Data

Go Beyond Engagement Scores

Use AI for insights

Daniel PappOctober 07, 2019

Universities from Amherst to Princeton are currently harnessing engagement scores to quantify their alumni's involvement with their institution. An alumni attends a university event—add one point. Hasn’t donated in the last 5 years—subtract two. Unfortunately, because these weighting systems are often selected intuitively, they are imperfect in predicting how willing to give your alumni really are. Since alumni relationships with their alma mater are intangible, quantifying these bonds can be a logistical nightmare and a drain on limited university resources.

For this reason, efforts have been made to standardize a disorganized engagement score playing field. For instance, the Council for Advancement and Support of Education (CASE) has recently embarked on a project to standardize engagement metrics throughout universities. The project strives to create a singular, informative set of engagement metrics that could be easily implemented by institutions across the globe.

However, even though this rigorous examination of engagement scores would likely decrease the error rate involved with the scores’ implementation, the data-collection process remains unoptimized. Tracking the effectiveness of engagement activities to assess engagement models over time is time-consuming, and requires a significant allocation of university staff efforts. In fact, tangible results may not be seen for years. For advancement professionals, quick results could make a million-dollar impact.

Fortunately, there is a faster, more accurate way to measure alumni engagement, while using a fraction of the resources. The surprising answer to your engagement metric woes? AI (Artificial Intelligence)

The Power of Prediction

Despite the wealth of information available to alumni professionals, human guesswork abounds in the field of advancement. Alumni professionals speculate freely about what factors increase and decrease the likelihood of giving, while often lacking the results to back up their claims. Such approaches are inefficient and leave the data-rich industry of higher education largely untouched by advances in predictive technology. However, this pattern is ripe for disruption. Without deviation from the norm, progress is not possible.

Industries from retail to medicine have already adopted AI technology, which they use to consistently provide the most relevant information to every consumer. The implementation of AI allows businesses to save considerable money and effort, with automation replacing countless hours of human labor. In fact, 72% of business executives stated in a survey by the Napa Group that they believe “AI will enable humans to concentrate on meaningful work and will be the business advantage of the future.”

Here is an example of machine learning in action. Recently, a small non-profit decided to implement predictive analytics to improve its member acquisition. As the AI technology analyzed response rates by zip code and zone level, the company was able to identify optimal targeting strategies for each zone and maximize its ROI. After only a couple of months of using the technology, the non-profit saw a 90% improvement in member acquisition and a drastic reduction in acquisition costs. The automation of a previously cumbersome process saved a struggling organization significant time and money, and the non-profit thrived through a greater understanding of its potential donors.

Such success is possible for any business willing to take the leap into AI. Here are the 4 steps necessary for a college to implement predictive analytics and maximize fundraising: 

1. Integrate Donor Data

Although volumes of information are already available within university databases, they are often not used to their full advantage. By organizing and analyzing alumni data, you can unlock deep insights that will be invaluable to your advancement team. For example, some key data points that CueBack includes in its predictive model are:

• Demographic data - such as age, income, occupation, family status, business, and personal relationships 

• Donation data - such as past outreach, responses, and giving history

• Behavioral data - when collected through CueBack’s engagement application, gives insights into alumni affinities, interests, connections, event attendance, needs, and preferences 

Since predictive models depend on inputted data to make accurate predictions, the model will be more effective as the number of various inputted data points increases. For this reason, the process of collecting data for optimizing the model may take time—and that is normal. Great results are not achieved overnight, but the insights gathered will pay off in the long run.

2. Harness Predictive Power

The information you put into your model can now be put to work. Your team will greatly benefit from the model’s ability to predict which alumni will be most likely to give back to their alma mater, allowing you to maximize the effectiveness of your advancement efforts.

Moreover, predictive analytics can predict how much an alumni is likely to give. This enables your university to optimize your Advancement efforts by allocating more resources to high-value donors. Once you establish a base of receptive alumni donors, the model can predict which campaigns and messaging will connect most with them, thus optimizing the decision-making process and increasing donation velocity.

Furthermore, the predictive model can help your team with donor segmentation. For example, a new campaign highlighting the diverse cultures of your university’s student body (such as the University of Sheffield’s #WeAreInternational), would likely be most meaningful to alumni from international or immigrant backgrounds themselves. A predictive model would help you identify and target these alumni, helping you consistently provide meaningful content that builds stronger, long-term relationships with your graduates.

Whereas engagement score calculations rely on a handful of arbitrarily chosen, easily measurable variables, predictive models can incorporate a virtually unlimited number of variables, identified and tested by the models themselves. There is no “guesswork” in machine learning models, no exhausting meetings in which it is decided that how often an alumni likes their alma mater’s Facebook posts must predict their likelihood of donating. Instead, machine learning discovers which variables are good predictors based on statistical analysis, testing each variables’ correlation with the desired outcome (in this case, increased alumni donations). Through this rigorous testing, the model is able to weigh these variables in accordance to their measured predictive power—not through random score assignments.

