Watched Star Trek with Aidan, Pat and Lke.

It is amazing. Its safe to say it is the best movie I’ve watched for the past 12 months (I haven’t watched many, actually).

I used to be a fan of Enterprise back in those days. I’m always amazed with quantum physics, astronomy and space/time travel. Why I didn’t pursue? Because in this field, you have to be the best. Its either hero or zero, there’s no 50% P, 65%C, 75%D.

I really liked the show and recommend anyone to go to the cinema to watch it.

A busy week is finally coming to an end. This leads us to a busy weekend, filled with meetings, appointments, yum char and pool sessions. No bowling today for whatever-god-kn0ws-what reason, but thats ok.

I’ll probably spend some time to complete my semester-long-ongoing Internet commerce assignment which requires us to collect articles (since 2005) regarding the subject matter. Must be a journal. Some people found the cheat to how to Google for it. I haven’t. So I’ll spend the weekend completing that task. I will have to do my Advertising Assignment #2 also which involves creating a promotional plan for a Broadband company.

After those two assignments are settled, then its time to hit the books in preparation for the final exams in 3 weeks time.

Oh, and to my sister who is reading this, remember our lunch date on Saturday. Preferbly in 1U / Ikea / PJ / Subang area lah, so its convenient for me. Thanks a lot Saturday is wesak day too.

Class start in a few moments. Anyway, since moving to the new house, I do not have internet access. I dongle my HP to Maxis’ 3G network, and subscribed to the Unlimited plan for about RM 3 a day. So about 1 week downtime (Thanks to Terrorkom and TM nut) I have to fork out RM 21 just to keep connected. Oh well, it is and has been a fun experience playing with 3G.

This is REALLY a true friend [link]

Good day everyone. Live long and prosper.

Related posts:

1. Phailed Mix May 2009
2. Events of early May 2009
3. Party 25 May 2009 @ 6pm
4. MTV VMA Awards 2009

I created a 33,000KB MS excel spreadsheet just to compute 1 thing. 1 fact.

E-mail and Snail mail are EFFECTIVE in getting Donations for Charity

Mmmm… I’ll be rich! But yeah, there are limitations… the other alternative was Telephone, and I prove that its not that good if you use Telephone.

One thing that would be totally awesome would be the Internet. I aligned my APM and DMM and IC and NF knowledge to deduce that internet is the way of the future. :”)
[sorry the images dun load properly cuz i didn't upload them]

Executive Summary

Moonlighters is a Not for Profit organization facilitating a system of giving money, food or help free to those who are in need because they are ill, poor or homeless, to provide money or help in this way. As a charitable organization, they rely on public donations to run their day-to-day activities.

During their recent mail campaign, only 5% of the donor list made a donation. On a database of 20,000 people, the net return was a paltry $10,000.

This report uses Microsoft Excel and some Statistical Analysis through Regression to find out a suitable model to explain and improve the current targeting methods. Out of the 26 variable provided, only 6-8 variables are useful in determining the amount donated. However, other demographical and socio-economical factors can help explain further on the models being used.

In short, we concluded that Mail and Email are the best ways to send promotional materials. Our ROI was between 5-8 times, whereas the potential for Telephone to convert into pledges was a very slim 2 times.

We mustn’t completely ignore the effectiveness of telephone. This is because, with a good telemarketer and an up-to-date contacts list and fact file, many people can be persuaded to contribute or donate.

Whilst the above 2 methods are good, the future lies within the Internet. With varying technologies in place, online marketing is the way of the future. Technologies such as social-networking sites like Facebook and MySpace helps links the wealthy and influential people on cyberspace. A lot of information can be displayed on a corporate website which will make the organization more transparent and relevant.

Another common promotional activity is through Blogs. Web logs or Blogs are websites created by an average Joe who loves telling the world about what has been happening. Various Web 2.0 technologies enable linking and searching of relevant and up to date sites. This introduces viral marketing where when someone learns about something interesting or beneficial, he can easily share it to the world, may it be through e-mail, Twitter, blogs, social-network sites, or social-news site. Once the awareness has been created in the blogosphere, the traditional News networks will pick it up, blowing the whole campaign to into an astronomical scale.

Introduction and Statement of the Management problem

This assignment requires us to find ways to better target our advertising campaign to obtain better returns.

A list of 20,000 records was given. We used a two-pronged attack configuration.

