Going Beyond Predictive Analytics in Higher Education

Higher education has evolved to be a data informed industry. Many institutions have implemented data analysis for decision-making after having been an anecdote driven industry for decades. It’s a great evolution in higher education’s data journey. Even with predictive analytics in place, however, that’s not enough to know how to support students.   Predictive analytics … Read more

Why Mindset Matters Most

Today’s consumers demand personalization, and learners are no different. Schools and colleges often use demographic data to personalize learner engagment, but that rarely provides a truly personalized experience for students 


Demographic data doesn’t tell the whole story. It provides important context about a person’s life and surroundings but doesn’t tell us who a person is on the inside. Mindsets, on the other hand, do. Using mindset and behavioral data allows you to humanize your dataset and offer learners a truly personalized experience.  


Mindsets are beliefs and attitudes we hold about the world and ourselves that affect our decision making and behavior. Public discourse is now more focused on acknowledging people’s demographic identities and experiences. While important to understand a social context through demographics, growing evidence suggests it is not enough.  


Why demographics don’t tell the whole story  

Mindset data can often tell us more about people than demographics can. In one study, researchers found that mindsets were twice as predictive of test scores when controlling for all other factors. Students with a highly motivated mindset who attended schools with low performance outcomes achieved similar outcomes to students who attended high performing schools. This mindset had a similar effect to vaulting into a higher socioeconomic bracket. Mindsets clearly play a powerful role in a student’s life.  


TikTok research is also showing us that mindsets transcend demographics. TikTok consumer research on the European sector showed that actions and mindset were more important than age in predicting user behavior. They identified four mindsets of TikTok users that transcended demographic categories. Using demographics in data has been the primary strategy for many data-driven decision makers, but Head of Europe, Global Business Solutions at TikTok says that “In reality, we are really just ascribing them [people] to all sorts of traits and behaviors that could apply to anyone.” In our everevolving social world, mindsets can tell us more than just demographics can.  


Many mindset challenges can also be present across demographics. A growth mindset–the idea that skills and abilities are developed over time instead of being stagnant–is found in people across all demographic categories, as is procrastination. And psychological phenomena like imposter syndrome or stereotype threat can be more common among underrepresented groups like women and people of color. Even so, they occur across these demographic groups. 


Demographics don’t tell the whole story because they overlook the individual. People can be from the same demographic group and still think and behave differently.  


Why mindsets matter 

Mindsets offer insight on people’s values and help us understand them on a deeper level. Demographic categories can’t fully explain why people interpret events the way they do in the same ways mindsets can 


Take a pair of identical twins—they grew up in all the same demographic categories but have distinct mindsets, so each sibling will approach situations differently. If Twin A’s has a growth mindset and Twin B’s mindset is anxious, how we encourage each of them to take advantage of tutoring resources will differ. The table below exemplifies how messaging differs by mindset.  









Twin A  


Growth mindset 


Need help sharpening your skills? Sunset College’s peer tutoring program is a great resource to help you work on topics that are challenging for you. Make an appointment with a tutor here.  


Twin B 




It can feel overwhelming when you’re struggling with a topic in class. That’s ok. Sunset College’s peer tutoring program can pair you with a student who can guide you through your challenges. You can make an appointment here.   


Despite being from the same demographic categories, these twins feel very differently about their challenges in a class. Demographics based, one-size-fits-all messaging misses key differences between the two individuals, but mindset based messaging gets to the core of how they feel and behave.  


Using demographic data is also increasingly complicated because people can defy stereotypes or assumptions that go along with these demographics. Increasingly, our idea of what a demographic group is like is not nuanced enough.  


Mindsets, on the other hand, are not tied to one demographic group. Mindsets explain what motivates an individual, what keeps them from doing something, and shed light on their perspective. They provide key data that informs how to effectively communicate with someone.  


Carol Dweck’s game-changing research on mindset is a prime example of how mindset can transcend demographics and provides personalization. In one case study, students across various demographic groups such as income level, race, and English learner status were exposed to growth mindset lessons. Those with a growth mindset reported higher standardized test scores and growth. Embracing mindsets and behavior humanizes data and avoids the oversimplifications of demographics. To put it simply, mindsets truly explain the “why” of human behavior. 


