Skip to content
BoldSignal.

Work in progress · September 2026

Grounded in neuroscience. Evaluated on advertising.

BoldSignal brings together computational models of neural responses and advertising research. The aim is a consistent view across creative, with scientific interpretation of what the estimates can — and cannot — tell us.

01

From creative to estimated response

A brain-encoding model learns a relationship between representations of content and measured neural responses in research data. Applying that relationship to video creative produces estimates that can be summarized and compared.

The customer supplies creative for analysis. The estimates come from the content; this workflow does not collect brain data from the customer's audience.

02

What a profile represents

A profile is a set of estimated neural-response summaries, rather than a direct reading of a viewer's attention, emotion, memory, or purchase intent. Mapping a neural target to a psychological interpretation requires care.

In this context, a mechanism is a proposed account of how content is processed. A model estimate by itself does not establish that mechanism or a causal effect on behavior. The model version, targets, aggregation, units, and applicable reference set should be specified for an engagement.

03

How we evaluated it

Consumer-evaluation analysis. Our September 2026 work in progress reports a common sample of 2,349 TV advertisements across 172 brands, analyzed alongside consumer pretest evaluations. Those evaluations are distinct from observed sales or measured brain responses.

External fMRI comparison. The work in progress separately evaluates selected estimated neural summaries against measured fMRI, with eligible samples of 37 U.S. ads and 32 Dutch ads. These are smaller comparisons, not brain measurements for all 2,349 advertisements. Performance varies by target and cohort.

This research is a work in progress, not a peer-reviewed validation of a commercial release.

04

Scope and interpretation

A useful interpretation depends on the creative format, model target, reference coverage, and decision at hand. Differences between estimated profiles do not, on their own, identify the cause of those differences or the creative that will sell more.

Sensitivity to a particular edit needs to be established before drawing conclusions from that edit. Consumer research and subsequent testing can help examine hypotheses that arise from a portfolio review.

For a proposed engagement, we discuss the question and confirm the model and profile specification, coverage, and interpretation scope before analysis.

Work in progress

Overgoor et al. · Work in progress · SSRN

NeuroViT: A Brain-Aligned Vision Transformer for Product Image Analysis

Our NeuroViT research develops brain-aligned image measures and studies their relationship with consumer responses to product images. This working paper is part of the broader research program; the video-advertising analysis described above is a separate study.

Published foundation

Chan et al. (2024) · Journal of Marketing Research

Neural Signals of Video Advertisement Liking: Insights into Psychological Processes and Their Temporal Dynamics

This published study examines measured fMRI responses during video advertisements using meta-analytic decoding. Hang Yee Chan is the lead author. It is part of the research foundation, separate from the current estimation pipeline, and does not validate a BoldSignal commercial model.

From research to a review

A portfolio review adds neural response analysis to the evidence you already have. We assess the estimated processing patterns alongside the creative and available audience outcomes, then explain what those relationships imply for your work.

See how the review is organized ↗