Page 1

Row

Total Number of Control Non-Responders(Do Not Distrub)

46912

Total Number of Control Responders(Sure Things)

4519

Total Number of Treatment Non-Responders(Lost Causes)

12207

Total Number of Treatment Responders(Persuaders)

2863

Row

Interpretation of Output Variables: Categorization of population classes using Probability of Control Responders and Treatment responders.

Output Results

Prob of Do_not_Distrub Prob of Sure Things Prob of Lost Causes Prob of Persuaders Uplift Score
0.4805905 0.0577063 0.3847666 0.0769365 0.1150542
0.6084216 0.0336630 0.3021231 0.0557923 0.3284278
0.3867488 0.0411368 0.4437217 0.1283927 0.0302829
0.4196815 0.0286245 0.4637970 0.0878969 0.0151569
0.4869971 0.0486918 0.4020364 0.0622747 0.0985437
0.6850376 0.0344083 0.2466115 0.0339426 0.4379605
0.5148878 0.0493959 0.3789463 0.0567700 0.1433156
0.2755052 0.0269928 0.5897287 0.1077733 -0.2334431
0.4241754 0.0262042 0.4782044 0.0714160 -0.0088172
0.5148878 0.0493959 0.3789463 0.0567700 0.1433156
0.4869971 0.0486918 0.4020364 0.0622747 0.0985437
0.5148878 0.0493959 0.3789463 0.0567700 0.1433156
0.4241754 0.0262042 0.4782044 0.0714160 -0.0088172
0.2739331 0.0269091 0.5909066 0.1082512 -0.2356315
0.6825135 0.0348184 0.2487732 0.0338949 0.4328168
0.2833884 0.0286967 0.5807494 0.1071655 -0.2188922
0.2751879 0.0270109 0.5892397 0.1085615 -0.2325012
0.6856817 0.0722518 0.2059692 0.0360972 0.4435579
0.4131307 0.0270368 0.4874240 0.0724086 -0.0289215
0.5330733 0.0373427 0.3765794 0.0530045 0.1721556
0.3687231 0.0359342 0.4927392 0.1026036 -0.0573467
0.4226642 0.0259128 0.4813038 0.0701193 -0.0144331
0.6828906 0.0743649 0.2043549 0.0383896 0.4425605
0.2534098 0.0335045 0.5451746 0.1679111 -0.1573582
0.2597467 0.0338699 0.5404238 0.1659595 -0.1485875
0.2597467 0.0338699 0.5404238 0.1659595 -0.1485875
0.2597467 0.0338699 0.5404238 0.1659595 -0.1485875
0.2534098 0.0335045 0.5451746 0.1679111 -0.1573582
0.3927491 0.0365136 0.4782639 0.0924733 -0.0295551
0.7355227 0.0834593 0.1505499 0.0304681 0.5319816
0.7355227 0.0834593 0.1505499 0.0304681 0.5319816
0.6953703 0.0360191 0.2325735 0.0360371 0.4628148
0.3927491 0.0365136 0.4782639 0.0924733 -0.0295551
0.6710938 0.0358577 0.2478039 0.0452445 0.4326766
0.5365250 0.0399449 0.3525219 0.0710082 0.2150663
0.6710938 0.0358577 0.2478039 0.0452445 0.4326766
0.2582233 0.0315125 0.5531451 0.1571192 -0.1693151
0.4087671 0.0307804 0.4623286 0.0981239 0.0137820
0.5365250 0.0399449 0.3525219 0.0710082 0.2150663
0.6893245 0.0890968 0.1860547 0.0355239 0.4496968
0.4687576 0.0568681 0.3776929 0.0966813 0.1308779
0.5400997 0.0398055 0.3493206 0.0707742 0.2217478
0.6893245 0.0890968 0.1860547 0.0355239 0.4496968
0.2623532 0.0372835 0.5353695 0.1649938 -0.1453060
0.4073443 0.0332310 0.4505225 0.1089022 0.0324930
0.2623532 0.0372835 0.5353695 0.1649938 -0.1453060
0.6613244 0.0426532 0.2626069 0.0334155 0.3894797
0.6841373 0.0663693 0.2160184 0.0334750 0.4352246
0.4896471 0.0478632 0.4000825 0.0624072 0.1041088
0.6841373 0.0663693 0.2160184 0.0334750 0.4352246

Page 2

Row

Categorization by Gender: Mens are responding more than womens for the offers

Categorization by Offer type: Most of population is responding for Discount offer type rather than information and bogo.

Categorization by Income: We need to target population who are more than 40K.

Row

Age wise Distribution: We need to target most of 40-80 age people in the population.

Categorization of Days as member: Members who joined <1500 days are interested for most of the offers.

Categorization by Channel-wise: Web and email are two important channels where most of the population are responding for the offers.

ROI

Row

Total amount Invested on Treatment Responders(Persuaders)

397750

Total amount Invested on Treatment Non-Responders(Sure Things)

1715800

Total amount loss

1318050

Row

Amount Spend on Individual offer types of Treatment Responders

offer_type x
bogo 114600
discount 219150
informational 64000

Amount Spend on Individual offer types of Treatment Non-Responders

offer_type x
bogo 520100
discount 833700
informational 362000

Amount spent on different channels

Page 3

Row

Month wise Number of Users: Most of users started joining from 2016 and Number of users has been increased gradually from 2017 also.

Page 4

Row

Year Wise Number of Users:

Timeseries plot

Page 5

Row

MAPE

76.78

RMSE

5.84

Row

Forecasted Output: Forecasted for Number of users going to join for next 6 weeks

Prediction of Number of Users in next 6 weeks: Prediction of Number of users going to join next 6 weeks with +/- 95% Confidence interval.

Future_Week Forecasted Value +95% CI -95% CI
2018-12-08 8.720730 2.692382 14.579222
2018-12-09 10.555102 4.962755 16.103870
2018-12-10 4.078022 -1.504871 9.886755
2018-12-11 7.302788 1.652872 13.355395
2018-12-12 8.530066 2.981126 14.233780
2018-12-13 8.207780 2.750743 13.625855

Page 6

Row

Uplift score interpretation: We need to concentrate on the high value of Uplift score from 0.4+

Qini Plot: When the population percentage is 40-80% there is high chance of Uplift score.

Page 7

Row

Response Rate per Decile

Uplift per Decile

Descrimination of Target class using LDA