Synthetic Survey Data Validated Against Walmart’s Retail Rewired 2025 Study

The Rise of Agentic Retail

Retail is entering a new phase. AI-powered product recommendations, conversational shopping assistants, and agentic experiences — where software acts on behalf of the consumer to find, compare, and even purchase products — are no longer speculative concepts. They are live features at the world’s largest retailers. Walmart’s Retail Rewired Report 2025 captures consumer attitudes toward these emerging experiences, measuring trust in AI recommendations, comfort with digital assistants, and willingness to delegate shopping decisions to autonomous agents.

For Simsurveys, the Retail Rewired study offered a compelling validation target. The questions sit at the intersection of technology adoption, trust, and consumer behavior — exactly the kind of attitudinal research where synthetic data needs to prove itself. These are not simple satisfaction questions; they probe emerging behaviors where consumer sentiment is still forming and where traditional panels may struggle to capture the full spectrum of attitudes.

Study Design

We generated 1,000 simulated respondents using the Simsurveys Consumer model, targeting a general adult population to match the Retail Rewired study’s sampling frame. We compared synthetic responses against the published topline results across the study’s core questions on AI trust, product search behavior, and digital assistant attitudes.

For single-select questions, we used KL divergence to measure distributional alignment. For multi-select items, we applied cosine similarity to compare the response vectors between real and simulated samples. Together, these metrics provide a comprehensive picture of how well the synthetic model reproduces both the shape of individual response distributions and the relative importance of different response options.

Trust in AI Product Recommendations

The first question examined who consumers trust most for product recommendations — AI algorithms, human influencers, friends and family, or none of the above. The KL divergence was 0.031, indicating strong distributional alignment. In the Walmart study, 49% of respondents said they don’t know or feel neutral about AI recommendations, compared to 46% in our synthetic sample. Twenty-seven percent of real respondents expressed trust in AI recommendations, while 38% of synthetic respondents did the same. Conversely, 24% of real respondents trusted influencer recommendations versus 17% in the synthetic sample.

The directional pattern is clear: both real and synthetic consumers show that AI trust is still nascent, with the largest segment remaining undecided. The synthetic model slightly overestimates AI trust and underestimates influencer trust — a pattern consistent with the observation that language models tend to reflect a somewhat more technology-forward consumer profile. Despite this, the overall distribution shape and the dominance of the “neutral/undecided” segment are well-captured.

Product Search Behavior

Q2 asked consumers how they typically search for products online — traditional search engines, retailer search bars, chatbot or AI assistants, or social media. The KL divergence was 0.091. Traditional search dominated in both samples: 69% in the Walmart study versus 75% in the synthetic sample. Chatbot and AI search tools were used by 13% of real respondents and 19% of synthetic respondents.

Again, the synthetic model captures the correct hierarchy — traditional search far outpaces AI-assisted search — while showing a modest tilt toward emerging digital behaviors. For most retail strategy and channel planning applications, the directional insight is the same: traditional search remains dominant, but AI-assisted shopping is a meaningful and growing minority behavior.

Digital Assistants: Improve or Detract?

Q3 asked whether digital shopping assistants improve or detract from the shopping experience. The KL divergence was 0.069. In the Walmart study, 48% said digital assistants improve the experience, compared to 46% in the synthetic sample. The proportion who said assistants detract, and those who were unsure, also tracked closely between the two samples.

This is one of the most balanced questions in the study — consumer opinion is genuinely split — and the synthetic model reproduced that balance accurately. The result is encouraging because it shows the model can capture ambivalence and divided sentiment, not just consensus positions.

Broader Patterns

Across the full set of Retail Rewired questions, the synthetic model showed strong directional alignment on the themes that matter most to retail strategists: consumer trust in AI is real but cautious, traditional shopping behaviors still dominate, convenience is the primary driver of AI adoption, privacy concerns remain a significant barrier, and willingness to delegate shopping decisions to autonomous agents is growing but far from mainstream.

These are the insights that drive real business decisions — channel investment, feature prioritization, messaging strategy, and competitive positioning. The synthetic data captures them reliably, even on questions about emerging behaviors where consumer attitudes are still crystallizing.

Implications for Retail Research

The Walmart validation demonstrates that synthetic consumer data can keep pace with the most forward-looking retail research. For brands, retailers, and technology companies exploring AI-powered shopping experiences, the Simsurveys Consumer model offers a fast, affordable way to test consumer attitudes, benchmark against industry data, and iterate on product and messaging strategies without the lead time and cost of traditional fielding.

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 retail study.