Business Intelligence/Decision Models
RFM Analysis
Outline
Explain how RFM works, and how you can use it to improve response rates and reduce costs
SPSS RFM With transaction files With customer files
Use of RFM with Target Variable2
Recency, Frequency, and Monetary (RFM) Analysis
According to Hughes! Always improves response and profits
Better than any demographic model
The most powerful segmentation method for predicting response
Old-fashioned RFM? Never assume a CHAID program or
even a regression model will outperform an old-fashioned RFM…
(David Shepard)
Response by Recency Quintile
3.49%
1.25% 1.08%0.63%
0.26%
0.00%0.50%1.00%1.50%2.00%2.50%3.00%3.50%4.00%
5 4 3 2 1
Resp
onse
Rate
Recency Quintile
Response by Frequency Quintile
1.99%
1.56%
1.31%
0.92% 0.93%
0.00%
0.50%
1.00%
1.50%
2.00%
2.50%
5 4 3 2 1
Re
spo
nse
Ra
te
Frequency Quintile
Response by Monetary Quintile
1.61%
1.45% 1.46%
1.22% 1.23%
0.00%
0.20%
0.40%
0.60%
0.80%
1.00%
1.20%
1.40%
1.60%
1.80%
5 4 3 2 1
Big Ticket: Response to $5,000 CD Offer by Monetary
Monetary Quintile
Percentage of households promoted who purchased
1.68
1.17
0.88
0.66
0.32
5 4 3 2 10
0.5
1
1.5
2
RFM Code Construction 1
FM
One SortFive Sorts
Twenty-five sorts
Database
5
4
3
2
1
35
34
33
32
31
335334333332331
R
RFM Code Construction 2
F M
1st SortDB
5
4
3
2
1
R5
4
3
2
1
5
4
3
2
1
DB DB2nd Sort 3rd Sort
RFM Code Construction 3 Predefined subjective breaks:
R = < 2 wks, 2-6 wks, 7-8 wks, and > 12 wks F = 10, 8, 6, 3, 1 M = > $500, $300-500, $200-300, $100-200, < $200,
11
RFM Code Construction 4 R = 5, 4, 3, 2, 1 @ weight = 10 F = 5, 4, 3, 2, 1 @ weight = 5 M =5, 4, 3, 2, 1 @ weight = 1 Weighted sum
Score = xiwi
Ex. RFM 421 = Score 51
Let’s stick with this one!
FM
One SortFive Sorts
Twenty-five sorts
Database
5
4
3
2
1
35
34
33
32
31
335334333332331
R
Break Even Response Rate BE = Unit Cost / Unit Profit
Example:• Unit cost = .55• Unit profit = $33• Thus $.55 / $33 = 0.0167 or 1.67%
Number of Cells 5 x 5 x 5 = 125 cells How many individuals per cell so that
response is significant? Minimum 4 / BE Ex. 4 / .0167 = 240 individuals per cell # of customers +/- 30,000
Use Common sense:5x4x4, or 5x4x3, or 5x2x2, or 4x2x3
15
Customer Database
Nth
Systematic Sampling:Selecting every Nth case
300,000 Records
30,000 Records
For Nth by 10, select every tenth record.