Furthermore, machine learning models are meant to excel at their titular function—learning. These models continually learn and improve based on the data provided to them; the more data they have, the more effective their predictions. Rather than remaining static and immovable, the model changes to fit current giving climates and alumni attitudes. As your alumni change, so do your predictions.

3. Personalize Donor Interactions

Your alumni are used to having a personalized online experience—and your university should not be any different. Ensure that the content you provide aligns with your alumni donors’ interests, every time.

Consider online shopping. As you click through products that interest you, ranging from skin creams to windshield wipers, the site’s algorithm collects information about your habits and interests. For example, you consistently browse kitchen products, it may conclude that you are an avid household chef. After gathering data about your online behavior, the site will recommend products based on your interests: KitchenAid mixers, Mediterranean cookbooks, colorful oven mitts, and lavender-scented dish soap. You, notorious among your friends and family for your love of cooking, will benefit from viewing relevant products without a cumbersome search process. At the same time, the site benefits from recommending items that you are most likely to spend money on. The more time you spend on the site, the more data the algorithm collects—and the better the shopping experience it provides.

Now, consider a different transactional experience: university donations. When alumni interact with your university online, they need a platform that tailors content to their interests in the same way online shopping does. The site’s algorithm should track the alumni’s online behavior over time: who they connect with, what topics they show interest in, and what content they share with others.  Based on this information, the algorithm only serves up the most relevant content to the alumni: fundraising campaigns aligned with their interests, pertinent archival material, and reunion events for the activities that had a lasting impact on them.

This targeted approach will ensure that alumni consistently see the content that appeals to them on an emotional level. Instead of scrolling through dozens of irrelevant posts, detailing reunions for clubs they never participated in or fundraisers for departments they were never a part of, alumni would only see content that has personal meaning to them. The online behavior of your alumni offers a window of insight into their emotions, preferences, and connections. An online retailer wouldn’t barrage a teenage girl with lawnmowers,  white New Balance sneakers, and barbecue grills, though a suburban dad would likely be interested in purchasing all three. Effective alumni engagement and fundraising, just like shopping, depend on knowing your audience.

4. Implement What you Learned

The last step to machine learning success is arguably the most important. To ensure that you are making the most out of your newfound insights, it is crucial to integrate your findings back into the analytical process to improve future performance. As you add more data sources to your model while refining your existing ones, you can dramatically improve the model’s accuracy.

Furthermore, predictive models can identify KPIs (key performance indicators) that will help you analyze the success of your marketing or campaign initiatives. As you isolate what worked and what did not within these campaigns, you will be able to put on more effective initiatives going forward—reducing your costs and improving efficiency.

Your predictive analytic insights can help your university answer key questions about how to allocate resources and what returns to expect from investments. These insights will dramatically improve planning, budgeting, and forecasting processes, as they can be used to identify how much of the budget to allocate to specific campaigns or to predict the costs of soliciting a certain category of donors.

However, as you receive more campaign data and donor information, it is crucial to continually revise the model. The major advantage of machine learning models is their ability to continually improve and change—and your university should take full advantage of that.

Facebook Isn’t Your Facebook

You may hesitate at the prospect of utilizing an independent prediction model, such as the one built by CueBack. After all, social media conglomerates such as Facebook already have algorithms that use information about user behavior to serve up the most relevant content. However, your university never receives the data Facebook collects about your alumni—instead, the company saves that information for itself, exploiting user data as a massive source of revenue.

On Facebook’s platform, you ultimately have no control over how much alumni data you receive, or what that data is then used for. As your alumni interact with your university’s page, Facebook uses that valuable data to advertise to your alumni on other sites. Worse, it may sinisterly sell alumni data to other companies such as Cambridge Analytica.

Your alumni are yours, not Facebook's. With CueBack, all alumni data belongs to the college. Moreover, you have control to shape the content through cues and prompts. The granular insights that are elicited through CueBack can then be utilized to serve a wide range of your college's interests, such as Advancement, Recruiting, Athletics and Career Services.

A Future of Success

The future of technology is here, and there is no reason for universities to fall behind. Given all the fundraising challenges faced by universities, engagement scoring systems that rely on the one-size-fits-all model of Facebook will not give your university the advantage it needs to thrive. After all, the predictive powers of AI have the power to make all human forecasting obsolete, with the accuracy and efficiency of the technology upending traditional methods of prediction. Rather than fruitlessly competing with such advanced technology, universities should look to adopt machine learning into their advancement processes.

By following the 4 step process outlined in this article, any university can implement machine learning technology to maximize the effectiveness of their marketing campaigns, increase the ROI of fundraising initiatives, and improve alumni engagement. Abandon engagement scores, and step into the future. Let CueBack help you leverage AI to increase giving, maximize efficiency, and decrease your costs.

About the Author

Daniel Papp

CueBack COOHi, I’m Daniel, the COO of CueBack. I love dogs (especially mine), whiskey, and Muay Thai. I enjoy reading anything, including nutritional information, higher-ed dissertations, and machine learning research papers. If you’d like to chat, connect with me on LinkedIn and send me a message!