Humanizing data driven decision making  

Data driven decision-making is the right way to go, but it’s important to focus on the right kind of data. Mindsets are powerful and can say more about an individual person than just their demographics can. Humanizing data by using behavioral and mindset data unlocks a new level of understanding that can help learners get closer to meeting their goals.  

How AI Enhances Advisors’ Ability to Support Students

It’s easy to feel that academic, financial, and career advisors in higher ed could be replaced by AI, especially with all the talk about ChatGPT and generative AI. As more colleges and universities adopt technology to help achieve desired student outcomes, this future seems more like an upcoming reality.   But the adoption of AI should … Read more

Why Behavioral Data is Better Than Demographic Data

Data is essential to understanding today’s learner. What type of data is used is equally important.  


Education has made the shift from anecdotal and descriptive data to a more predictive data focused approach over the past decade, but more needs to be done to embrace different types of data and analytics approaches that enable better understanding, support, and connections between instructors and students. That’s where behavioral data comes in.  


What is behavioral data?  

Behavioral data is data on a person’s engagement with websites, resources, a school campus, and more. Most of these data are currently collected through a customer relation management (CRM) platform, websites, LMS and automated systems, among other sources. Behavioral data is dynamic—it represents what a person does while they are engaged in their learner journey. Demographic data, on the other hand, is stagnant data that segments people into different groups based on demographics such as age, gender, race, sexual orientation income, and zip code.   


What demographic data does  

Demographic data can be useful for generating insight into a person’s circumstances–it can tell you about their age, where they live, or what their gender identity is, for example. Demographics can provide additional contextual awareness of a student and what she may be experiencing in her individual learning journey. This type of information can also help understand high level group experiences 


But demographic data also has downfalls. Some demographic data serves as a proxy for information without truly telling you about a person or digging deep into the diversity of a large group. Gender, for example, can serve as a proxy for preferences. Using traditional demographic approaches could lead marketers to advertise baby products to middle-income women ages 20-35. Even with disaggregating the data, they could end up wasting their time and money advertising to women who don’t have children or are not interested in having children. If these same marketers used behavioral data instead, they could advertise to people who have shown interest in baby products and target older women with children, current or future fathers, and non-binary parents. Behavioral data digs deeper into the diversity within demographic groups.  


Demographic data can also perpetuate bias. Data is simply a quantitative representation of our world, so inequities in our society can easily be replicated. Even without using data on protected groups like race or disability status, some demographic data can serve as a proxy for this sensitive information and perpetuate inequities. As an example: zip code data can appear neutral at first, but given that systemic inequities have pushed certain groups into particular neighborhoods, that data can lead to similar results as having used race or socioeconomic status data.  


Why behavioral data is better 

Behavioral data, on the other hand, represents patterns in people’s behavior and give insight on their mindsets and decisions. Where demographics tell the “who”, behavioral data tells you “why” and “how”.  


Implementing behavioral data in your student success efforts is better than just using demographic data because it gives a more holistic view into your students. Behavioral data gives greater context into who your audience is, especially when combined with demographic data. With behavioral data, you can understand the causal factors of the decisions people are making and how those decisions affect their outcomes instead of just knowing people’s characteristics.  


Using behavioral data also allows your institution to create a better connection with students. You can understand how students are engaging with material, what motivates them to continue their education, and what is keeping them from progressing. Empowered with this information, you can provide scalable and one-to-one personalized messaging and support to help each student with exactly what they need. Much like the personalization engines driving value for many consumer facing apps (Uber, Spotify, Netflix), properly applied behavioral data fosters a better connection between students and the institution.  


Why using behavioral data is important  

Personalization, connection, and belonging are more important than ever in the education sector, and using behavioral data can help you foster that for better results. This year, just 12 percent of college students reported feeling a strong sense of belonging at their institutions, an essential component of student success. Connecting with students of all ages and levels is crucial during the post-pandemic recovery, and using the right data can help. Using behavioral data allows you to create a more personalized experience for learners at a time when students are seeking connection in ways that demographic data alone can’t.  