1. Remove all Non Donors (or people who have never donated)
We ended up with only 989 donors.
We then analysed on how best to reach them; either by Phone or Email or Mail

2. Using the same 20,000 records, we found out who can be reached; some by phone, some by e-mail and the rest by mail. We ended up with 11,067 records.

Some assumptions we have to make are:

1. Homephone is the only telecommunication option available to them. There is no mention of handphones, pagers or other methods which are common in today’s world. Therefore, these findings and models are based on Home telephone only, which not many people have nowadays.

On the same note, this could be an advantage because when a person is at home, he is usually relatively free and available to talk. On a handphone, he might be driving or having a class or meeting when the telemarketer calls.

2. Since this project is a Donation for Charity, the amount different people of different demographics gives may vary greatly. Not all donors are rich. Not all rich people are donors.

3. When disclosing information to researchers and marketers, people tend not to be entirely honest. Certain private information might be misconstrued.

4. Data might have changed between the time of capture and when the data is processed. A person might have changed job, changed house, bought a PC / phone, bought more Cars or a senior citizen might have died. To ensure a good model, the list must be current and updated.

Description of Analytical Model

Step 1: Creating dataset containing Current donors

Using the sort function to sort the Donation Column, I deleted all records with the values “0” meaning that this person did not donate to the charity.

Step 2: Variables which are useful indicators

I found that the variables Lifetime, Gifts, Average Gift, Promotions, RecentGifts, Sales, Cars, Technical are significant depending on the model used.

Years since First / Last is not that useful because donator’s behavior changes over time. In different phases of a person’s life, he might have heavier personal commitments like house and car loans, marriage or holiday expenditures. As long as a person has donated before, it is safe to assume they are willing to donate again.

It is harder to get non-donors to donate, and the Returns might not be as good.

The Major Donor Indicator was not used because only 48 records were found. There was no notable pattern observed.

Age is not a factor.

Home Phone / PC Owner / Catalog Shopper – these three variables do not affect whether or not a person donates.

MedHouseInc – Donation is voluntary. Not all donors are rich. Not all rich people are donors.

Occupation based percentages which are significant are included in the results below.

Percentage of Senior Citizens in the community did not affect the Regression results.

Education is also not a factor when targeting for donors.

Results

Basic Information

1. There are 989 Current Donors.

2. Using the CountIf function in excel, here are the results:

10,116 people can be contacted by phone
2,227 people can be reached by e-mail
1,276 of the above can be contacted by BOTH phone and e-mail

Of the 989 Current donors:

Total 989
Phone 476
Email 113
Phone & Email 58

No phone/email 458

Step 2: About the Model

As a general guide to whether or not a value is Significant, the t Stat must be greater than modulus / absolute 2. The P-value must be less than 0.05

Step 3: The models

Total Current Donor

Donations (Y) = 18.403 + 0.059 (Lifetime) – 0.542 (Gifts) + 1.317 (AvgGifts) – 2.935 (RecentGifts) + 29.483 (Sales)

By Phone

Donation (Y) = 3.998 + 2.407 (AvgGift) – 1.325 (RecentGifts) – 46.954 (Technical) + 32.577 (Sales)

By E-mail

Donation (Y) = 13.881 + 0.085 (Lifetime) + 1.201 (AvgGift) – 0.265 (Promotions) – 3.922 (RecentGifts) + 62.225 (Sales) + 33.913 (Cars)

Final Model Chosen

Out of TOTAL Current Donors

SUMMARY OUTPUT

Regression Statistics
Multiple R 0.711633
R Square 0.506422
Adjusted R Square 0.503911
Standard Error 16.44714
Observations 989

ANOVA
df SS MS F Significance F
Regression 5 272829 54565.81 201.7157 5.7E-148
Residual 983 265909.8 270.5084
Total 988 538738.8

Coefficients Standard Error t Stat P-value Lower 95%
Intercept 18.403 2.031 9.063 0.000 14.41797
Lifetime 0.059 0.009 6.586 0.000 0.041164
Gifts -0.542 0.096 -5.638 0.000 -0.7307
AvgGift 1.317 0.074 17.862 0.000 1.172694
RecentGifts -2.935 0.513 -5.718 0.000 -3.94203
Sales 29.483 9.603 3.070 0.002 10.63782

Donations (Y) = 18.403 + 0.059 (Lifetime) – 0.542 (Gifts) + 1.317 (AvgGifts) – 2.935 (RecentGifts) + 29.483 (Sales)

By Phone

SUMMARY OUTPUT

Regression Statistics
Multiple R 0.7478
R Square 0.5592
Adjusted R Square 0.5554
Standard Error 14.6394
Observations 476

