Use RFM to Boost Your Response Rate DMA
Use RFM to Boost Your Response Rate DMA Monday, October 17, 2005 1: 00 – 2: 00 PM Georgia World Congress Center Atlanta, Georgia Arthur Middleton Hughes Vice President / Solutions Architect Knowledge. Base Marketing, Inc.
How a modern database system works Customer Transactions Marketing Database Inputs from Retail, Phone, Web Data Access & Analysis Software Appended Data & Modeling Web Site Marketing Staff Customer Service
Two Kinds of Database People n Constructors People who build databases Merge/Purge, Hardware, Software n Creators People who understand strategy Build loyalty and repeat sales n You need both kinds!
Responsiveness & Profitability are not the same
Recency Frequency Monetary (RFM) Analysis • Used for marketing to customers • Always improves response and profits • Better than any demographic model • The most powerful segmentation method
How to Apply Recency Codes • Put most recent purchase date into every customer record • Sort database by that date - newest to oldest • Divide into five equal parts - Quintiles • Assign “ 5” to top group, “ 4” to next, etc. • Put quintile number in each customer record
Responsive customers may not be the most profitable Profitable Customers Responsive Customers RFM LTV Not all responsive customers are profitable Not all profitable customers will respond when you write them.
RFM Can Predict Responders • For product launch, select SICs with highest penetration ratios • Use RFM to select most likely responders • Use combination of mail, phone, and sales visits to responsive relationship buyers.
How to Apply Recency Codes • Put most recent purchase date into every customer record • Sort database by that date - newest to oldest • Divide into five equal parts - Quintiles • Assign “ 5” to top group, “ 4” to next, etc. • Put quintile number in each customer record
Response by Recency Quintile
How to compute a Frequency Index • Keep number of transactions in customer record • Sort Recency Groups from highest to lowest • Divide into five equal groups • Number groups from 5 to 1 • Put Quintile number in each customer record
Response by Frequency Quintile
How to compute a Monetary Index • Store total dollars purchased in each customer record • Sort Frequency Groups from highest to lowest • Divide into 5 equal groups (Quintiles) • Number Quintiles 5, 4, 3, 2, 1 • Put Quintile number in each record
Response by Monetary Quintile
Monetary Response to $5, 000 Product Percentage of households promoted who purchased 2 1. 68 1. 5 1. 17 0. 88 1 0. 66 0. 5 0. 32 0 5 4 3 Monetary Quintile 2 1
RFM Code Construction R 5 F 35 4 34 3 33 2 32 31 1 Database One Sort Five Sorts M 335 334 333 332 331 Twentyfive sorts
Appended RFM Codes
Creating an Nth 300, 000 Records Customer Database For Nth by 10, select every tenth record. Nth 30, 000 Records Result will be statistical replica of database
Result of Test Mailing to 30, 000
Test Response Rate by RFM Cell
Profit from Test Mailing
Determine Break Even and Test Sizes
How to Compute the Response Rate • Divide number of responses by number mailed. Multiply by 100 • Example: Responses = 1034 Mailed = 40, 000 Rate = 1034 / 40, 000 Rate = 2. 59%
Test, Full File & RFM Selects Compared
Test Vs Rollout Response Rates
Retroactive RFM Test • Many times there is not enough time or funding to run an Nth test in advance • Solution: apply RFM codes to your last completed outgoing promotion. • Since you know who responded, you can determine response rates by cell • Use previous rates to govern this rollout.
How Many RFM Cells Needed? • Test File = (Test Budget) / (per piece cost) • Example = $15, 000 / $0. 76 = 19, 737 • Cells Needed = 19, 737 / 274 = 72
Cell Division Determination • To create 72 cells, some must be less than 5 • Recency most powerful. Do not scrimp. • Example R-F-M = 6 X 4 X 3 = 72 • Is this best? Test and see.
