Scrub the Data
1. Filter out invalid ZIP Codes (eg “00000”)
2. Alter email addresses (eg change [email protected] to [email protected])
3. Remove very common names, null names, and non-names (eg “a gift for you”)
II. Refine the Data
4. Separate full account names into meaningful pieces
5. Remove extraneous characters (eg extra spaces or slashes)
6. Match names in “forward” order
7. Match names in “backward” order
III. Group the Data
Please see the next slide for a detailed outline of the proposed matching process
Customer Rollup Methodology - 2
Below is the proposed order in which matches could be made; if wanted, varying levels of matches may be awarded a higher or lower “score” based on completeness of the match
10
9
7
6
4
3
Possible Point Allocation
1Full name or reverse refers to a match like “Jane Smith” or “Smith Jane”2Due to the many variations on name formats, the first initial may actually be the first letter of the middle or last name; thus the last name may actually be a first or middle name
Customer Rollup Methodology - 2