Using behavioral data to support students and learners allows you to meet them where they are at and better support their needs. Behavioral data offers deep insight into who a person is in the moment. Contact us to learn how we can help you offer more personalized support for your students.  

What we talk about when we talk about personalization in Higher Ed

What we talk about when we talk about personalization in Higher Ed

By Keith P. O’Brien

7 minutes estimated time to read



Part one of a two-part series

People's faces are unique like their attitudes

If you spend any time with enrollment leaders at colleges and universities, you get a sense of what the phrase “canary in a coalmine” means. Enrollment management sits at the forefront of the struggle to maintain a rapidly crumbling business model. One perceptive take on this business stated: “The entire multi-billion-dollar, 2,000-campus American college system depends overwhelmingly for its very existence on one resource: an ever-renewing supply of fee-paying undergraduates.”  

The “perfect storm” has arrived where the diminishing supply of high school grads intersects with the contracting international student pool (a mainstay of tuition revenue). In fact, the higher ed sector operates in a “buyer’s market” where supply exceeds demand for college seats and most institutions must compete harder for undergraduates.

I use the term buyers’ market deliberately because higher ed operates as a marketplace where prospects and students—both traditional and nontraditional—are the customers.  Students choose an institution and a course of study based on personal need and preferences, price and brand in return for a tuition expense. No other voluntary purchase by a person comes close to the life changing effect of choosing a college. How the university interacts with the student across the enrollment cycle (a uniquely long sales cycle) greatly influences the ultimate enrollment outcome. 

Consumers inherently expect universities to understand them as individuals and provide the personalization, convenience and service they experience daily through Spotify, Netflix, Amazon Prime, WhatsApp, and Uber. Enrollment leaders struggling to seat a class that meets academic profile, tuition and diversity targets must treat their prospects and families as consumers. To influence consumers’ decision making institutions must embrace personalization: “delivering tailored, meaningful and relevant customer communications.” Personalization is no longer optional but rather a competitive necessity for business-to-consumer firms. And, higher ed institutions that ignore this market reality do so at their peril. Institutional leadership striving to spur enrollment quickly would do better to invest in personalization rather than other common (and much imitated) strategies such as tuition resets, going test optional, or launching market- orientated degree programs (cybersecurity, data analytics or health sciences).

Personalization and Nudging

Personalization influences a target audience’s behaviors by providing the right message at the right time. True personalization is neither easy to design nor implement: every prospect and admitted student have unique academic, financial, and wellness needs. And, “Dear Jane” emails no longer clear the bar. Data, however, are the lifeblood of personalization. Universities possess unequaled demographic and behavioral data across the student lifecycle. Implementing true personalization at scale–right message, right time, right channel—requires leveraging insights about human decision making.

A new approach to personalization lies in the application of behavioral economics. This discipline’s concept of “nudging” provides new insight on how people think, make decisions, and behave. A nudge involves “changing the presentation of choices for people so they’re more likely to choose one option rather than the other” — Michael Hallsworth, Behavioral Insights Team.

Nudging is not new to higher ed. Recently two studies found large-scale nudging had marginal impact on FAFSA completion and college enrollment, respectively: Nudging at a National Scale: Experimental Evidence from FAFSA Completion Campaigns and College Board’s Realize Your College Potential Campaign for Postsecondary Enrollment. The studies’ findings, however, miss the mark somewhat by not assessing the level of personalization employed. A comment from the report on nudging FAFSA completion is illuminating: “Our results suggest a more effective path to scale may depend on using advances in data science and technology to develop more targeted, personalized nudges.” 

Reimagining Personalization Personalization is like assembling a jigsaw puzzle

Personalization has traditionally been based on demographic attributes: age, gender, location, socio-economic status, and ethnicity. Two people with exactly the same socio-economic and demographic characteristics can choose products in very different ways: one may focus on value and cost, the other on quality and durability. Traditional “personalization” approaches would message them the same way. But this approach will likely fail because the quality-conscious person won’t respond positively to a price-focused message. It is the individual’s attitude (value versus quality) towards the decision that motivates purchasing behaviors.