RFM For Business Databases • Business databases are small • For small databases, use quartiles or thirds • Quartile = 4 X 4 = 64 Cells • Thirds = 3 X 3 = 27 Cells • Custom = 5 X 2 = 20 Cells
Recent Case History • User sells personalized product by mail • 45, 000 selected for a test
Second Recency Quintile Had More Responses. Why?
Even so, First Recency Quintile Had Higher Sales
Recent buyers spend more per order
Lowest two recency quintiles did not break even
Frequency was very predictive of response
Monetary did not predict response rate very well
But Monetary does predict average sales by quintile
RFM Cells clearly show who to mail to, and who to drop
When NOT to use RFM • If you use it all the time, half your customers will never hear from you • They will be lost • The others will suffer from File Fatigue • Use it sparingly • Product launch is ideal use
Response by Recency Quintile
How to compute a Frequency Index • Keep number of purchases in customer record • Sort records in each recency quintile from highest to lowest • Divide into five equal groups (Quintiles) • Number quintiles from 5 to 1 • Put Quintile number in each customer record
Response by Frequency Quintile
How to compute a Monetary Index • Store total dollars purchased in each customer record • Sort the records in each frequency quintile from highest to lowest • Divide into 5 equal groups (Quintiles) • Number Quintiles 5, 4, 3, 2, 1 • Put Quintile number in each customer record
Response by Monetary Quintile
RFM Code Construction R 5 F 35 4 34 3 33 2 32 31 1 Database One Sort Five Sorts M 335 334 333 332 331 Twentyfive sorts
Appended RFM Codes
Result of Test Mailing to 30, 000
Test Response Rate by RFM Cell
Profit from Test Mailing
What is the break even rate? • Each test segment must be measured • A segment breaks even if the profit from sales exactly equals the cost of the promotion • BE = (Per Piece Cost) / (Net revenue from one sale) • BE = ($0. 48) / ($28) = 1. 71%
How large must test segments be? • Large enough for predictive accuracy • Small enough to keep test costs down • Size = 4. 00 / (Break Even Rate) • Size = 4. 00 / 1. 71% = 234 pieces mailed • You should adjust the “ 4. 00” based on your experience -- up or down.
How to Compute the Response Rate • Divide number of responses by number mailed. Multiply by 100 • Example: Responses = 1034 Mailed = 40, 000 Rate = 1034 / 40, 000 Rate = 2. 59%
Test Response Rate by RFM Cell
Test, Full File & RFM Selects Compared
Test Vs Rollout Response Rates
RFM Deals With Very Small Numbers • Only a small percentage (such as 5%) of customers respond to the typical offer • 95% or more will not respond at all • RFM tells you which customers are most likely to be in the responsive 5% • Those who respond may not be your most profitable customers
Retroactive RFM Test • Many times there is not enough time or funding to run an nth test in advance. • Solution: apply RFM codes to last year’s completed outgoing promotion. • Since you know who responded, you can determine response rates by cell. • Use last year’s rates to govern this year’s rollout.
Recent Case History • User sells personalized product by mail • 45, 000 selected for a test
Second Recency Quintile Had More Responses. Why?
Even so, First Recency Quintile Had Higher Sales
Recent Buyers Spend More per Order
Lowest Two Recency Quintiles did not Break Even
Frequency was Very Predictive of Response
Monetary did not Predict Response Rate Very Well
But Monetary does Predict Average Sales by Quintile
RFM Cells Clearly Show who to Mail to, and who to Drop
When NOT to use RFM • If you use it all the time, half your customers will never hear from you • They will be lost • The others will suffer from File Fatigue • Use it sparingly; when you need a boost • Use it to identify your best customers • Don’t go hog wild!
Half Life Data
Graphing Half Life
Half Life by Revenue
What should you do? • Maintain a customer database • Maintain the most recent date, frequency of orders and total dollar amount • Put RFM cell codes into your records • With each mailing, see which cells respond. • Increase response and profits by NOT MAILING non responsive cells
Books by Arthur Hughes From Mc. Graw Hill. Order at www. dbmarketing. com Contact Arthur: arthur. hughes@kbm 1. com
- Slides: 72
