You are on page 1of 5

Andrea Talic CRM Spring 2020

Gabriel Hansen Foundation


Individual Assignment 1: Customer Valuation and Prediction

Question 1
The signs tells you whether there is a positive or negative effect of each independent variable to the
dependent variable. A positive coefficient indicates that as the value of the independent variable
increases, the dependent variable also increase.

Question 2
For this question, we take the results from the outofsample prediction on Excel. It is also very
important to multiply by 200 each donor as each donor of the segment represents 200 donors in the
database.

Total amount of donations = 13’723,91€


Total number of active donors = 2293*200 = 458’600
Average amount donated = 5,985€
Revenue for active donors = 5,985*2293*200 = 2’744’782,87€
Variable costs = 1,10€*458’600 = 504’460€
Gross margin for active donors = 2’744’782,82€ - 504’460€ = 2’240’322,87€
Fixed costs = 7500€
Total net revenue for active donors = 2’240’322,87€ - 7500€ = 2’232’822,87€

Question 3
For this question, we take the results from the outofsample prediction on Excel. It is also very
important to multiply by 200 each donor as each donor of the segment represents 200 donors in the
database. Moreover, to have the active donors expected to be profitable, we have to only consider
those who have a predicted donation greater than 1,10€.

Total amount of donations from profitable active donors = 13’411,09€


Total number of profitable active donors = 1111*200 = 222’200
Average amount donated from profitable active donors = 12,07€
Revenue for profitable active donors = 12,07*1111*200 = 2’682’218,17€
Variable costs = 1,10€*222’200 = 244’420€
Gross margin for profitable active donors = 2’682’218,17€ - 244’420€ = 2’437’798,17
Fixed costs = 7500€
Total net revenue for profitable active donors = 2’437’798,17 - 7500€ = 2’430’298,17€

1
Andrea Talic CRM Spring 2020

We see that it is much more profitable to solicit only those active donors expected to be profitable. By
doing so, the total net revenue increases of 197’473,3€ (8,84% of increase).

Question 4

For this question, we take the results from the outofsample prediction on the second Excel. It is also
very important to multiply by 200 each donor as each donor of the segment still represents 200 donors
in the database. We also take the expected gross margin found in question 3.

Total amount of donations from cold donors = 73,87€


Total number of cold donors = 1023*200 = 204’600
Average amount donated from cold donors = 0,0722€
Revenue for cold donors = 0,0722*1023*200 = 14’773,82€
Variable costs = 1,10€*204’600 = 225’060€
Gross margin for cold donors = 14’773,82€ - 225’060€ = -210’286,18€
Fixed costs = 7500€ (already counted in the net revenue found before)
Total net revenue = 2’430’298,17€ - 210’286,18€ = 2’220’011,99€

We see that the operation of solicitating all cold donors can only cause a loss. By adding the cold
donors in their March campaign, their net revenue diminishes of 210’286,18€, being now of
2’220’011,99€ when before they got a higher revenue of 2’430’298,17€. This isn’t profitable for them
and, therefore, Gabriel Hansen Foundation should not spend resources trying to convince all cold
donors.

Question 5

For this question, we take again the results from the outofsample prediction on the second Excel. We
also have to multiply by 200 each donor for the same reason as before. We take also again the
expected gross margin found in question 3. Moreover, to have the cold donors expected to be
profitable, we have to only consider those who have a predicted donation greater than 1,10€.

Total number of profitable cold donors = 1*200 = 200


Average amount donated from profitable cold donors = total amount = 65,37€
Revenue for profitable cold donors = 65,37*1*200 = 13’073,79€
Variable costs = 1,10€*200 = 220€
Gross margin for profitable cold donors = 13’073,79€ - 220 € = 12’853,79€
Fixed costs = 7500€. (already counted in the net revenue founded before)
Total net revenue for active donors = 2’430’298,17€ + 12’853,79€ = 2’443’151,96€

We see that, by soliciting the profitable cold donors in addition, we get a higher net revenue
(2’430’298,17€ + 12’853,79€ = 2’443’151,96€). However, the increase in revenue represents only

2
Andrea Talic CRM Spring 2020

0,53%, so is it really worth it? Maybe not, as they represent cold donors, “re-activation” campaigns
may be needed and would represent additional costs.

