Blending Database Systems to Prepare for Predictive Modeling - - PowerPoint PPT Presentation
Blending Database Systems to Prepare for Predictive Modeling Kristen Salomonson Dean of Enrollment Services Ferris State University Jon MacMillan Senior Data Analyst Rapid Insight June 11 th , 2019 Agenda About Rapid Insight The Data
Kristen Salomonson Dean of Enrollment Services Ferris State University Jon MacMillan Senior Data Analyst Rapid Insight June 11th, 2019
Founded in 2002 and headquartered in Conway, NH Predictive analytics and data preparation software company empowering professionals of all skill levels to turn raw data into actionable insights Serving thousands of users worldwide, ranging from healthcare to higher education The Veera platform enables users to easily build predictive models, perform advanced data analysis, and share insights Code free (but code friendly) self-service analytics platform
Veera Construct enables everyone from citizen data scientists to PhD statisticians to turn any data into actionable information Veera Predict enables users of any skill level to analyze data and build predictive models in a fraction of time required by other tools Veera Bridge empowers organizations by democratizing data with its cloud-based collaboration platform
Data Sources Data Destinations Data Preparation Predictive Modeling
Rapid Insight User Conference - June 23-25th
North Conway, NH
An opportunity for users, old and new, to learn about the benefits and pitfalls of bringing real data analysis to their organization More information at: rapidinsight.com/riconf/
Percent Change High School Graduates, 2013-2032
Source: WICHE
Hard Demographic Truths
In 2013 From 2013 to 2025 From 2025 to 2032 CA 456,000
TX 314,000 61,000 19%
NY 212,000 2,000 1%
FL 176,000 17,000 10%
IL 153,000
PA 146,000
OH 135,000
MI 111,000
NJ 109,000
NC 101,000 9,000 9%
Source: WICHE
Multiple systems generating data about a greater number of recruitment and retention activities than ever before. Many existing analytic tools use only a small fraction of available data for predictions with wider error margins. Limits on in-house staff with the time and knowledge to do the work.
The Veera platform enables us to create automatic linkages with our multiple data sources. It’s flexible so we can add and remove what we include. The data delineation process is surprisingly enjoyable and extremely useful.
Initial Apply to Enroll Model
from our CRM.
the first model. *Greater # of people brought to event – higher likelihood of enrolling
New Data Source & Instant Connection
Yada this year for roommate match.
connected so we could integrate the data into our Apply to Enroll model.
Rapid Insight tools help us to take advantage of our multiple data sources and use them to create the best possible models to enhance predictive utility. Our Apply to Enroll model analysis included
every last drop of predictive utility. An earlier off-the-shelf model gave us 6. The results enable us to target individual applicants, segments, colleges and institution-wide performance.
Model Output
Rapid Insight predictive model developed
Likelihood to Enroll Score Generated
Daily Scores loaded into Salesforce for Recruitment Staff
Tailored Action Paths Deployed Select applicants targeted with robust
enrollment probability Impact on FTIAC Enrollment
Fall 17: +24 (1.5%) Fall 18: +70 (4%)
Not Just Higher Student Counts
Individual – A sought-after student in one of your top programs hasn’t signed up for Orientation. Q: What is the likelihood that Applicant K will enroll in the Fall? Action: Score in range where a personal call is warranted. Group – A recruiter wonders how to make time management decisions to optimize their yield. Q: How many Applicants in my territory are at least 70% likely to enroll? Action: Reaches out via text to these students and asks if there are follow-up questions. Institution – A budget director asks for assistance in estimating tuition revenue next year. Q: What is the predicted enrollment of new and returning students? Action: Fuse results from Admissions and Retention Models to develop an estimate. 2018/19 estimate was off by .2%.
Jon MacMillan
Rapid Insight Senior Data Analyst Jon.MacMillan@rapidinsight.com
Kristen Salomonson
Ferris State University Dean of Enrollment Services KristenSalomonson@ferris.edu