Human attitudes explain the “why” of human behavior. And personalization based on a person’s attitudes will generate more favorable responses than other approaches. For higher ed institutions, accordingly, the mindsets of prospects, students and alumni play a critical role in how they make decisions.

Identifying each student’s mindset and tailoring communications to individual attitudes may sound complicated, time consuming, and costly. But attitudinal personalization is viable as:

  • The necessary student data already exist on every campus in the CRM, LMS and SIS.
  • Artificial intelligence can translate existing behavioral data into a profile of the student’s attitudes. (A process that does not use any Personally Identifiable Information.)

In my next blog, I’ll show how one university used just 18 data points on admitted students to deliver attitudinal nudges that reduced melt.

Two free resources to learn more about:

  1. Personalization in higher ed, attitudinal nudging and AI’s game-changing potential: download our free eBookArtificial Intelligence and Higher Ed
  2. How attitudes, “think-alike mindsets”, and nudging drive human decision making: get our free whitepaperWhy Attitudes Drive Decision Making: Reimagining Personalization

Want to learn more or schedule a product demo? [email protected] 

Personalized Nudges. Improved Outcomes.

4929 Bethesda Ave #200, Bethesda, MD 20814


© 2022 All Rights Reserved.

Effective Personalization: Student Mindsets Matter Most

Effective Personalization: Student Mindsets Matter Most

By Keith P. O’Brien

8 minutes estimated time to read

Part two  of a two-part series

Head shots showing uniqueness of people

Summer melt remains a persistent issue for heads of enrollment at colleges and universities. Even a few admitted students failing to turn up or dropping out can undermine headcount, academic profile, diversity, and tuition revenue objectives. Admitted students choose one college over another for an array of reasons; institutions that demonstrate they “get” the student on some personal level tend to win out. Personalizing outreach and interactions builds affinity and mitigates melt.  My previous blog proposes that the most effective personalization aligns to a student’s attitudes. The following case study shows how one university engaged student attitudes to cut FAFSA verification melt..

What is Verification Melt?

“Each year, approximately one in three aid applicants gets another chore. The U.S. Department of Education requires millions of new and returning students to submit additional information to colleges, which then must verify the accuracy of each FAFSA flagged for review. Students must comply to get their money. If you haven’t been through the time-consuming procedure, then you’re probably not poor.”   — Eric Hoover, Chronicle of Higher Ed

The federal government requires every institution to verify selected student’s FAFSA information. Verification melt occurs when students fail to initiate or complete this (laborious) process and don’t qualify for Title IV grants and loans. Underrepresented and first-generation students are disproportionately affected

Case Study: Identifying Student Mindsets    

At the university, 150 students from the Fall 2019 cohort were selected for FAFSA verification. The VP for Financial Aid partnered with Discourse Analytics (DA) to help these students—the target population—complete the process and enroll.  DA used a four-step process to profile the attitudes of these students towards verification and develop nudges targeting these attitudes.)

Four steps to identify a student's attitudes

  1. Use existing student data as indicators for attitudes: for each student in the target population, DA’s artificial intelligence (AI) platform ingested 18 data fields from the CRM (Slate) and financial aid software (PowerFAIDS). These data included the number of AP classes; household debt to income ratio; and social club participation. No Personally Identifiable Information (PII) was used.
  2. Pinpoint each student’s attitudes: the AI-platform evaluated 12 non-cognitive attributes (capacity, growth mindset, proactivity, resiliency, self-awareness, self-confidence, self-discipline, social intelligence, comfort with technology, etc.) to create an attitudinal profile of every student. The profiles pinpointed a student’s attitudes that drive their behavior regarding FAFSA verification.
  3. Cluster like-minded students: DA’s “Think-alike” engine identified three clusters of students with shared mindsets. Each mindset cluster predicts how its students will behave and prescribes nudges towards the desired action. Clustering enables personalization at scale by tailoring interventions to shared mindsets. (See Exhibit for details of one mindset cluster.
  4. Nudge student mindsets: DA and Financial Aid created nudges for each mindset cluster. The nudges help students make better decisions for themselves by changing the context in which they make choices. The nudges were sent via the university’s email platform to streamline access to students and maintain confidentiality. DA’s platform uses machine learning to assess student responses and continuously refine the mindset clusters; each nudge shapes attitudes so re-calibration is essential. 


