Regression analysis is a statistical tool for comparing variables. In the past, mailer's analysis
was most often based on penetration reports and indexing various demographics separately,
referred to as univariate analysis. A strong advantage of regression is that it takes into
consideration the fact that index scores can quickly change when combined with other
qualifiers/variables.
For example, an univariate study may show a high index score for "x" variable. However, a
Regression would observe that "x" is a good prospect "only" if it meets a certain criteria.
The regression model can predict variables that are most strongly related to your objective,
such as: higher response, higher profit, or greater purchase frequency.
The regression model will also stratify your file into several categories in relationship to sample
size requirements and/or profit/cost restrictions. The below GAINS CHART is an example. From this
chart, you may decide to only mail the top percentile in order to achieve your revenue goals.
GAINS CHART SUMMARY
Node
Node: n
Node: %
Resp: n
Resp: %
Gain (%)
Index (%)
Node: n
Node: %
Resp: n
Resp: %
Gain (%)
Index (%)
2
1,327
9.6
1,327
16.1
100.0
167.0
1,327
9.6
1,327
16.1
100.0
167.0
22
635
4.6
635
7.7
100.0
167.0
1,962
14.2
1,962
23.8
100.0
167.0
30
565
4.1
565
6.8
100.0
167.0
2,527
18.3
2,527
30.6
100.0
167.0
7
443
3.2
443
5.4
100.0
167.0
2,970
21.5
2,970
36.0
100.0
167.0
13
443
3.2
443
5.4
100.0
167.0
3,413
24.7
3,413
41.3
100.0
167.0
45
107
0.8
107
1.3
100.0
167.0
3,520
25.5
3,520
42.6
100.0
167.0
32
95
0.7
94
1.1
98.9
165.2
3,615
26.2
3,614
43.7
100.0
166.9
9
114
0.8
111
1.3
97.4
162.6
3,729
27.0
3,725
45.1
99.9
166.8
23
142
1.0
134
1.6
94.4
157.6
3,871
28.1
3,859
46.7
99.7
166.4
38
244
1.8
229
2.8
93.9
156.7
4,115
29.8
4,088
49.5
99.3
165.9
3
222
1.6
208
2.5
93.7
156.4
4,337
31.4
4,296
52.0
99.1
165.4
15
143
1.0
129
1.6
90.2
150.6
4,480
32.5
4,425
53.6
98.8
164.9
10
207
1.5
172
2.1
83.1
138.7
4,687
34.0
4,597
55.6
98.1
163.8
28
206
1.5
165
2.0
80.1
133.7
4,893
35.5
4,762
57.6
97.3
162.5
16
757
5.5
548
6.6
72.4
120.9
5,650
41.0
5,310
64.3
94.0
156.9
33
185
1.3
121
1.5
65.4
109.2
5,835
42.3
5,431
65.7
93.1
155.4
11
362
2.6
234
2.8
64.6
107.9
6,197
44.9
5,665
68.6
91.4
152.6
41
244
1.8
147
1.8
60.2
100.6
6,441
46.7
5,812
70.4
90.2
150.7
19
446
3.2
266
3.2
59.6
99.6
6,887
49.9
6,078
73.6
88.3
147.3
4
672
4.9
374
4.5
55.7
92.9
7,559
54.8
6,452
78.1
85.4
142.5
37
243
1.8
134
1.6
55.1
92.1
7,802
56.6
6,586
79.7
84.4
140.9
27
545
4.0
299
3.6
54.9
91.6
8,347
60.5
6,885
83.3
82.5
137.7
8
440
3.2
203
2.5
46.1
77.0
8,787
63.7
7,088
85.8
80.7
134.7
36
234
1.7
106
1.3
45.3
75.6
9,021
65.4
7,194
87.1
79.7
133.1
25
567
4.1
256
3.1
45.1
75.4
9,588
69.5
7,450
90.2
77.7
129.7
39
751
5.4
282
3.4
37.5
62.7
10,339
75.0
7,732
93.6
74.8
124.9
47
114
0.8
41
0.5
36.0
60.0
10,453
75.8
7,773
94.1
74.4
124.2
40
444
3.2
121
1.5
27.3
45.5
10,897
79.0
7,894
95.6
72.4
120.9
6
52
0.4
13
0.2
25.0
41.7
10,949
79.4
7,907
95.7
72.2
120.6
17
216
1.6
43
0.5
19.9
33.2
11,165
80.9
7,950
96.2
71.2
118.9
43
782
5.7
147
1.8
18.8
31.4
11,947
86.6
8,097
98.0
67.8
113.2
26
833
6.0
111
1.3
13.3
22.2
12,780
92.7
8,208
99.4
64.2
107.2
14
172
1.3
16
0.2
9.3
15.5
12,952
93.9
8,224
99.5
63.5
106.0
42
168
1.2
15
0.2
8.9
14.9
13,120
95.1
8,239
99.7
62.8
104.8
46
308
2.2
13
0.2
4.2
7.0
13,428
97.4
8,252
99.9
61.5
102.6
24
101
0.7
3
0.0
3.0
5.0
13,529
98.1
8,255
99.9
61.0
101.9
20
72
0.5
2
0.0
2.8
4.6
13,601
98.6
8,257
99.9
60.7
101.4
34
193
1.4
5
0.1
2.6
4.3
13,794
100.0
8,262
100.0
59.9
100.0
13,794
100%
8,262
100%
CUSTOMER ADDRESS SENSITIVITY
Even though there is a "Data Confidentiality Agreement," provided with Dirmark's research,
some firms may have a corporate policy preventing the release of company data. Fortunately,
the regression model does not need company name, address, or phone data. The regression model
only needs a unique record ID number and all data elements that can predict the stated objective
(response, profit, size, SIC, etc.)
Moreover, there are numerous techniques for masking data that will assure complete confidentially
of a client's database. Thus, the regression model may state the following:
Node#14 = highest likelihood of profit
Node#14 + SIC + 23,49,58,7311,57, 59
and yellow +01,01,03
and years in business + 5+
and
red = $55
Number of "ids" within Node #14: 3,587
REGRESSION-DRAWBACK
The Regression Analysis is good in creating a profile of the ideal target market. However, because
the profile may consist of many qualifiers, finding a sufficient mail quantity may be difficult.
ADVANTAGES
Simply, a Regression will define your most profitable segments within your customers file and
allow you to apply these models to mail less and gain more profit.. In addition, a Regression
will also allow you to make your prospect lists perform more like your house lists, along with
identifying new prospect segments that were formerly unidentified.