EEG for Ad Testing: Protocols and Metrics Topical Map

EEG measures voltage fluctuations produced by synchronized neuronal activity at the scalp with millisecond precision; in ad testing we use those signals to extract event-related potentials (ERPs), spectral power (e.g., alpha/beta/gamma bands), and connectivity metrics that index attention, cognitive workload, emotional valence, and memory encoding during specific ad moments.

Commonly used EEG metrics are frontal alpha asymmetry (approach/avoidance, proxy for positive valence), P300 amplitude and latency (attentional capture and novelty), theta power (memory encoding), and overall high-frequency broadband power (arousal/activation); mapping to commercial KPIs requires validation — for example, P300 and theta often predict ad recall, while frontal asymmetry correlates with purchase intent in multiple applied studies.

For within-subject designs testing short video or creative variations, 20–40 clean participants typically detect medium effects (d≈0.5) at 80% power; for between-subject comparisons or small effect sizes you should plan 50–100+ participants and incorporate robust preprocessing to reduce noise.

Best-practice protocols use time-locked stimulus markers, a pre-stimulus baseline of 200–500 ms, randomized stimulus order or counterbalancing across participants, at least 1–3 seconds of inter-stimulus interval for short ads, and multiple repetitions or longer exposure for measuring memory-related metrics; include calibration tasks and filler trials to monitor attention and reduce expectancy effects.

Preprocessing should include band-pass filtering appropriate to your metrics (commonly 0.1–45 Hz), line-noise removal, bad-channel interpolation, ICA or regression-based artifact correction for blinks and muscle noise, epoching around stimulus events, baseline correction, and trial rejection criteria documented in a reproducible pipeline.

EEG can predict downstream outcomes when combined with behavioral and demographic data: engagement and memory-related EEG metrics have been shown to explain a meaningful portion of variance in ad recall or short-term lift (commonly 10–40% in applied reports), but predictions of actual sales require larger multimodal models and out-of-sample validation.

Portable/consumer EEG (8–32 channels) can capture broad engagement trends and are useful for larger-sample, naturalistic tests but have lower spatial resolution and are more susceptible to movement/artifact; lab-grade systems (32–128 channels) offer better signal quality, source estimation and advanced ERP analyses but cost more and require trained technicians.

Synchronize all devices with shared timestamps or hardware triggers, align events in a master log, use eye-tracking to disambiguate visual attention and map EEG responses to gaze-driven exposures, and embed EEG-derived features as explanatory variables in A/B test models or uplift regressions for combined inference.

EEG is sensitive biometric data: obtain informed consent specifying data use and retention, anonymize raw signals, store data securely (encrypted at rest), limit commercial claims to validated endpoints, and follow local regulations on biometric profiling and data subject rights.

Use a layered reporting approach: executive summary with 1–3 actionable insights tied to KPIs, visual timelines of moment-by-moment metrics (attention, valence, memory), benchmark comparisons across creatives, statistical effect sizes with confidence intervals, and an appendix with methodology, preprocessing steps, and raw metric definitions for reproducibility.