Exhibit:  “Motivated Achievers” Mindset Snapshot  
Attitudes: highly motivated, very self-confident, low self-awareness, average capacity. Limited self-awareness and high confidence often lead these students to overcommit on tasks and responsibilities.  
Expected FAFSA Verification Behaviors: unlikely to track deadlines, have tendency to delay gathering documents because self-confidence underestimates task complexity, and misjudges capacity to get projects completed. Student’s view of personal capacity doesn’t match commitments and related workload.  

Nudging Strategy: calls to action target overcommitment so capacity issues don’t undercut FAFSA verification obligations. Nudges highlight the importance of timeliness in task completion and the support resources available. Nudges also engage the student’s task-orientation and proactivity by emphasizing the benefits of a using a step-by-step approach to the process. 


Motivating Student Attitudes Boosts FAFSA Verification

DA created a randomized control group of students going through verification to benchmark the results. The target group students receiving attitudinal nudges were much more likely to complete verification: 25% increase versus the control group.  

How engaging student attitudes boosts FAFSA verification

The Personalization Imperative

Consumer habits shape a student’s approach to decision making: how to parse information sources, consider options, and make choices. This consumer orientation and its expectations around responsiveness, convenience and personalization will rarely, if ever, cease at the campus gate, or on the college website. And students’ needs are increasingly distinct as societal and demographic trends blur the distinctions between traditional and nontraditional students. Institutions serve cohorts with highly disparate characteristics and needs: using personalization has become a non-negotiable for every institution, from the most selective to open access.

Data are indispensable for personalization and institutions have ever-growing data sets for the entire student lifecycle. AI and data science grant universities the opportunity to turn student data into attitudinal profiles and deliver personalization at scale—the right mindset message on the right channel at the right time. Attitudinal personalization helps an institution rise above today’s buyers’ market by boosting student and institutional outcomes. 

Two free resources to learn more about:

  1. Personalization, mindsets, the FASA Verification project, and AI’s game-changing potential:  download Discourse Analytics’  free eBookArtificial Intelligence and Higher Ed
  2. How attitudes, “think-alike” mindsets and nudging drive human decision making: get our free whitepaper,  Why Attitudes Drive Decision Making: Reimagining Personalization 

Personalized Nudges. Improved Outcomes.

4929 Bethesda Ave #200, Bethesda, MD 20814


© 2022 All Rights Reserved.

Your campus and AI: 6 Questions to Prepare for Success

Your campus and AI: 6 Questions to Prepare for Success

By Keith P. O’Brien

8 minutes estimated time to read


Light bulb showing power of new technology                

 Why Artificial Intelligence is the New Electricity

“Just as electricity transformed almost everything 100 years ago, today I actually have a hard time thinking of an industry that I don’t think AI will transform in the next several years.”

— Andrew Ng, co-founder of Coursera and Adjunct Professor of Computer Science, Stanford University  


About a century ago, we replaced steam powered machines with those using electricity, transforming everything from transportation to healthcare.  Equating Artificial Intelligence (AI) with electricity is apt given its impact on industry and markets. I’m not trying to up the ante on AI’s importance but I was taken by the assessment of Google CEO, Sundar Pichai, “AI is one of the most important things humanity is working on. It’s more profound than electricity or fire.”  The reference to fire resonated because I’d read where a professor used fire as an analogy with AI: fire became life changing for humans not for our understanding of its chemical properties but because humans discovered remarkable ways to use fire. Leaders across all industries including higher ed have to determine how to apply AI in the most consequential ways without getting distracted by technological minutiae.  

AI presents colleges and universities with the opportunity to deliver personalized experiences across the student lifecycle. But many administrators are wary of the hype and perceived complexity of AI. Such concerns are overstated—only AI can translate the terabytes of student data on campuses into experiences that students value and benefit from. The time to act is now; the rewards of adopting AI far outweigh the risks from procrastination given today’s enrollment and tuition revenue challenges. So how does a college get started? 

