The development of present-day assessment culture in higher education has led to a disciplinary turn away from statistical definitions of reliability and validity in favor of methods argued to have more potential for positive curricular change. Such interest in redefining reliability and validity also may be inspired by the unsustainable demands that large-scale quantitative assessment would place on composition programs. In response to this dilemma, we tested a mixed-methods approach to writing assessment that combined large-scale quantitative assessment using thin-slice methods with targeted, smaller-scale qualitative assessment of selected student writing using rich features analysis. We suggest that such an approach will allow composition programs to (a) directly assess a representative sample of student writing with excellent reliability, (b) significantly reduce total assessment time, and (c) preserve the autonomy and contextualized quality of assessment sought in current definitions of validity.