Implementing a Scoring Algorithm of NJ SEEDS Alumni

I recently got the chance to attend this year’s NJ SEEDS Alumni Association’s alumni retreat. NJ SEEDS helps bright students from low-income families improve their life outcomes by putting its scholars into environments that increase students’ chances of getting into highly competitive colleges. Anyway, the alums got to brainstorm ideas that we would like to try in order to improve alumni commitment to the program. One thing that was mentioned was the idea of a scoring algorithm that stands as a proxy for how committed an alum is to the organization; apparently Stanford does this with their alumni for a variety of reasons:

They quantitatively know which alums are the most likely to help out when the organization needs something, because people who have invested a lot in terms of time, energy, and/or money probably enjoy doing so and therefore are likely to keep investing and supporting the organization.
If Alum $X$ needs something from the organization, they know how invested they should be in entertaining a graduate’s request.

Other reasons to have a scoring algorithm might be to actually incentivize good behavior and utilize the competitive nature of SEEDS alum, like how Fitbit shows individuals how they are doing every week, compared to their peers:

Boat-Shoe Badge by Fitbit — Reward for reaching a milestone

Weekly Stats by Fitbit — Show stats and encourage friends to push each other

In this article, I propose a simple algorithm that takes into account several features that signify alumni behavior that is beneficial to NJ SEEDS.

Algorithm

Version 1: All-Time

An alumni’s score can take into consideration the following features:

total amount of money donated
total number of times donated
number of events attended
number of events hosted
average number of alum “pulled” in per event

The first two metrics are money-related. Number of events attended, number of events hosted, and average number of alums “pulled” into events, on the other hand, focus more on social aspects that I think are important in maintaining and improving relationships within alumni.

Using the total amount of money donated as a metric is obvious – more money donated to SEEDS is a good thing and helps the non-profit organization run.

The total number of times donated, on the other hand, is less obvious. The benefit of this metric is that alums who might not be in good financial status – and therefore, might have a harder time donating money – can still do well in supporting SEEDS. A lot of potential donors apparently decide whether or not they are going to contribute on the basis of alumni participation in donations. In this case, the total amount of money donated by one alum is not as important as having very many alum donating, even if the donations are individually small.

Number of events attended measures the time commitment of alumni to the organization. This could be related to social events or even volunteering events that help SEEDS in one shape or form.

Number of events hosted rewards alumni who are hosting events. A problem I see right now is that the people hosting alumni events might be “pulling” in the same people most of the time. By having other people host events, they might be able to “pull” people that might not show up otherwise.

Lastly, the average number of alum “pulled” in per event is a measure of social influence. If person $A$ invites person $B$ to a SEEDS event, and person $B$ says in a survey that part of the reason he/she/(insert preferred pronoun here) attended an event is because of Person $A$, then person $A$ has “pulling” power. Over time, if it becomes a habit for Person $B$ to attend events, and person $B$ attends events because it is a habit and does not list person $A$ as a reason of attendance, then Person $A$ has to bring in other alumni that haven’t been going to the events, in order to improve performance on this metric.

Examples

For example, let’s say that Jenny, an alumnus who has contributed so much to the organization in countless years has donated a total of $8,000, donated a total of $30$ times, attended $50$ SEEDS events, hosted $3$ events, and had an average of bringing $4$ alumni to events:

$\begin{align} J_{total\ donated\ dollars} &= 10,000 \\ J_{donation\ absolute\ frequency} &= 30 \\ J_{events\ attended} &= 50 \\ J_{events\ hosted} &= 3 \\ J_{avg\ num\ alum\ pulled} &= 4 \end{align}$

A recent and excited graduate of SEEDS, George, has donated $$11,000$, donated $5$ times, attended $3$ SEEDS events, hosted $0$ events, and had an average of bringing $0$ alumni to events:

$\begin{align} G_{total\ donated\ dollars} &= 11,000 \\ G_{donation\ absolute\ frequency} &= 5 \\ G_{events\ attended} &= 3 \\ G_{events\ hosted} &= 0 \\ G_{avg\ num\ alum\ pulled} &= 0 \end{align}$

To calculate the scores for each person, we find the maximum values of each feature:

$\begin{align} max(total\ donated\ dollars) &= 11,000 \\ max(donation\ absolute\ frequency) &= 30 \\ max(events\ attended) &= 50 \\ max(events\ hosted) &= 3 \\ max(avg\ num\ alum\ pulled) &= 4 \\ \end{align}$

We then divide each person’s performance in each category by the corresponding maximum values and average them out.

