Education analytics: enrollment, retention and student success metrics
Enrollment funnels, yield and melt, fall-to-fall retention, completion, DFW rates and early warning signs, each with a formula and round-number example, plus how to handle student data responsibly and what a BI tool can and cannot do here.
Colleges, online course businesses and training companies all face the same three questions: are enough of the right students enrolling, are they staying, and are they succeeding? Each has a small set of metrics with standard definitions. This post gives the formulas, works through each one with round numbers, and ends with the privacy rules that apply to student data and what a BI tool like clariBI can and cannot do in this area.
Enrollment: the funnel from inquiry to first day
Most institutions and programs have the same stages, even if they use different names:
- Inquiry: someone requests information, attends an open day or signs up for a webinar.
- Application: a completed application.
- Admission: an offer is made.
- Deposit or acceptance: the student commits, often with a payment.
- Enrollment: the student is still enrolled after the census date (or, for a course business, has paid and started).
Worked example for one intake:
| Stage | Count | Conversion from previous stage |
|---|---|---|
| Inquiries | 5,000 | |
| Applications | 1,500 | 30% |
| Admissions | 900 | 60% |
| Deposits | 360 | 40% |
| Enrolled after census | 324 | 90% |
Three metrics come straight out of this table:
- Yield rate = enrolled / admitted = 324 / 900 = 36%. Falling yield usually means admitted students choose someone else.
- Melt rate = (deposited - enrolled) / deposited = 36 / 360 = 10%. Summer melt tends to be highest among first-generation and lower-income students, so it is worth tracking by group.
- Cost per enrolled student = recruitment and marketing spend / enrolled. With $180,000 spent: 180,000 / 324 = about $556. Split it by channel (search ads, social ads, events, referrals) to see which channels bring students who actually enroll, not just inquiries.
The same table works as a forecast. If this year's inquiries are 5,500 and the stage rates hold, expect 5,500 x 0.30 x 0.60 x 0.40 x 0.90 = about 356 enrollments. Check the forecast against actuals each intake and update the rates.
For an online course business the stages are shorter (visit, sign-up or free lesson, purchase), and the funnel analysis guide applies almost unchanged.
Retention: keeping the students you enrolled
Retention is usually a bigger lever than recruitment: a student who leaves after one term is lost tuition and, often, a student who was poorly served.
- Fall-to-fall retention = first-year students who return the next fall / first-year students who started. With 400 starting and 300 returning: 75%.
- Term-to-term persistence: the same calculation from one term to the next. It catches students who leave mid-year, which annual retention hides.
- Completion rate = students who complete within a set time / students who started. In US higher education the usual window is 150% of normal program length, so six years for a four-year degree. If 220 of those 400 finish within six years: 55%.
- Stop-out versus drop-out: a stop-out leaves and may return; a drop-out is gone. Tracking them separately tells you who a re-enrollment campaign should target.
For national comparisons, use a published source such as the National Student Clearinghouse's annual persistence and retention reports, and compare like with like (institution type, full-time versus part-time).
The best way to look at retention is by cohort: one row per starting term, one column per term since start. It shows whether newer cohorts are doing better than older ones. The cohort analysis guide explains how to build and read one.
Early warning signs
The point of retention analytics is to reach students before they leave. Commonly used leading signs:
- Academic: low midterm grades or a first-term GPA below a set level. Midterms give earlier warning than final grades.
- Engagement: fewer logins to the learning platform, late or missing assignments, missed classes.
- Financial: unpaid balances, holds on the account, gaps in financial aid.
- Registration: not registered for next term by the priority deadline, or frequent schedule changes.
A simple rule ("two or more of these signs triggers an advisor call") is better than waiting for a student to fail to re-register. If you later build a statistical risk score, check it for bias: a model trained on past outcomes can flag students from groups the institution served poorly, when the real problem was support rather than the students.
Student success: are they learning and progressing
- Credit accumulation = credits earned / credits expected at this point. A student with 21 of 30 expected credits after year one is at 70% and unlikely to finish on time.
- DFW rate = students receiving D, F or Withdraw in a course / students enrolled in it. In a course of 120 with 18 Ds, 12 Fs and 6 withdrawals: 36 / 120 = 30%. High DFW rates identify gateway courses that hold many students back.
- Time to completion: median and average time from start to credential.
- Outcomes after completion: employment rate in a related field within 6 to 12 months, certification or licensure pass rates, further study.
For course and training businesses, the equivalents are course completion rate, refund rate, repeat purchase rate ("how many students bought a second course within six months?") and pass rates on any exam the course prepares for.
Privacy and responsible use
Student data is sensitive and regulated.
- FERPA governs education records at US institutions that receive federal funding. GDPR applies to personal data of people in the EU, students included.
- Share only what each role needs. An advisor needs their own students; a dean needs aggregates for the college.
- Prefer aggregated data for analysis: counts and rates by program, term and group rather than named individuals.
- Tell students how their data is used, and use analytics to offer support, not to sort or penalize.
What clariBI can and cannot do here
Be clear about the fit before you connect anything.
What it does not do: clariBI has no connector for student information systems or learning platforms (Canvas, Moodle, Blackboard, PowerSchool and similar). It is not meant to hold identifiable student records or grades, and it has no FERPA-specific arrangement. For an institution's student-level data, use your SIS reporting or institutional research tools.
Where it fits: course, training and tutoring businesses, and the enrollment and marketing side of larger programs. As described on the education use case page, you can connect:
- payments from Stripe, PayPal or Square;
- marketing from Google Ads, Meta Ads, Google Analytics 4 and Search Console;
- email tools such as Mailchimp, Klaviyo or ActiveCampaign;
- aggregated exports (enrollments by course and term, completion counts, capacities) as CSV or Excel, as described in the file upload guide.
With those connected, questions like "How many enrollments did each campaign bring in this term, and at what cost?" or "Which courses have the highest refund rate?" get answered in plain English on the Trial and on Starter and above. On Starter and above, forecasting can project enrollments or revenue per term from your history, with ranges that show how uncertain the projection is.
A next step: export last year's enrollments by course and term, with no student names or IDs, connect it together with your ad accounts, and ask to see ad spend and enrollments by month side by side. Divide one by the other for cost per enrollment. If it differs a lot from what your team assumes, you have found the first number worth watching.
