It is one of the most widely repeated rules in marketing: don’t overpromise. Set customer expectations too high, the thinking goes, and people will inevitably be disappointed when reality falls short. This idea has shaped advertising strategies, customer service training, and business school curricula for decades. But what if the accumulated research on the topic tells a different story?
A large-scale analysis of 150 published records, covering data from 58,597 consumers, found that higher expectations were actually associated with higher satisfaction, not lower. The analysis, published in the Journal of the Academy of Marketing Science, examined over 40 years of research on how expectations shape how pleased or displeased consumers feel after buying a product or service. The researchers found virtually no evidence supporting the popular idea that high expectations set consumers up for disappointment.
A decades-old debate with no clear resolution
The research was led by Tom Schiebler of Ludwig-Maximilians Universität München (now at FOM University of Applied Sciences in Munich), along with Nick Lee of Warwick Business School and Felix C. Brodbeck, also at Ludwig-Maximilians Universität München. Their goal was to settle a long-running theoretical debate in marketing research.
For more than four decades, the dominant framework for understanding customer satisfaction has been something called expectancy-disconfirmation theory. In plain terms, this theory says that how satisfied you feel after buying something depends on how the product’s actual performance stacks up against what you expected before the purchase. If performance exceeds expectations, you experience “positive disconfirmation,” a feeling that things went better than expected. If it falls short, you get “negative disconfirmation,” the sense that you were let down.
Within this framework, two competing ideas have long coexisted. The first, called assimilation, suggests that people tend to adjust their feelings of satisfaction toward their original expectations. If you expected a hotel to be wonderful, you might unconsciously overlook minor flaws and report being satisfied. The second, called contrast, predicts the opposite: if reality doesn’t match high expectations, people exaggerate the gap and feel even more let down. The question of which pattern actually holds in the real data has remained unresolved. As recently as 2010, one of the field’s founding researchers wrote that essentially any combination of outcomes was possible and none could be assumed in advance.
How the team pooled decades of research
To tackle this gap, the team conducted a meta-analysis. A meta-analysis is a statistical technique that combines the results of many individual studies to look for overall patterns. Rather than running a new experiment, the researchers gathered and analyzed the findings from existing research conducted by other scholars over several decades.
They searched five academic databases and screened the reference lists of earlier literature reviews. After evaluating thousands of records, they identified 150 that met their inclusion criteria, containing 168 independent studies and data from nearly 60,000 individual participants. Each study had to measure consumer satisfaction with a product or service and either measure or manipulate performance expectations.
Two coders independently categorized the studies according to several features: what type of expectation was measured (a prediction of what would happen versus a standard of what should happen), whether the product was a tangible good or a service, and what kind of study design was used (a one-time survey, a study tracking consumers over time, or a controlled experiment). They converted all reported results into a common statistical format, the correlation coefficient, and then used a random-effects model to calculate average effect sizes. This type of model accounts for the fact that effect sizes can genuinely vary from one study to the next, rather than assuming all studies are estimating an identical underlying number.
Higher expectations, higher satisfaction
The central finding was a positive correlation of r = .29 between performance expectations and consumer satisfaction. In other words, across roughly a hundred comparisons, people who went in with higher expectations tended to report being more satisfied, not less. This supports the assimilation idea and runs against the contrast prediction. Among the 99 studies that tested the expectations-satisfaction link, only two found a significantly negative relationship.
To check whether this pattern held even under more rigorous conditions, the team looked specifically at the 32 experimental studies in their dataset. Experiments are considered stronger evidence for cause-and-effect relationships because they involve manipulating expectations directly, rather than simply measuring them after the fact. The positive link between expectations and satisfaction remained significant in this subgroup, though smaller (r = .14).
The researchers also found that perceived performance was strongly and positively linked to satisfaction (r = .63), making it the single most important factor in the model. This is consistent with common sense: how well consumers think a product actually performed matters a great deal for how satisfied they feel.
Disconfirmation: predictor or just a close cousin of satisfaction?
