Know which actions will improve student outcomes, and by how much
Start with one decision, using data you already have. We show it first on synthetic data shaped like yours.
Universities we work with
Prediction shows risk. Causation shows the way forward.
Most platforms tell you who is at risk. CML+ estimates which action will actually change the outcome, and by how much.
Who is at risk?
- A risk score for each student
- Little guidance on what to do next
- Can't tell which programs actually worked
What will help, and did it?
- The expected effect of each action, for each group of students
- Every recommendation shows its evidence and uncertainty
- Measured impact, so budgets follow what works
How it compares with what you already have
CML+ builds on your dashboards and analysts rather than replacing them, starting from your SIS, LMS and CRM data.
| Compared on | BI dashboards | In-house analysts | Consultants | CML+ |
|---|---|---|---|---|
| Answers | What happened | The questions they have time for | One question, once | Which action changes the outcome, for whom, by how much |
| Keeps learning | No | When staffed | Ends with the report | Each term's outcomes update the evidence |
| Who can use it | Analysts and leaders | Analysts | Sponsors | Leaders, advisors, faculty and IR/IE, each in their own terms |
| Works with | The reports you already build | The data they can reach | Extracts for one engagement | Your SIS, LMS and CRM data |
BI dashboards
- Answers
- What happened
- Keeps learning
- No
- Who can use it
- Analysts and leaders
- Works with
- The reports you already build
In-house analysts
- Answers
- The questions they have time for
- Keeps learning
- When staffed
- Who can use it
- Analysts
- Works with
- The data they can reach
Consultants
- Answers
- One question, once
- Keeps learning
- Ends with the report
- Who can use it
- Sponsors
- Works with
- Extracts for one engagement
CML+
- Answers
- Which action changes the outcome, for whom, by how much
- Keeps learning
- Each term's outcomes update the evidence
- Who can use it
- Leaders, advisors, faculty and IR/IE, each in their own terms
- Works with
- Your SIS, LMS and CRM data
Higher Education Products
Most student success tools stop at a risk score. Every Dial recommends the action most likely to help, measures what it actually changed, and learns what works for each student.
for Instructors
Course Dial
Act before the first failing grade, then see what each action actually changed.
- Early alerts with likely causes, not just a risk score
- The action with the largest expected effect for this student
- Every action tracked against the student's expected path without it
for AdvisorsNew
Advising Dial
Learns what works for each student, not just on average, through advisor-approved n-of-1 trials.
- Ranked interventions with expected effect and uncertainty
- N-of-1 trials compare options for one student over time
- Talking points for each conversation; every result updates the model
for Learners
Career Dial
Shows which choices change a student's path over the next 3–5 years: salary, debt, credit and wealth.
- Side-by-side pathway projections, with uncertainty
- The levers that change the outcome, such as soft skills workshops, internships and financial literacy
- Post-graduate outcomes, not just job titles
for Administrative Decision-Makers
Outcome Dial
Fund programs by their causal impact on student success per dollar, at every level.
- Causal impact against cost, by program and intervention
- Evidence for accreditation and program review
- Drill down from institution to college to program
Causal instrumentation, built into every Dial
Flag early
Spot the student or program drifting off track, and why.
Recommend
Rank actions by their expected effect for this student.
Instrument
Log each action as a treatment and track the outcome against the expected path without it.
Learn
N-of-1 and cohort results update the model every term.
Screens are illustrative.
What each role gets
Provost, VP student success, CFO
Which programs and supports to fund, and which to stop. Impact estimates you can take to the board and accreditors. Evidence from a pilot before any larger commitment.
Advisors, faculty, IR/IE analysts
Who to reach, why, and what is most likely to help. Talking points that explain each recommendation. Causal reasoning skills built through daily use. Access to app analytics data for research and grants (IR/IE).
CIO, CISO, data governance
Demonstrated first on synthetic data shaped like your institution. Works from SIS, LMS and CRM extracts you control. Recommendations support staff; no automated decisions about students.
How we handle securityWhere these applications are working
Each item is CML Insight's own work with the named partner. Status is shown as it stands today.
UT Arlington
Identified program sweet spots and quantified social-mobility impact using time-series data.
Challenge
No unified data to assess program effectiveness and social-mobility outcomes.
Approach
Institutional data linked with Equifax post-graduate employment data in the CML Insight SaaS app.
Result
Identified program sweet spots and quantified social-mobility impact using time-series data.
National University
Six lifecycle models, from lead to persistence, on scalable Kubernetes MLOps.
Challenge
An end-to-end modeling platform for a complex term structure.
Approach
CRM, SIS and LMS data with relative and absolute engagement features; MLOps on Kubernetes with Kubeflow orchestration.
Result
Excellent, equitable model performance on a scalable platform.
Colorado State University
Running a randomized controlled trial that uses prediction scores and LMS behaviour trends.
Challenge
Build an n-of-1 causal learning platform with evidence-based intervention design.
Approach
SIS and LMS data integrated through the Unizin Data Platform, predictive and prescriptive models, and causal analytics agents.
Result
Running a randomized controlled trial that uses prediction scores and LMS behaviour trends.
Southern New Hampshire University
Agentic flexible learning pathways with n-of-1 causal learning, now in beta testing.
Challenge
Create a flexible, agent-driven learning environment that learns what works for each learner.
Approach
Flexible learning pathways with n-of-1 causal learning, used by LLM agents working with learners.
Status
In beta testing.
Bring one decision you want to get right.
Book a working sessionStart with a pilot, not a transformation
A low-lift pilot on one program or cohort, using data you already have: evidence before any larger commitment.
Pick one decision
One program, cohort or intervention you need to get right, such as gateway-course completion or advising outreach.
Use the data you have
SIS, LMS and CRM extracts. We first show your views on synthetic data shaped like your institution.
Decide on evidence
See which actions worked, for whom and by how much, then decide whether to scale.
Find the actions that will move your outcomes
Tell us the decision you're working on. We'll reply within one business day and set up a 45-minute working session.
Book a working sessionWhat happens next
- We reply within one business day.
- A 45-minute working session works through your decision with you.
- We show a prototype on synthetic data shaped like yours.