Practice notes from the field.
Shorter and more opinionated than the research papers. What we have learned deploying analytics inside real institutions, including the parts that went badly.
Everything we’ve published.
Closing the equity gap: what we got wrong in our own early-warning model
We removed protected attributes from the feature set and assumed that made the model fair. It did not. The problem was a feature that looked…
Read →
Integrate assessment timing first. Everything else can wait.
We ran a 36-cell grid on model families against data completeness. One data source was worth more than every model upgrade combined, and it is…
Read →
Half of your student success initiatives are probably doing nothing
We assessed 220 initiatives with matched comparison. 47% showed a measurable effect. Here is what separated them, and it was not…
Read →
Write the data dictionary before you buy the analytics
The most common reason an analytics project fails has nothing to do with software. Your registrar and your IR office do not agree what retention…
Read →
Why we will not sell you one model for every institution
It would be a much better business. We tested whether it works, across fourteen institutional contexts, and the answer was clear enough to…
Read →
Students are not telling you about the timetable, and it is ending their degrees
We compared what 340,000 feedback responses complained about with what actually predicted outcomes. The correlation was 0.21. The gap is where…
Read →New writing, roughly once a month.
Practice notes, method notes, and the occasional argument. No product announcements dressed up as insight.