Linking survey scores to student comments

Updated Nov 06, 2026

A survey score can identify a group worth listening to more closely. Its comments can show what that group wants to discuss. Annalina Sarra and colleagues bring these perspectives together in a study of student experience at an Italian university, linking questionnaire ratings with the language students use about strengths and improvements.

For institutional research and Student Experience teams, the useful question is specific: what do students with different satisfaction profiles praise, and what do they want to change?

How the study connects ratings and free text

The researchers analysed an online survey conducted during the 2023/24 academic year. Students rated aspects of their degree programme and wider university experience, then answered separate open questions about what they appreciated and what could improve.

The team used structural equation modelling to estimate satisfaction dimensions from the questionnaire items. They then explored comment themes with a Structural Topic Model and a guided topic model, SeededLDA. For the guided model, distinctive words from groups with lower, middle and higher estimated satisfaction supplied the starting vocabulary. An additional unguided topic captured other content.

What the comment analysis revealed

Praise and improvement requests brought different aspects of university life into view. In the university-experience comments, positive language emphasised friendships, relationships, opportunities and personal growth. Suggestions for improvement concentrated on exams, administration, classrooms and organisation. Reading these as separate kinds of feedback helped explain what students valued and where they encountered friction.

A shared topic could contain different messages. Within the topic concerning the university environment and academic resources, positive comments emphasised teaching and staff competence, while improvement comments focused on classrooms and practical activities. For analysts, a topic label is therefore a starting point for reading the account behind it. Our student feedback analysis glossary explains the distinction between categorisation and sentiment.

More satisfied students still described changes they wanted. The guided analysis associated their improvement comments with spaces, classrooms, cafeterias, schedules and materials. Lower-satisfaction comments more often centred on teaching and learning processes. A report that reads only the lowest scores would leave useful suggestions from satisfied students outside the discussion.

Social experience remained visible alongside operational concerns. The positive-comment topics included relationships with peers and lecturers, campus resources and personal development. The analysis gave these strengths space alongside requests about administration and timetabling, offering a more detailed account of the university experience than an overall score alone.

A practical approach for university survey teams

Our recommendation is to choose a focused reporting question before combining datasets. For example, ask which aspects of course organisation appear in comments from students with different responses to an organisation question. Decide how the rating groups will be defined, then retain the original question and collection period with each comment set.

Give each group a short account of both strengths and requested changes. Separate the topic from what students are saying about it, and use anonymised examples to help the responsible team understand the issue. A timetable comment about room changes calls for a different response from one about clashes with placements.

Agree who will review each finding and what decision it should inform. Our comment analysis governance checklist provides prompts for access, review responsibilities and traceability. The practical aim is a report in which readers can move from a survey pattern to a clearly described experience and a named next step.

FAQ

Q: What does linking scores and comments add?

A: It lets a survey team examine what students with different rating profiles discuss. In this study, that brought relational strengths and operational improvement requests into the same analysis.

Q: How were the topics guided?

A: The researchers used distinctive vocabulary from groups defined by estimated satisfaction to guide SeededLDA. A separate unguided topic allowed other themes to emerge.

Q: Where could an institutional team start?

A: Select one reporting question, agree the rating groups, and review strengths and improvement requests within each. Assign the resulting issues to teams able to investigate and respond.

References

Sarra, A., Cucco, A., Fontanella, L., Di Battista, T. and Evangelista, A. (2026). Semi-supervised evaluation of University quality: from structured measures to student voices. Quality & Quantity, 60, 18525–18549. Published online 31 July 2026. Read the open-access paper.