One of the more provocative findings involved the concept of disconfirmation itself. Disconfirmation, in this context, refers to a consumer’s feeling that a product performed better or worse than expected. The researchers found a very strong correlation between disconfirmation and satisfaction (r = .61). At first glance, this might seem to validate the idea that exceeding expectations is the key driver of satisfaction.
But the team raised an important caution. They pointed out that the correlation was so high, with 39 individual studies showing correlations of .70 or above and 16 exceeding .80, that it raises questions about whether disconfirmation and satisfaction are really measuring different things. They also conducted a path analysis, a statistical technique that models multiple relationships at once, and found no indirect path from expectations through disconfirmation to satisfaction. In other words, the data did not support the popular chain of logic where higher expectations lead to greater disconfirmation, which then leads to lower satisfaction.
Adding another layer of surprise, the researchers discovered an unusual type of publication bias. Normally, when researchers analyze publication patterns, they find that very small, statistically insignificant results tend to be missing from the published record because journals are less likely to publish them. Here, the opposite appeared to be happening: very large effect sizes seemed to be missing. The team speculated that researchers who found extremely high correlations between disconfirmation and satisfaction might have chosen not to report them, perhaps because such high numbers raised uncomfortable questions about whether the two measures were truly distinct.
When do expectations matter most?
The analysis also explored conditions under which the link between expectations and satisfaction was stronger or weaker. Three factors stood out.
First, the type of product mattered. The positive association between expectations and satisfaction was stronger for services (r = .32) than for physical goods (r = .21). The researchers suggested this might be because the experience of using a service, like a restaurant visit or a consulting session, is inherently more ambiguous than evaluating a tangible product. When the experience is harder to evaluate objectively, consumers may rely more on their prior expectations to form judgments.
Second, study design played a significant role. The effect was largest in cross-sectional surveys (r = .42), where expectations and satisfaction were both measured at a single point in time after consumption. It was smaller in longitudinal studies (r = .29), where expectations were measured before the purchase, and smallest in experiments (r = .14). The researchers attributed this partly to hindsight bias: when people recall their expectations after the fact, their memories may already be colored by the experience itself.
Third, there was a trend suggesting that predictive expectations (“I think this product will perform at a certain level”) were more strongly linked to satisfaction than normative expectations (“I think this product should perform at a certain level”). However, this difference narrowly missed the threshold for statistical significance.
What businesses can take away
For marketers and business leaders, the results suggest several practical points. The common instinct to lowball customer expectations as a safety strategy may be misguided. Across the accumulated evidence, higher expectations were linked to higher satisfaction, not lower satisfaction. That said, this does not mean businesses should deliberately make false promises. The study found that perceived product performance remains far and away the strongest factor associated with satisfaction.
If a product or service does have genuine weaknesses, the findings suggest that lowering expectations is not the best remedy. Instead, businesses might focus on improving the actual performance of their offerings or preparing customers for potential shortcomings in a way that does not diminish their overall expectations.
The researchers also offered a warning about how companies interpret customer feedback. When a customer says a product “fell short of expectations,” the natural conclusion is that expectations were set too high. But the data suggest a different story: it is more likely that the product’s performance was perceived as too low, rather than that the initial expectations were too high. Negative disconfirmation, in other words, is more of a signal about performance problems than about an expectations-management failure.
Caveats to keep in mind
There are some important limitations. A meta-analysis can only work with the data that existing studies provide, and a large number of potentially relevant studies had to be excluded because they did not report the necessary statistics. The analysis also could not test for non-linear effects, such as the possibility that expectations help satisfaction up to a point but then start to hurt it once the gap between expectations and reality becomes too large. Most of the underlying studies used linear statistical methods, so this remains an open question.
Additionally, the strongest associations between expectations and satisfaction appeared in cross-sectional surveys, which are more vulnerable to hindsight bias and other measurement artifacts. The more conservative estimates from experiments, while still positive, were notably smaller. Businesses should keep this in mind when interpreting the size of the effect.