ANOVA
df SS MS F Significance F
Regression 4 128046.9 32011.72 149.370 0.000
Residual 471 100940.5 214.3111
Total 475 228987.4

Coefficients Standard Error t Stat P-value
Intercept 3.998 2.893 1.382 0.168
AvgGift 2.407 0.114 21.125 0.000
RecentGifts -1.325 0.632 -2.097 0.037
Technical -46.954 27.071 -1.734 0.083
Sales 32.577 11.618 2.804 0.005

Donation (Y) = 3.998 + 2.407 (AvgGift) – 1.325 (RecentGifts) – 46.954 (Technical) + 32.577 (Sales)

By E-mail

SUMMARY OUTPUT

Regression Statistics
Multiple R 0.7480
R Square 0.5595
Adjusted R Square 0.5346
Standard Error 13.55
Observations 113

ANOVA
df SS MS F Significance F
Regression 6 24723.48 4120.58 22.443 0.000
Residual 106 19461.86 183.6025
Total 112 44185.35

Coefficients Standard Error t Stat P-value
Intercept 13.881 8.551 1.623 0.107
Lifetime 0.085 0.026 3.244 0.002
AvgGift 1.201 0.278 4.319 0.000
Promotions -0.265 0.109 -2.427 0.017
RecentGifts -3.922 1.332 -2.945 0.004
Sales 62.225 26.093 2.385 0.019
Cars 33.913 12.367 2.742 0.007

Donation (Y) = 13.881 + 0.085 (Lifetime) + 1.201 (AvgGift) – 0.265 (Promotions) – 3.922 (RecentGifts) + 62.225 (Sales) + 33.913 (Cars)

3. Conclusion / Recommendation

When doing regression analysis, we should be careful to never assume and generalise who will donate and who would not donate. Donation is a form of free will and although the data shows that donation is wholly contributed by specific groups, we can never be too sure to assume the other specific groups will not donate later on in the future.

The reason for them not donating could be probably because it is not the right timing at the time of survey.

Therefore with this theory we should still include the other specific groups but minimize the campaign cost on these “less-value” targeted groups. As for the specific groups that have been successfully donating, we should focus more on them by focusing our attention more to them.

Pros & Cons of 3 alternative campaign strategies:

Strategy 1: Mail

After filtering the Current Donors, we are left with 989 records. These people are the ones we intend to send mail to. The total cost is $2890.

The estimated donations received using the model is $27099 which garners a profit of $24209

Strategy 2: Telephone

There are 476 people from current donors who are contactable by Telephone.

Putting in the numbers into the model developed for Telephone, we get $6324 estimated donation giving $4325 profit.

Strategy 3: E-mail

Very few people have access to a Home computer. There were 113 records found. The donations estimated is $10414 with a profit of $8858

Below is a table with further information

TOTAL COST ESTIMATED DONATIONS Profits ROI (X)
Strategy A Mail
Fixed Cost 2000
Individual cost 0.9
Count 989 2890 27099 24209 8.4

Strategy B Telephone
Fixed Cost 1500
Individual Cost 1
Count 476 2000 6324 4325 2.2

Strategy C Email
Fixed Cost 1500
Email 1
Count 113 1557 10414 8858 5.7

From the previous table, we can deduce several interesting findings:

1. Traditional Mail is still the most effective way to encourage donation, closely followed by E-mail.

This is so because a lot of information can be delivered through these two media.

A potential donor can take his time to read the material in his leisure time and make a decision whether or not to donate. There is nobody rushing him, and he would psychologically feel more generous to have the proposal in front of him before committing an amount of money to the charity.

With modern software, email and letters can be personalized with salutations.

2. Telephone is not so successful because it is not the right way to sell or market to a client / donor. The target might be busy driving, cooking, chatting with friends, busy with work or having a bathe. If they are not reachable, they are unlikely to call back. Even if they do, the response would not be that good.

In general, tele-marketers have a bad name. They are usually associated with credit card scams, fake lottery winnings and pyramid schemes as well as unsolicited spamming or request for information via surveys. People usually hate it when they receive phone calls from marketers.

A grand total of 11,067 people can be contacted either by Home Phone or E-mail.

1,587 people have been accounted for in my models.

Therefore, a total of 9480 have yet to be reached or has not made a donation.

Recommendations

A. Who to TARGET?

We have identified 9,480 people being potential targets because they have not yet donated to the charity. There is no way to categorise this group of people because we do not know how much they are willing to donate.

