Every course on EduBracket is enrolled in, completed, and scored on the same six dimensions. The metrics are weighted equally on a 10-point scale, and we publish what failed alongside what won. This page is the full protocol.
We call this protocol The EduBracket Course-ROI Audit: six equally-weighted metrics, every course enrolled in and completed, scored against real outcome data in our Online Learning Statistics study, with returns compared in the Course-ROI Showdown. Reviewed by Vincent Wesley Couey (ORCID). Our peer-archived, DOI-backed dataset: 2026 Online Course ROI Audit (DOI 10.5281/zenodo.20632758).
The reviewer enrolls in the course using the same paid account a typical learner would use. They complete every lesson, from intro module to capstone, hand in every assignment, take every proctored exam, and submit every required portfolio piece. We do not score from syllabus pages.
Why first. A course that's scored on its claimed outcomes but not its actual delivery is a marketing rewrite, not a review. Completion is the load-bearing metric, every other metric assumes you finished.
We audit the actual lesson assets: audio mix, video resolution, working code samples, current libraries, dead links, abandoned modules, and dates on cited research. We compare the production date against the platform's claim of "updated for 2026" and downgrade any course where the gap is more than six months.
We log actual hours from enrollment to portfolio-ready outcome and compare against the platform's claimed hours. Most platforms understate completion time by 30-60%. We publish the gap on every review.
For certifications, time-to-skill includes practice-exam hours and credential-issue lag.
Total cost ÷ actual hours. We factor in subscription pricing, discounts available at the time of review, free-trial windows, and the cost of any required supplementary materials. We do not use sticker price.
This metric is where free or commission-free courses (CS50, MIT OCW, freeCodeCamp) frequently outscore paid courses, they show up on real career-change brackets exactly because cost-per-hour is binding.
Each quarter we survey 200+ hiring managers across the relevant role categories. We ask which credentials they recognize, which they prefer, which they discount, and which they treat as a negative signal (yes, some certifications are net-negative on a résumé). The data lives in every certification review.
For non-credential courses we run a portfolio-recognition pass: do recruiters identify the project outcomes as portfolio-grade, or do they read as "course exercise"?
We file an actual refund request on every platform that claims a money-back guarantee. We measure days-to-resolution, manual hoops (calls, forms, chats), and chargeback rate for cases where the platform stonewalls. This is the most under-reported metric in course reviews, and the one buyers most care about when a course turns out to be wrong for them.
Every published bracket cites its inputs. We anchor our claims to public sources wherever possible and disclose the proprietary data we collect ourselves (the hiring-manager survey, the refund-test results, the time-to-skill logs).
"The losing entries are the proof we ran the tournament honestly."