How do different audience segments rate the same product differently?
This explores why the same product earns different ratings from different groups of people — and the corpus suggests the gap is rarely about the product itself, but about who self-selects to rate, what priors they bring, and who they imagine is watching.
This explores why the same product earns different ratings from different groups of people. The corpus's recurring answer is unsettling: most of the variation isn't about the product at all. It's about which segment shows up to rate, what they expected before they arrived, and who they think is reading.
Start with who even enters the rating pool. Review aggregates aren't a sample of everyone — they're a sample of people who expected to be satisfied enough to buy in the first place Do online reviews actually measure product quality or just buyer preferences?. On top of that filter, participation itself is lopsided: only people with strong opinions bother to rate, producing U-shaped distributions where the lukewarm middle goes silent Why do people bother writing online ratings at all?. So before any segment 'rates differently,' the segments that rate at all are already skewed in opposite directions.
Then there's the prior each segment carries in. One striking finding is that the recommender that surfaces a product effectively selects its audience: 'frequently-bought-together' and 'co-viewed' networks pull in different people with different expectations, which is why connected products converge in one network and diverge in another Do different recommender types shape opinion convergence differently?. The path to the product shapes the verdict on it. This is the same lesson the persuasion-research note delivers from a different angle: linguistic features that look like they 'cause' a positive response often vanish once you control for the reader's ideology — what looked like persuasive language was really audience-text matching all along Do linguistic features of persuasion stay the same across audiences?. Segments differ because the text lands on different priors, not because the text is different.
The most human mechanism is self-presentation. Posters systematically lower their public ratings after reading negative reviews — even when their own experience was positive — because negative reviewers read as more intelligent. Private raters, with no audience to perform for, show no such shift Why do online reviewers publish negative ratings despite positive experiences?. So 'segment' isn't only demographics; it's context. The same person is a different rater in public than in private. That instability compounds: the same user gives the same item ratings that swing by multiple stars across sessions from mood, anchoring, and rating style alone Why do the same users rate items differently each time?, and early ratings then bias the ratings that follow them, so segment differences harden into the historical record Do online ratings actually reflect independent customer opinions?.
Here's the part you might not have known you wanted: because so much rating variance is about audience composition rather than product quality, the useful move is to model the audience directly. One line of work shows you can cluster raters by latent traits — expertise, learning style — rather than by the words they use, capturing *who people are* instead of *what they said* Can LLMs extract audience traits better than comment similarity?. Once you can see the segments, the divergence stops being noise and becomes signal — and difference can even be designed in, as with recommenders that deliberately lean on friends with *unlike* tastes instead of averaging everyone toward the same opinion Can friends with different tastes improve recommendations?.
Sources 9 notes
Only consumers expecting satisfaction purchase and review, creating two selection filters. Research shows early reviewers shape later perceptions, altruism affects learnability, and summary statistics can actually slow quality discovery. Observed ratings misrepresent the satisfaction distribution of all potential buyers.
Lafky's experiments show raters care about both buyers and sellers rather than purely one or the other. Small participation costs create U-shaped distributions where only strong-opinion raters engage, biasing average ratings away from true quality.
Research shows that frequently-bought-together and co-viewed recommendation networks produce different opinion convergence patterns. The mechanism: each recommender type attracts different audience segments with different prior expectations, shaping both who sees products together and how they rate them.
The linguistic features that predict persuasion success change dramatically once political and religious ideology are added as statistical controls. Features appearing predictive in standard analyses often reflect audience-text matching rather than true language effects, making many published findings potentially artifacts of audience composition.
Posters systematically reduce their ratings in public when exposed to negative reviews, even with positive personal experience—because negative reviewers appear more intelligent. Private raters show no such shift, revealing a self-presentational mechanism tied to multiple-audience communication.
Amatriain et al. found that the same user gives substantially different ratings to the same item across sessions, shifting by multiple stars. This noise stems from temporal inconsistency, rater-specific biases, and anchoring effects—making ratings reflect both preference and rating-behavior rather than stable preference alone.
Moe and Trusov decomposed ratings into baseline quality, social-dynamics influence, and error, finding that prior ratings meaningfully affect subsequent ones. These effects have both immediate sales impact and long-term compounding effects through future ratings, though high opinion variance can eventually dampen the distortion.
LLM-extracted latent characteristics like expertise and learning style produce more homogeneous audience clusters than k-means on comment text alone. This captures who people are, not just what they say.
Social Poisson Factorization uses friends' diverse tastes to recommend items outside users' usual preferences, outperforming methods that pull friends' representations together. Networks add value through influence on anomalous choices, not taste similarity.