Industry
Financial
Use Case
Consumer Preference Modeling
The Challenge
Annuities are among the most financially complex products consumers encounter. Attributes like inflation-adjusted payment increases, period-certain guarantees, and company financial ratings all affect the true expected value of a policy, but most consumers aren't equipped to calculate that value on their own. Traditional research assumed that rational buyers simply maximize their expected payout — but that assumption left a big gap between what research predicted and how real people actually behaved.
The team needed a research approach that could separate two distinct forces: the financial impact of each product attribute and the psychological weight consumers assign to that attribute beyond its financial impact. Understanding that gap, and what closes it, was the key to both better product design and smarter communication strategies.
The Solution
The researchers used CBC analysis to design a discrete choice experiment in which participants aged 40 to 65 evaluated 20 sets of annuity options. Each option varied across four attributes: starting monthly income, annual payment increases, period-certain guarantee length, and the financial strength rating of the issuing company. Participants also had the option to opt out entirely, simulating a real-world "keep managing it myself" choice.
To test whether information presentation changes decisions, the team ran two versions of the survey. The first presented annuities using standard industry descriptions. The second added a cumulative payout table showing what each annuity would pay out if the buyer lived to ages 70, 75, 80, 85, 90, or 95. The same underlying attributes, just displayed with the math already done.
Sawtooth tools and methods used:
- Choice-Based Conjoint (CBC): Structured the discrete choice experiment across 20 tasks per respondent, capturing individual-level preference data at scale
- Hierarchical Bayes estimation: Modeled population-level heterogeneity while preserving individual-level preference estimates
- CBC survey design optimization: Generated statistically efficient choice sets to maximize data quality given the complex, interrelated attribute structure
The Outcome
- Consumers systematically undervalue inflation protection. In the standard information condition, participants treated percentage-based annual increases as a liability rather than an asset — willing to accept lower expected payouts just to avoid them. This wasn't irrational stubbornness: it largely reflected an inability to mentally calculate compounding over decades.
- Enriched information partially corrects the bias. When the cumulative payout table was shown, undervaluation of annual payment increases dropped substantially. Participants who expected to live longer than average showed the strongest de-biasing effect, consistent with a rational model. The enriched format didn't add new information — it just made existing information legible.
- Period-certain preferences follow an inverted-U curve. Consumers overvalued medium-length guarantees (around 20 years) and undervalued both short and very long ones. This pattern held across both information conditions, suggesting it reflects something deeper than miscalculation.
- Smart repackaging can more than double demand. By restructuring the mix of attributes — shifting payout structure without changing the expected financial value — researchers showed that demand for the same-value annuity could more than double. The optimal product design differed depending on how it would be described to consumers.
- Perceived fairness drives baseline annuity acceptance. Across all conditions, individuals who viewed annuities as fair showed significantly higher preference scores. This psychographic signal could be a valuable segmentation lever for marketers targeting undecided consumers.