SELECTED WORK
Case studies where measurement changed the conversation
Client names anonymised where confidentiality agreements apply. Each summary reflects the measurement problem, our learning-science approach, and what shifted afterward. Past metrics are illustrative — not guarantees for your institution.
POST-SECONDARY · BC · AT-RISK SIGNAL
Vancouver school district — at-risk early signal
Grade eleven math attrition followed a pattern visible in LMS clickstreams by week three — shallow quiz attempts, widening spacing intervals, declining confidence ratings on low-stakes checks. Faculty saw nothing until the withdrawal list landed. We delivered an at-risk early signal view with educator sign-off gates: counsellors contacted students before midterm, tutorial hours reallocated to courses with mastery gaps rather than login spikes.
Illustrative past result: 14% fewer late withdrawals in the pilot cohort. Not a guarantee for your district — outcomes depend on counsellor follow-through, class size and data quality. Engagement value: C$62,000 over twelve weeks including knowledge tracing calibration and responsible-AI documentation.
HIGHER ED · NATIONAL · OUTCOMES INSIGHT
Canadian university — outcomes insight report
Faculty senate wanted retention analysis reads tied to programme interventions, not vanity metrics. Dashboards celebrated rising LMS logins while licensure exam pass rates stagnated. We traced the disconnect: the event layer conflated access with practice quality. We redesigned telemetry to capture attempt sequences, time-between-retrieval intervals and instructor feedback latency.
Within one term, department heads identified courses where students clicked often but practiced shallowly. The outcomes insight report linked cohort performance trends to three documented interventions — supplemental tutorials, revised low-stakes checkpoints, peer study group prompts — with prediction confidence ranges on each. First data cut delivered in 4.2 weeks; full reporting integration completed at C$78,000.
EDTECH · PILOT · ASSESSMENT SUPPORT
AI writing assistant — structured pilot evaluation
A literacy platform wanted evidence its AI drafting tool improved revision quality without encouraging uncritical acceptance of generated text. We co-designed a staggered rollout: control sections used the legacy editor; treatment sections received scaffolded prompts requiring students to annotate AI suggestions before submission. Primary outcomes included rubric-scored revision depth and self-explanation items; secondary telemetry tracked edit distance and time-on-task.
The study gave product leadership defensible evidence for procurement conversations — and identified cohorts where the tool increased surface-level copying. Automated feedback remained educator-gated throughout. Pilot design and analysis: C$44,000.
CONTINUING ED · INSTRUMENT DESIGN
Micro-credential confidence calibration
A professional development provider issued digital badges for project management modules but lacked evidence learners could apply frameworks on the job. We built brief scenario checks at enrolment, midpoint and thirty days post-completion, paired with confidence ratings. Calibration gaps — high confidence with low performance — became the leading indicator for content revision.
Marketing could still promote completion rates, but internal teams prioritised modules with demonstrable transfer signals. Instrument design sprint plus learner modelling profiles: C$36,000.
CLIENT VOICE
What partners say about working with us
"Smart Edu AI did not hand us a dashboard and disappear. They sat with our faculty, translated pedagogy into events, and documented every assumption so we could defend metrics to our board." — Director of Teaching & Learning, anonymised BC institution
We measure our own success by whether clients still use the frameworks six months after delivery — not by project count alone. Every engagement ends with educator sign-off on published views and a responsible-AI assurance packet your privacy officer can file.
Have a similar challenge?
Share your context — anonymised descriptions welcome — and we will outline how an analytics review could apply. Typical entry point: C$14,000.
Request an analytics review