How our algorithm works
StepUp is an explainable recommendation system, not a black box. This page describes exactly what we collect, how a recommendation is produced, and how the model improves over time.
What data we use
Four input families feed the model. Students provide all of them directly, and any input left blank simply lowers confidence rather than being guessed.
Interests
Responses to the 48-item Holland RIASEC inventory, scored across Realistic, Investigative, Artistic, Social, Enterprising and Conventional dimensions.
Academics
Course history, current GPA, grade level and prerequisite completion. Grades are used to measure readiness, never to gate a student out of a pathway.
Strengths and self-efficacy
Confidence ratings and self-reported strengths that indicate where a student already performs well and where support would help.
Context and goals
Stated career interests, time remaining in school, and program preferences such as college or university preparation.
From answers to recommendation
Collect
Students complete the RIASEC inventory and add grades and courses. Progress autosaves after every item, and nothing is inferred from data a student has not provided.
Normalize
Answers are averaged per dimension and scaled to 0 to 100. Grades are converted to a common GPA band so schools with different scales are comparable.
Score
Every career in the catalogue receives Fit, Readiness and Feasibility scores. They combine into one deterministic match score, weighted by Confidence.
Explain
Each recommendation ships with the reasons behind it: which interests drove the match, which prerequisites are missing, and what the next concrete step is.
The scoring model
A match score is a weighted blend of Fit, Readiness and Feasibility. Confidence weights the whole result and is always shown next to it.
Fit
40%How closely a career's Holland profile matches the student's RIASEC results. Each career carries an ordered affinity, and the student's dimension scores are weighted by position (first affinity counts most).
Fit = Σ (dimension score × position weight) ÷ maximum possible
Readiness
30%Whether current coursework and grades support the pathway today. Completed prerequisites add credit, missing ones subtract it, and a recent upward trend is rewarded.
Readiness = GPA band match + prerequisite coverage ± recent trend
Feasibility
20%How realistic the path is given the terms a student has left, pathway difficulty, admissions requirements and program cost. Harder paths rank lower, they are never hidden.
Feasibility = terms remaining vs. terms required, adjusted for difficulty
Confidence
Reliability flagHow much evidence sits behind the result: completeness of the inventory, how clearly the top dimensions separate, and response variety so straight-lining lowers confidence.
Confidence = completeness × differentiation × engagement
The model gets more accurate over time
The deterministic core stays fixed so results remain reproducible. On top of it, StepUp learns from aggregated, de-identified usage: which recommended pathways students actually pursue, which milestones get completed, which prerequisite gaps predict later difficulty, and how counselors adjust plans. Those patterns tune the weights and the career catalogue across cohorts, so recommendations sharpen year over year without changing how any single score is explained.
Learning happens on aggregated cohort data. Individual student records are never sold, never used to train third-party models, and remain owned by the school.
Our commitments
Deterministic first
The same inputs always produce the same ranking. A counselor can reproduce and audit any result a student sees.
Explainable, not opaque
Every score traces back to specific items, courses and weights. No recommendation appears without a reason attached.
Advisory, never prescriptive
Scores rank options and surface gaps. They do not lock a student out of a pathway or replace counselor judgement.
Research-backed
The interest model is grounded in Holland's RIASEC framework, a widely validated vocational theory used in school counseling for decades.
Want the score-by-score breakdown? See how StepUp scores careers.
Research & Sources
The frameworks, statistics and counseling research behind StepUp’s recommendation model and product claims.
- 01
National Center for Education Statistics. Public High School Students' Use of Graduation, Career, or Education Plans.
- 02
American School Counselor Association. School Counselor Roles & Ratios (2024-2025).
- 03
Poynton, T. A., Lapan, R. T., & Schuyler, S. W. Reducing Inequality in High School Students' College Knowledge: The Role of School Counselors. Professional School Counseling, 2021.
- 04
Paolini, A. C. School Counselors Promoting College and Career Readiness for High School Students. Journal of School Counseling, 2019.
- 05
Holland, J. L. Making Vocational Choices: A Theory of Vocational Personalities and Work Environments. Psychological Assessment Resources, 1997.
- 06
Bandura, A. Self-Efficacy: The Exercise of Control. W. H. Freeman, 1997.
StepUp is built using research-backed career development frameworks and transparent recommendation logic. Citations are provided for institutional review.