As said earlier, not all donors are rich. Not all rich people are donors.

B. How to campaign (which methods)?

Based on my findings, the campaign to use is Postal Mail and E-mail.

Week 1 – Send out postal mail
Week 3 – Send out phone / e-mail reminders to remind the people whose records are available

C. Various alternative mixes to get the maximum donations.

The most important asset of this charity is the amount of existing donors (or people who have donated before). This is the most valuable group because they are known to be quite generous.

Given a budget, we should allocate about 60% of available funds to entice them (the 1,587 people) to give more donation.

The balance of 40% should be targeted to the 9,480 people who are potential targets but have not yet made a donation.

Another thing to look out for is % sales in the neighbourhood. This group of people have consistently been giving generously.

4. Technical Appendices – Methods

1. Make copies of the Original Data for manipulation.

2. Clean the list using the sort function
a. Make copies of the cleaned data

3. Use the sort function again to find HomePhone = 1 and PCOwner = 1
a. Make copies of the cleaned data

4. To build a model, select Donation as (Y) and up to 16 variables at one time.

5. There are some variables which are not able to be compute, like Age. In this case, we merely remove them from our model.

6. Using t Stat > |2| and P-value < 0.05, Anova value F > 5 and significance F < 0.05 , we can evaluate whether how good the model is.

7. R square is the percentage change of Y (in this case Donation) for a unit change of X (in this case, the 6-8 different variables chosen to be Significant)

8. To calculate Estimated Donations (assuming 100% response), we just apply the X value being the Count. For example, there are 989 people who can be contacted by e-mail, so we replace X with 989 in our model.

9. Residual Plot in a regression offers a good example of what a problem-free plot should look like. There is no odd fan or curved trends in the plot, the average of the residuals is zero, and the points are equally represented about the x-axis. This residual represents the difference between the observed response variable Y and the value predicted by the regression line.

10. Line of best fit (trend line) - A line on a scatter plot which can be drawn near the points to more clearly show the trend between two sets of data.

The line of best that rises quickly from left to right is called a positive correlation.
The line of best that falls down quickly from left to the right is called a negative correlation
Strong positive and negative correlations have data points very close to the line of best fit..
Weak positive and negative correlations have data points that are not clustered near or on the line of best fit.
Data points that are not close to the line of best fit are called outliers.

Formulae

1. Use
Data | Sort | Sort By
Then Select the Field to sort by.
It doesn’t matter ascending or descending, so as long as the Zeros are separated from the Ones. We then delete the appropriate values and records accordingly.

2. Using Cell-Referencing to ensure completeness and accuracy of data

='Current Donor FINAL'!B17+'Current Donor FINAL'!B18*Master!C13+'Current Donor FINAL'!B19*Master!C13+'Current Donor FINAL'!B20*Master!C13+'Current Donor FINAL'!B21*Master!C13+'Current Donor FINAL'!B22*Master!C13

In this example, the regression output on “Current Donor FINAL” sheet is used in the Master sheet pointing to the Count (number of donors). The predicted output is extremely accurate but for analytical purposes, we round it to 3 decimal places.

=ABS('Phone REGRESSION FINAL'!B17+'Phone REGRESSION FINAL'!B18*Master!C18+'Phone REGRESSION FINAL'!B19*Master!C18+'Phone REGRESSION FINAL'!B20*Master!C18+'Phone REGRESSION FINAL'!B21*Master!C18)

Sometimes in regression, a negative value may be obtained. In this case, the Absolute or Modulus function is used

Residual Plots

If the residual plot is random (and the conditions t Stat > |2| and P-value < 0.05, Anova value F > 5 and significance F < 0.05 are met), this variable is significant

Line Fit Plot

Note that there are many “outliers” where the blue dots do not align with the pink dots.

Due to the uncertain nature of donations being unpredictable, the R-Square for all models described in this assignment is between 50-55%. This means that based on a demographic or characteristic profile of a donor, there is a coin-flip chance that he will donate to our charity and proving that my model is correct.

[the end]

Student Name: Philip Khor
Student ID: 0602D56499
Tel: 016 9465088
E-mail: philip@taylorians.com

(word count: 2700 words)

Related posts:

1. Dice roll of Life
2. Pork Dinner
3. TM Net is pretty good
4. First time you met her

Related posts:

1. In response to Lukas Foo
2. Pool at Club 9
3. The Advertures of Lukas Foo: The Good Samaritan
4. Trapped in my TRAP!