Result will be a statistical replica of the database
Result of Test for 30,000# RFM Reach Response Rate 1 555 240 20 8.15% 2 554 240 16 6.56% 3 553 240 13 5.62% 4 552 240 10 4.33% 5 551 240 11 4.51%
6 545 240 9 3.78% 7 544 240 12 4.98% 8 543 240 6 2.88% 9 542 240 10 4.26%
10 541 240 7 3.10%
11 535 240 10 4.13% 12 534 240 9 3.83% 13 533 240 8 3.35% 14 532 240 6
2.70%
Case Study # of customers 2,100,000 Test with 30,000 Left with 2,070,000 If 555 RFM 125 cells 16,560 customers per cell
18
Test Response Rate by RFM Cell
-200
-100
0
100
200
300
400
500
555 455 355 255 111
Index of Response 0 = Break Even
34/125 cells above BE
Break Even Index (BEI) BE: .0167 set to 0 or 100 as a reference point Ex. Assume Response = .025 for a given cell
If BEI set to 0 BEI = [(Resp – BE)/BE] * 100 Ex. [(.025 - .0167)/.0167]*100 Ex. 49.7 rounded 50
If BEI set to 100 BEI = [Resp/BE] * 100 Ex. [.025/.0167] * 100 = 150
Profit from Test Program
Quantity Rate Amount
Gross Profit 402 $40.00 $16,080
Fixed Costs 30,000 $0.55 $16,500
Profits (Loss) ($420)
Test, Full File & RFM Selects Compared
Test Full File RFM Select
Response Rate 1.34% 1.17% 2.76%
Responses 402 23,412 15,540
Net Revenue $16,080 $936,480 $621,596
Quantities 30,000 2,001,585 563,040
Fixed Costs $16,500 $1,100,581 $309,672
Profits -$420 -$164,101 $311,924
RFM Procedure Transaction file
• Run RFM on transaction file• Create an new RFM dataset• Merge with customer file
Customer file• File is already prepared• Run RFM on time since last purchase,
number of purchases, and money spent
24
SPSS Means Procedure Compare Means Means
DV Response Variable (0/1) IV RFM (Means and N)
Copy output table to Excel Sort all columns by mean response Descending order
Determine BE and economics
25
Selected RFM
26
RFM score Mean N Cost Income515 .46 547 $ 410 $ 1,008 512 .42 1097 $ 823 $ 1,860 514 .42 530 $ 398 $ 892 513 .41 185 $ 139 $ 304 511 .38 318 $ 239 $ 488 411 .34 302 $ 227 $ 416 415 .34 565 $ 424 $ 760 414 .32 561 $ 421 $ 708 412 .31 1205 $ 904 $ 1,516 413 .30 185 $ 139 $ 224 312 .28 1543 $ 1,157 $ 1,704 311 .27 527 $ 395 $ 576 314 .27 724 $ 543 $ 768 315 .25 749 $ 562 $ 744 313 .24 132 $ 99 $ 128 211 .21 1053 $ 790 $ 904 213 .18 1318 $ 7,667 $ 13,000 $ 5,333
Total .07 82882 $ 62,162 $ 21,920 -$ 40,242
U. Cost U. Profit BE
0.75 4 0.1875
Bits and Pieces Recency with continuity:
Change in buying decisions Frequency?
Purchases in year or # of items Transactions in a year # of visits, # of calls
Tests and Rollouts
Caution about RFM Customer reach
File fatigue
Use RFM for testing
Use it to identify your best customers
Objectives? Profit, ROI, Reach
RFM Customer File
RFM Customer File
RFM Customer File
RFM Customer File
RFM Customer File
RFM Customer File
RFM Sort Analyze
Compare MeansMeans
RFM Sort, Import into Excel
36
RFM Mean N111 .1975 319112 .1937 320113 .1531 320114 .1438 320115 .1531 320121 .0562 320122 .0281 320123 .0438 320124 .0438 320125 .0344 320131 .0094 320
………..…
535 .1219 320541 .0719 320542 .0688 320543 .0594 320544 .0656 320545 .0500 320551 .0219 320552 .0188 320553 .0281 320554 .0250 320555 .0281 320Total .1484 39999
RFM EconomicsEx. Catalogue
Fixed Cost $4.00
Gross Profit $20.00
BE $4 / $20 = .20
Sent to all 39,999 Cost $159,996
Resp .1484 5,936 Gross $118,720 - $41,276
Sent to > .20 10,240 Cost $40,960
Buyers 4,367 Gross $87,340 $46,380
37
RFM EconomicsEmail
Fixed Cost $0.02 Unsub. .025
Gross Profit $20.00 Worst Case
.050
BE .02/20 = .001 CLV $100
Sent to all 39,999 Cost $800
Resp .1484 5,936 Gross $118,720
Cust. Loss 20,000 $2,000,000 - $1,882,100
Sent to > .20 10,240 Cost $205
Buyers 4,367 Gross $87,340
Cust. Loss 256 $25,600 $61,53538