United StatesEducation2016 – ongoingGeorgia State University
Pounce admissions text chatbot
Georgia State's admissions chatbot cut 'summer melt' by about a fifth in a randomized trial
Problem
At Georgia State, the share of admitted students who committed to enroll but never arrived in the fall, known as summer melt, had risen from about 12% to nearly 19%. Email, letters and phone calls had not been enough to guide students through financial aid, documents and other enrollment steps.
Approach
In spring 2016 the university introduced Pounce, a text-message chatbot built by the startup AdmitHub (later Mainstay), which sent reminders, collected survey answers and answered questions around the clock. Researchers Lindsay Page and Hunter Gehlbach ran it as a randomized controlled trial: of more than 7,000 admitted students with US mobile numbers, 3,114 got Pounce and the rest received the usual email and mail.
Results
Takeaway
Staff were needed for under 1% of more than 50,000 student messages, and a randomized trial measured the chatbot's effect on enrollment. Mainstay reported a 3.3% enrollment gain, while Campus Technology reported 3.9%.
Categories
- Size
- Large enterprise
- Industry
- Education
- Business function
- Customer serviceMarketing and sales
- Innovation type
- Process innovationProduct and service innovation
- AI technology
- Conversational and voice AI
- Adoption
- Vendor solution
- Evidence
- Independent study
Sources
Source summaries
Campus Technology describes the trial overseen by Page and Gehlbach, with 3,114 students in the treatment group out of more than 7,000. It reports nearly 200,000 messages exchanged, under 1% needing staff, and the university's estimate that it would otherwise have needed 10 full-time staff. It reports 21.4% lower summer melt and 3.9% higher enrollment in the treatment group.
Mainstay reports that Pounce launched in April 2016 and that results were measured among students who committed to enroll by June 1. It reports a 21.4% reduction in summer melt and a 3.3% increase in enrollment among those students. It adds that 90% of eligible students opted in and that first-generation and Pell-eligible students messaged the bot more than average.
Full details are in the original sources.
