Anchoring: How Does the First
Positioned Offer Affect a Purchase
Decision in Hotel Choice.
Herman Blöte
Marketing Management | Marketing Intelligence Thesis Defence
Goal of the
Research
81% of all holiday bookings are made online. (Holiday Habits Report, 2018)
Theory
The Anchoring and Adjustment heuristic (Tversky & Kahneman, 1974).
➢ The human tendency to rely on the first piece of information received.
➢ “an ubiquitous phenomenon in human judgement” (Furnham & Boo, 2011). Error cost
Hypothesis
The Data
POSITIVE NEGATIVE › +/- 10.000.000 rows of data. › 172 countries. › Represents 106.911.133,00 USD of purchases.› Data on both the choice, and non-chosen alternatives.
› Search and offer level data.
› MNAR data on every 5th position.
› Data scarcity.
› Anonymized.
› Mix of different pricing laws.
› Varying number of seen alternatives.
Methodology
Measures
➢ Dissimilarity to the anchor: Minkowski Distance (Price, Star rating, Location score). ➢ Error cost: number of individual + number of searched days.
Means-end Theory
➢ Every individual has an utility for an attribute - an individual will pick the offer that maximizes total utility.
➢ Adjust utility estimates, to best fit the observed choice.
Method
➢ Bayesian Inference – using Markov Chains.
➢ Increasing sampling bias with an extra metropolis step.
Results
Posterior distributions attribute utility
H1: The more dissimilar an option is to
the anchor, the less likely it is selected = rejected
➢ People lean towards, preferring offers similar to the anchor. But to much unexplained variance
Results
Minkowski distribution of heterogeneity for individual level variables
H2: Search importance weakens the
anchoring effect in online searches. = rejected
➢ Complete lack of an effect of search importance.
➢ Suggesting the confirmatory
hypothesis* is a better explanation for the mechanism of the
Implications
Additional finding
➢ higher error cost brings the utility for positioning closer to zero.
➢ Searches with higher error cost, are more indifferent towards positioning Theoretical
➢ Additional error cost, does not affect anchoring.
Methodological
Limitations Future research
› Inability to connect individuals over multiple searches.
› Some offer attributes are not in the data, for example: visual information or a pool.
› Lot of real-world effects, not controlled for.
› First positioned offer is not always the anchor.
› Testing the anchoring effect using an A/B test, include some outliers on the first position.
› Does error cost affect the utility of
Thank you.
Herman BlöteMarketing Management | Marketing Intelligence Thesis Defence