It''s Not Just Quality — It''s Met Expectations: Expectancy-Disconfirmation and Course Evaluations

In short: Two cohorts can take an identical, equally well-taught course and rate it very differently — because satisfaction is not a direct read-out of quality but of the gap between what students expected and what they got. Oliver''s (1980) expectancy-disconfirmation model holds that satisfaction is a function of prior expectations and the disconfirmation of those expectations: exceed them and satisfaction rises, fall short and it drops, even when objective quality is held constant. For course evaluation this has a sharp implication: a "satisfaction" or "overall" score partly measures the expectations students arrived with — which the instructor can shape — rather than the instruction alone. Managing expectations is therefore a legitimate, evidence-based lever, and a confound to interpret carefully.

What the research says

In 1980, Richard Oliver published "A Cognitive Model of the Antecedents and Consequences of Satisfaction Decisions" in the Journal of Marketing Research — the paper that established the expectancy-disconfirmation paradigm, the dominant account of how satisfaction forms for more than four decades. Oliver''s model is deceptively simple. A person enters an experience with an expectation (a prediction of how good it will be). They then perceive the actual performance. Satisfaction is driven not by performance alone but by disconfirmation — the direction and size of the gap between performance and expectation. Positive disconfirmation (performance exceeds expectation) produces satisfaction; negative disconfirmation (performance falls short) produces dissatisfaction; confirmation (performance matches expectation) is neutral. Oliver supported the model with a two-stage field study and demonstrated that expectation and disconfirmation jointly predict satisfaction and downstream attitudes and intentions.

The corollary is the part that unsettles any naive use of satisfaction scores: expectation has both a direct and an indirect effect on satisfaction. Higher expectations raise the bar performance must clear, so — holding quality constant — raising expectations can lower satisfaction through the disconfirmation term. This is why a course that markets itself as transformative can underperform on evaluations relative to an identical course that promised less and quietly over-delivered.

Oliver later refined the model to incorporate affect and attribution ("Cognitive, affective, and attribute bases of the satisfaction response," 1993, Journal of Consumer Research), showing satisfaction is not purely a cognitive calculation but also carries emotional and causal-attribution components — students ask not only "was it as good as I expected?" but "how did it make me feel?" and "whose fault was the gap?"

The direct translation to higher education comes from Appleton-Knapp and Krentler (2006), "Measuring Student Expectations and Their Effects on Satisfaction" (Journal of Marketing Education). Across two studies they tested whether the expectancy-disconfirmation paradigm predicts student satisfaction with a course — and found a crucial methodological wrinkle. When students'' recalled expectations and current perceptions were measured together at the end of term, the extent to which expectations were met was a good predictor of satisfaction. But when expectations were measured at the start of term, the degree to which they were later fulfilled was a weak predictor of satisfaction. Expectations, in other words, are partly reconstructed after the fact — and satisfaction tracks the remembered gap more than the original one.

Why it matters for course evaluation in practice

Three practical consequences follow for any quality-assurance office that takes satisfaction or "overall" items seriously.

First, a satisfaction score is a relative measure, not an absolute one. It encodes the expectations the cohort brought as much as the teaching they received. Comparing the satisfaction of a first-year service course (students often expecting little, easily pleasantly surprised) with a flagship final-year elective (students expecting brilliance) is comparing different disconfirmation baselines, not just different teaching. This is a close cousin of reference bias and a reason raw satisfaction comparisons across very different courses can mislead.

Second, expectation management is a legitimate pedagogical lever — and a potential gaming vector. Setting accurate, slightly conservative expectations at the start of a course (clear scope, honest difficulty, realistic workload) tends to reduce negative disconfirmation and improve satisfaction without changing the teaching at all. Used honestly, this is good practice: students who know what they are getting are less likely to feel let down. Used cynically — over-promising in recruitment, under-promising at the start of term — it is a way to move evaluation scores without improving the course, which committees should be alert to.

Third, the Appleton-Knapp and Krentler result warns against over-trusting retrospective expectation items. Asking students at the end of term "did this course meet your expectations?" largely measures a reconstructed expectation, not the one they actually held in week one. If you want to understand and manage the expectation gap, you have to capture expectations early and compare — not rely on end-of-term recall.