Jenny’s score can be calculated as follows:

$\begin{align} score(J) &= \frac{\frac{J_{total\ donated\ dollars}}{max(total\ donated\ dollars)} + \frac{J_{donation\ absolute\ frequency}}{max(donation\ absolute\ frequency)} + \frac{J_{events\ attended}}{max(events\ attended)} + \frac{J_{events\ hosted}}{max(events\ hosted)} + \frac{J_{avg\ num\ alum\ pulled}}{max(avg\ num\ alum\ pulled)} }{5} \\ &= \frac{\frac{10,000}{11,000} + \frac{30}{30} + \frac{50}{50} + \frac{3}{3} + \frac{4}{4} }{5} \\ &= \frac{10/11 + 4}{5} \\ &\approx 0.982 \end{align}$

The number reflects the idea that Jenny is a superstar alumni.

Let’s calculate George’s score:

$\begin{align} score(G) &= \frac{\frac{G_{total\ donated\ dollars}}{max(total\ donated\ dollars)} + \frac{G_{donation\ absolute\ frequency}}{max(donation\ absolute\ frequency)} + \frac{G_{events\ attended}}{max(events\ attended)} + \frac{G_{events\ hosted}}{max(events\ hosted)} + \frac{G_{avg\ num\ alum\ pulled}}{max(avg\ num\ alum\ pulled)} }{5} \\ &= \frac{\frac{11,000}{11,000} + \frac{5}{30} + \frac{3}{50} + \frac{0}{2} + \frac{0}{4} }{5} \\ &= \frac{1 + \frac{5}{30} + \frac{3}{50}}{5} \\ &\approx 0.245 \end{align}$

So even though George was able to raise the most amount in terms of donations ($$11,000$), he did not perform as well as Jenny because she performed strongly in all the metrics – not just one section. The benefit of this approach is that it considers the long term performance of alumni, and is biased towards high-performing senior alum. A senior alum who has been active in the community many years will generally have a higher score than someone who just became an alumni but performs strongly. However, this benefit could also a weakness; new members, even though they might be giving their all in supporting SEEDS as alum, might be turned off because their current efforts are just not as successful as the seniors because the seniors have had years of a head start.

Version 2: Averaged Per Year

Maybe we could augment this all-time scoring algorithm with one that takes more averages into account. Junior members of the SEEDS community can better compete with veteran alums by considering:

average amount of money donated per year
average number of events attended per year
average number of events hosted per year
average number of times donated per year
average number of alum “pulled” in per event.

Examples

A scenario for Jenny and George in this case could be the following:

Jenny raised an average of $$750$ per year, attended an average of $5$ events per year, hosted an average of $2$ events per year, donated an average of $5$ times per year, and brought $4$ other alumni. On the other hand, George has raised $$4,000$ per year, attended $4$ events per year, hosted $0$ events, donated $5$ times per year, and brings $5$ alumni to events per year on average.

Calculating the score this way is as follows:

$\begin{align} score(J) &= \frac{\frac{J_{avg\ donated\ dollars}}{max(avg\ donated\ dollars)} + \frac{J_{avg\ donation\ freq}}{max(avg\ donation\ freq)} + \frac{J_{avg\ events\ attended}}{max(avg\ events\ attended)} + \frac{J_{avg\ events\ hosted}}{max(avg\ events\ hosted)} + \frac{J_{avg\ num\ alum\ pulled}}{max(avg\ num\ alum\ pulled)} }{5} \\ &= \frac{\frac{750}{4,000} + \frac{5}{5} + \frac{5}{5} + \frac{2}{2} + \frac{4}{5} }{5} \\ &= \frac{3.9875}{5} \\ &= 79.75\% \end{align}$ $\begin{align} score(G) &= \frac{\frac{G_{avg\ donated\ dollars}}{max(avg\ donated\ dollars)} + \frac{G_{avg\ donation\ freq}}{max(avg\ donation\ freq)} + \frac{G_{avg\ events\ attended}}{max(avg\ events\ attended)} + \frac{G_{avg\ events\ hosted}}{max(avg\ events\ hosted)} + \frac{G_{avg\ num\ alum\ pulled}}{max(avg\ num\ alum\ pulled)} }{5} \\ &= \frac{\frac{4,000}{4,000} + \frac{5}{5} + \frac{4}{5} + \frac{0}{2} + \frac{5}{5} }{5} \\ &= \frac{3.8}{5} \\ &= 76\% \end{align}$

In the case of averaging, veteran superstar alumnus Jenny and junior superstar alumnus George do comparably well.