Where and How Should We Use AI? Six Questions to Get Started

AI can do many things on campus from automating routine administrative tasks to identifying the mindsets of students that drive their behaviors. Use the following six questions to shape how AI is deployed strategically and operationally on campus:

  1. What are we trying to do here?  The starting point for any AI project is pinpointing the business problem the technology can address. You must consider the university’s strategic priorities, the nature and urgency of the problems, and the organizational and cultural implications. Problem identification and prioritization are critical because deployment of AI solutions often raises unrealistic expectations. TIP: Campus leaders can start by addressing these five basic decisions: forge consensus on the problem, identify the project owner, pinpoint the operations and campus units  involved, clarify the role of the IT department upfront, and define the performance metrics. 
  2. Do we have leadership buy-in?  The AI project needs sponsors who clearly and publicly commit to the initiative. The project sponsor(s) must have the authority to allocate resources to the project and the backing of campus leaders to drive implementation across campus units. TIP: Communicate regularly to foster support and dispel misinformation. At a minimum, communicate the project’s strategic importance, operational impact (faster, smarter, cheaper), and benefits for students and the institution.  
  3. What’s the timeframe for expected returns?  Some administrators expect substantial returns quickly given the aura around AI. The timeline and ROI are a function of the problem being tackled first and foremost. For example, deploying AI to help more students complete FAFSA verification targets a well-defined issue, a set timeframe and a clear performance metric. The campus units involved and the performance metrics are critical factors too.  TIP: Groundbreaking technologies don’t fit neatly with existing ways of operating. Remember the advent of the Internet? At first it was very disruptive but now is a competitive necessity. Don’t let false precision over timing or outcomes let your institution get blindsided: early adopters of AI gain a market advantage while skeptics must play catch up. 
  4. Have we addressed data readiness and governance issues? AI generates vast new troves of information. Data readiness means that data can be connected across systems and campus units throughout the institution. Matching data using student id represents one such opportunity. Admissions systems don’t employ a student id and so alternate match keys are required. Policy readiness requires that data governance, identity resolution, privacy and other policies authorize the university to use specific data and systems for specific purposes, including informing student communications. 
  5. How receptive is the Institution to change?  Every AI project operates in the context of an institution’s unique culture, governance models, and campus unit processes. AI project sponsors must gauge the institution’s cultural receptivity to change, and plan accordingly. TIP: Here’s two criteria help assess campus receptivity:
    • Has a data-driven culture—the ability to test new data driven initiatives in a controlled environment indicates organizational readiness for AI. Taking a data-driven approach helps test, validate and implement process improvements. Data-driven Institutions tend to generate greater returns from AI.
    • Embraces continuous improvement—a track record of embracing change is a key marker of process readiness. Process driven self-evaluation indicates leadership has fostered a “growth-mindset” campus climate that embraces new technologies and decision-making approaches.
  1. How do we educate the campus on AI?  While all technology projects and integrations pose challenges on campus, AI-led  projects can generate unease among some staff: a common misconception is that AI serves to automate work and erode job security.  However, as one university CTO observed about the reality of AI: “Nothing will ever replace the human touch aspect of an educational professional interacting directly with a student.”  Higher ed has many federal- and state-mandated reporting processes that can be automated by AI, freeing up campus staff to focus on the more value-adding aspects of their roles. Project leaders must communicate proactively to set and manage expectations by sharing the goals, expected benefits, and the exact role the AI performs across campus. 

AI can streamline red tape, reduce or eliminate data siloes, and improve staff capacity to engage students one-on-one. It turns disparate student data sets into unparalleled insight on individual behaviors and needs. Just as it’s impossible to imagine a campus without electricity,  AI has to be part of how colleges and universities engage students. In my next blog, I’ll look at how institutions can approach working with vendors on AI-led initiatives. 

This blog is based on our free eBook about AI’s game-changing potential, Artificial Intelligence and Higher Ed: Cutting Through the Noise

To learn more contact: [email protected]


Personalized Nudges. Improved Outcomes.

4929 Bethesda Ave #200, Bethesda, MD 20814


© 2022 All Rights Reserved.