Why Food and Health Attitudes Matter
The food and health research space is one of the most active areas in consumer insights. Brands, retailers, and policymakers all rely on attitudinal data to understand how consumers think about nutrition, dietary guidelines, food processing, and health claims. The International Food Information Council (IFIC) has published its annual Food & Health Survey for nearly two decades, making it one of the most respected benchmarks in the field.
The 2025 edition surveyed a nationally representative sample of 3,000 U.S. adults on topics ranging from dietary guideline familiarity to protein consumption attitudes, fiber awareness, and perceptions of ultra-processed foods. For Simsurveys, this presented an ideal benchmark: a large, well-constructed survey covering a diverse range of consumer attitudes in a domain where synthetic data could deliver real value to CPG companies, food brands, and health-focused organizations.
Study Design
We generated 1,000 simulated respondents using the Simsurveys Consumer model and compared their responses against the published IFIC topline results. The survey instrument covered 26 items organized into 16 question groups, spanning both single-select attitudinal questions and multi-select behavioral items.
For single-select questions, we measured alignment using KL divergence — our standard metric for comparing probability distributions. For multi-select questions, where respondents could choose multiple options from a list, we used Rank-Biased Overlap (RBO) to evaluate how well the synthetic model reproduced the ordering and relative frequency of selected items. RBO is particularly well-suited for comparing ranked lists where the top items matter more than the tail.
Strong Alignment on Core Attitudes
The Consumer model performed well across the majority of the IFIC survey items. Dietary guidelines familiarity — a foundational question about whether consumers are aware of and follow official dietary recommendations — achieved a KL divergence of just 0.041. Protein consumption attitudes scored an exceptionally tight 0.009, and fiber awareness came in at 0.008. These are among the lowest divergence scores we have recorded across any validation study.
Beyond the headline numbers, the model showed strong directional alignment on the rank ordering of consumer attitudes. When the IFIC survey found that a particular concern ranked first among consumers, the synthetic model consistently placed it first or second. The magnitude of concern segments — the proportion of consumers expressing high, moderate, or low concern about specific food and health topics — also tracked closely between the real and simulated samples.
This kind of structural alignment is arguably more important than point-estimate accuracy for most marketing and strategy applications. When a brand team is trying to understand which health claims resonate most, or which dietary trends are gaining momentum, the relative ranking and magnitude of consumer segments matters more than whether a specific percentage is off by two or three points.
Key results: Dietary guidelines familiarity (KL=0.041), protein consumption attitudes (KL=0.009), fiber awareness (KL=0.008). Strong directional alignment on rank ordering and magnitude of concern segments across all 16 question groups.
Where Synthetic Respondents Diverged
The most consistent deviation we observed was that simulated respondents tended to show somewhat stronger stated preferences or awareness compared to the real IFIC sample. On questions about familiarity with dietary guidelines, awareness of specific nutrients, and stated intent to follow health recommendations, the synthetic sample skewed slightly more informed and more intentional.
This is a pattern we have observed in other consumer validation studies as well. The underlying language models tend to generate responses that reflect a slightly idealized consumer — one who is marginally more aware, more health-conscious, and more articulate about their preferences than the average survey respondent. For most research applications, this bias is small enough to be manageable, but researchers should be aware of it when interpreting results on self-reported awareness and behavioral intent questions.
It is worth noting that this awareness inflation is also present in traditional survey research. Panel respondents who opt in to surveys are themselves more engaged and informed than the general population. The synthetic bias and the panel bias may, in some cases, be operating in the same direction.
Research-Grade Data for Food and Health
Based on the results of this validation, we are confident that the Simsurveys Consumer model produces research-grade data suitable for marketing analysis, insight generation, and decision support in the food and health domain. The alignment on core attitudinal questions is strong, the directional patterns are reliable, and the areas of divergence are well-understood and predictable.
For CPG companies, food brands, and health-focused organizations, this means the ability to run consumer studies on dietary attitudes, nutrition perceptions, and food processing concerns at a fraction of the cost and timeline of traditional panel research. The model supports all standard question types and can be segmented by demographic, psychographic, and behavioral variables.
The full validation report is available for download on our validation studies page. To explore the Consumer model, visit the model page or create a free account to run your first consumer study.