CoChat
Sign Up
Cited by 330Consumer Market Behavior and PricingInformation Retrieval and Search BehaviorDigital Marketing and Social Media

The Power of Rankings: Quantifying the Effect of Rankings on Online Consumer Search and Purchase Decisions

Open access
Ursu, Raluca M.·2018·Marketing Science
DOI 10.1287/mksc.2017.1072
Open this paper in CoChat — free account

Abstract

Online search intermediaries, such as Amazon or Expedia, use rankings (ordered lists) to present third-party sellers’ products to consumers. These rankings decrease consumer search costs and increase the probability of a match with a seller, ultimately increasing consumer welfare. Constructing relevant rankings requires understanding their causal effect on consumer choices. However, this is challenging because rankings are endogenous: consumers pay more attention to highly ranked products, and intermediaries rank the most relevant products at the top. In this paper, I use the first data set with experimental variation in the ranking from a field experiment at Expedia to make three contributions. First, I identify the causal effect of rankings and show that they affect what consumers search, but conditional on search, do not affect purchases. Second, I quantify the effect of rankings using a sequential search model and find an average position effect of $1.92, which is lower than literature estimates obtained without experimental variation. I also use model predictions, data patterns, and a feature of the data set (opaque offers) to show rankings lower search costs, instead of affecting consumer expectations or utility. Finally, I show a utility-based ranking built on this model’s estimates benefits consumers and the search intermediary. Data and the online appendix are available at https://doi.org/10.1287/mksc.2017.1072 .

Integrity signals

Registry

Crossref · 10.1287/mksc.2017.1072

Cites

39 of 39 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Claims the paper makes

We couldn’t retrieve this paper’s full text, so its claims and self-check weren’t run. The registry and citation checks still apply.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Online search engines lower the cost of locating information without affecting expectations or utility for consumers.

    “For example, previous research has found that online search engines lower the cost of locating information without affecting expectations or utility for consumers [25,26].”

    From Consumer Responses to Generative AI Chatbots Versus Search Engines for Product Evaluation (2025) · cites this paper as [26]

  • The position coefficient on inspection is negative and significant for all tested LLMs, ranging from -0.0002 to -0.0005, which is four to ten times smaller than the human benchmark of -0.0019.

    “The position coefficient on inspection is negative and significant for all LLMs, ranging from -0.0002 to -0.0005 (all ps<0.001;Table 2, left panel), whereas the human benchmark is -0.0019 (p<0.001).”

    From Does Rank Still Matter? Position Bias When AI Agents Shop on Our Behalf (2026) · cites this paper as Ursu 2018

  • In the hotel search environment, AI agents search more deeply than human consumers, with mean inspections per session ranging from 1.63 to 5.83 across the tested LLMs versus 1.12 for humans.

    “AI agents search more deeply than humans and never decline to buy.”

    From Does Rank Still Matter? Position Bias When AI Agents Shop on Our Behalf (2026) · cites this paper as Ursu 2018

  • The effect of position on the final choice varies by LLM: it is indistinguishable from zero for Flash and Pro, significant but smaller than humans for Sonnet, and significant and similar magnitude to humans for Flash Lite.

    “For Flash and Pro, it is indistinguishable from zero (-0.000014, p=0.338 and -0.000008, p=0.606, respectively).”

    From Does Rank Still Matter? Position Bias When AI Agents Shop on Our Behalf (2026) · cites this paper as Ursu 2018

  • Under the alternative prompt, search depth is marginally lower (M=2.92 vs. M=3.12) but remains far above the human benchmark (M=1.12).

    “Search depth is marginally lower ( M Alternative = 2.92 vs. M Original = 3.12 ) but remains far above the human benchmark ( M Human = 1.12 ).”

    From Does Rank Still Matter? Position Bias When AI Agents Shop on Our Behalf (2026) · cites this paper as Ursu 2018

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Related

Authors
Ursu, R. M.
Affiliations
New York University (US)
Corresponding author
Ursu, Raluca M.
ORCID
Ursu, Raluca M.
Venue
Marketing Science
Publisher
Institute for Operations Research and the Management Sciences (INFORMS)
Published
2018-08
Volume
37
Issue
4
Pages
530-552
ISSN
  • 0732-2399 (ISSN-L)
  • 1526-548X
DOI
10.1287/mksc.2017.1072
OpenAlex
W3124312273
Cited by
328
References
39
Citations by year
2026: 47 · 2025: 53 · 2024: 57 · 2023: 35 · 2022: 44 · 2021: 35 · 2020: 26 · 2019: 11 · 2018: 12 · 2017: 4 · 2016: 1 · 2015: 3
FWCI
37.64
Type
journal article
Language
en
Open access
Licence
cc-by-nc-nd
Locations
  • Marketing Science (publishedVersion)
  • Knowledge@UChicago (University of Chicago) (submittedVersion, cc-by-nc-nd)
  • University of Chicago
Topics
  • Social Sciences › Business, Management and Accounting › Marketing › Consumer Market Behavior and Pricing
  • Physical Sciences › Computer Science › Information Systems › Information Retrieval and Search Behavior
  • Social Sciences › Sociology and Political Science › Digital Marketing and Social Media
Keywords
search engine ranking, consumer search behavior, purchase decision, position effect, search costs, consumer welfare, Expedia, third-party sellers

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review