Predictive probability that it will repond ?

 ajouter le calcul de regression mais vraiment utile ??

We would basically take our X variable (price of 20, no reminder of email, first possibility or second
possibility of add spend), then we plug that in in our regression model, do the logistic regression to
get the probability estimates and then we obviously multiply this with price and margin to obtain
estimates of profitability.

Question 6

The problem is that the same number of solicitation is send to donors having donated in the last 3
years without taking into acount any another information, when some donors have donated more or
less and frequently or not. A strong and special relationship should be build with frequent and loyal
donors, with more customized and gratitute messages ; while, a more persuasive marketing strategy
should be elaborated for less loyal donors. To adress the right message to the right donor, other data
about customers’ profiles should be added as, for instance, proven loyalty or average donation amount.
In fact, all attributes of the RFM model should be taken into account for clustering.
A modified version of RFM model could eventually be used here : the Weighted RFM version. Each
variables of the model have a different weight which reveal its relative importance and help the
clustering analysis and the evaluation of Customer Lifetime Value. 1
Moreover, classification rules, determined by considering demographic data (as age, gender, etc.) and
RFM vairables would, help in the prediction of future customer behavior and in a clearer targeting of
donors’ profiles.1
Furthermore, associating the Markov Chain Model could help managers consider the probability of
donors’ behavior changes at some periods and allocate some additional (e.g. re-acquisition) marketing
strategy to change customer profile and become high value donors.
In addition to all this, other data may appear valuable, as the reason of the donation which generally
impact frequency and amount of donation. 2 Also, scoring tests shoul be repeated to have an eye on the
most effective and impactful campaigns, as donations vary according to the time of the year. 3
The proposed models could help managers to target more easily the right donor profiles and to develop
for them efficient marketing strategy resulting from effective prediction of donors’ behaviors based on
weighted RFM variables and demographic variables. Different marketing campaigns according to
different segments could then be set.

Question 7

3
Andrea Talic CRM Spring 2020

The scoring approach with the addition of the data and models proposed in question 6 should be
generalized to all campaigns, so the management of the foundation can adapt throughout the time and
have a clear targeting strategy. Their initial strategy of targeting only active donors was not profitable
enough. In addition, donor behavior can change over time, leading to wrong predicted data or
interpretation. That’s why, it is important to keep track of results of each and every campaign.
To efficiently allocate resources, the management has to have a clear segmentation with full
understanding of each segment’s differentiations. Marketing strategies have to be adapted according
to the targeted donors. Indeed, different approaches should be taken depending on the segments as
loyal customer would need less “shaping behavior strategy” than other more resilient potential donors.
It is very difficult to shape donor behavior by mailing. For this reason, it is really important for the
management to base their marketing strategies on scoring tests and analysis. Campaigns and
solicitations should then be adapted based on this and could vary between reminders, re-activation,
gratitude messages, calls, etc. With the right strategy, overtime, the management would eventually
manage to shape some behaviors and increase the number of high value donors.

4
11

1
Derya Birant (2011). Data Mining Using RFM Analysis, Knowledge-Oriented Applications in Data Mining, Prof. Kimito
Funatsu (Ed.), ISBN: 978-953-307-154-1, InTech, Available from: http://www.intechopen.com/books/knowledge-oriented-
applications-in-data-mining/data-mining-using-rfm- analysis

22
« A donor who has made regular and generous donations to the Gabriel Hansen Foundation over the last 15 years,
without discontinuity, is much more likely to make future significant donations than a new donor who has made a first,
paltry donation after the passing away of a distant cousin. » p.4 from the case

33
« Fundraising campaign activity varies during the year, with most of the activity occurring near the end:
between 50 and 65% of donations occur during the last three months of the year. » p.3 from the case

You might also like