SQAM

Database courses often reduce to writing SELECT queries, graded all-or-nothing by string matching. Students who are nearly right get nothing and learn little from the mark, and they rarely practise defining schemas or managing data, the skills industry says it needs most.

SQAM, the SQL AutoMarker, grades a query by what it returns rather than how it is written, awarding partial credit and personalized feedback. Across five course offerings its marks matched TA marks wherever TAs gave partial credit, and averaged 14.5 points above a binary automarker.

Weekly scaffolded schema and data exercises with immediate feedback now make data definition and manipulation a core, assessed part of a third-year database course. Across three offerings (669 students, 9,071 submissions), students needed fewer attempts on schema and data tasks than on queries, and early schema performance predicted later SQL performance.