A concrete illustration makes the stakes clear. Imagine two parallel sections of the same statistics module, taught from the same materials to the same standard. Section A is described in the handbook as a gentle, applied introduction; Section B is promoted as a rigorous springboard into advanced econometrics. Students in Section B arrive expecting more, so identical teaching produces larger negative disconfirmation and lower satisfaction — not because the teaching was worse, but because the promise was bigger. A naive reading of the two evaluation reports would wrongly conclude that Section A is the better-taught course. Expectancy-disconfirmation reframes the comparison: the scores differ because the baselines differ, and the fair question is how each section performed relative to what its own students were led to expect.

Limitations and honest caveats

The expectancy-disconfirmation model is powerful but not the whole story, and a careful reader should resist over-applying it. First, it originates in consumer marketing, and importing it wholesale risks the "student-as-consumer" framing that much of the higher-education literature regards as reductive; education is a co-produced developmental process, not a transaction, and satisfaction is at best a partial proxy for educational value. Second, the model has known empirical tensions — the disconfirmation versus perceived-performance debate questions whether the gap term adds much beyond perceived performance itself, and effects vary by context. Third, Appleton-Knapp and Krentler''s reconstruction finding complicates measurement: if expectations are reshaped after the fact, both prospective and retrospective measures are imperfect in different ways. Fourth, expectations are heterogeneous within a cohort — students do not share one baseline — so a single satisfaction mean blends many disconfirmation calculations. Finally, none of this addresses whether satisfaction tracks learning; a course can satisfy by meeting low expectations while teaching little, which is exactly why satisfaction must be triangulated with outcome evidence rather than trusted alone.

How Koji incorporates this

Koji for Education is designed to make the expectation gap visible rather than letting it hide inside a single satisfaction number:

• Early and end-of-cycle expectation capture. Because Appleton-Knapp and Krentler show retrospective expectation items measure a reconstructed baseline, Koji supports formative, start-of-term collection so a programme can capture expectations prospectively and compare them with later perceptions — measuring real disconfirmation rather than end-of-term recall.
• Probing the gap, not just the score. When a student reports dissatisfaction, Koji''s AI-moderated conversational interview asks what did you expect, and where did the course diverge? — distinguishing a genuine quality shortfall from an unmet (and perhaps unrealistic) expectation. That distinction is invisible to a bare overall rating. Koji''s core research platform at koji.so applies the same expectation-probing engine to product and customer research, where met-expectations dynamics are identical.
• Expectation-aware reporting. Koji is designed to flag that satisfaction and "overall" items are relative to cohort expectations, discouraging naive cross-course satisfaction comparisons in the way reference bias and disconfirmation baselines warrant.
• Structured items that separate constructs. Using scale, single_choice, and open_ended questions, Koji lets you measure perceived performance, met expectations, and affect as distinct signals — mirroring Oliver''s 1993 refinement that satisfaction has cognitive, affective, and attributional components — rather than collapsing them.
• Closing the loop on expectation-setting. Where unmet expectations stem from unclear scope or workload signalling, Koji''s action-tracking supports concrete fixes (clearer module descriptions, honest difficulty framing) and re-measurement, so expectation management is done transparently and improves the student experience rather than merely the score.

Koji is designed to mitigate the confound that expectations introduce into satisfaction scores; it cannot make satisfaction a substitute for evidence of learning, and treats it as one triangulated signal among several.

Related Resources

• Are Students Customers? The SERVQUAL Gap Model, HEdPERF, and What Service-Quality Thinking Adds to Course Evaluation
• The Student-as-Consumer Effect: What a Consumer Mindset Does to Course Evaluations
• Beyond the Lecturer: What the Course Experience Questionnaire (CEQ) Measures and Why It Predicts Learning
• Should You Use Net Promoter Score for Courses? The "Would You Recommend" Question in Higher Education
• Why "I Learned a Lot" Can't Be Compared Across Courses: Reference Bias

References

• Oliver, R. L. (1980). A cognitive model of the antecedents and consequences of satisfaction decisions. Journal of Marketing Research, 17(4), 460–469. https://doi.org/10.1177/002224378001700405
• Oliver, R. L. (1993). Cognitive, affective, and attribute bases of the satisfaction response. Journal of Consumer Research, 20(3), 418–430. https://doi.org/10.1086/209358
• Appleton-Knapp, S. L., & Krentler, K. A. (2006). Measuring student expectations and their effects on satisfaction: The importance of managing student expectations. Journal of Marketing Education, 28(3), 254–264. https://doi.org/10.1177/0273475306293359