The Researcher

Dmitry Zharnikov

Independent Researcher

Brands are not what companies build. Brands are what observers perceive — and different observers, processing the same signals through different spectral profiles, perceive structurally different brands. This is not a communication failure. It is the fundamental mechanism of brand perception.

I spent 25 years discovering this across every vantage point the industry offers.

As a journalist, I learned to see how organizations present themselves — the gap between intended signal and received meaning. As a press secretary, then communications director of an industrial holding, then deputy director for brand management at one of the largest railway companies, I learned what works and what doesn't from inside organizations operating at national scale. As a PR and marketing consultant, I synthesized patterns across industries and clients — the recurring structural problems that no existing framework could adequately diagnose.

Then large language models arrived, and with them a new kind of research instrument. I could vibe-code market research tools, run systematic brand analyses at scale, and — for the first time — look at two decades of practitioner intuition through a data lens. What emerged wasn't a refinement of existing theory. It was a different framework entirely.

I formalized these observations into Spectral Brand Theory: a computational framework that decomposes brands across eight perceptual dimensions, models audience cohorts through formal spectral profiles, and runs an AI-native analytical pipeline that compresses multi-week consulting engagements into single analytical sessions.

The framework has been validated across five brands — Hermès, IKEA, Patagonia, Tesla, Erewhon — producing nine novel mechanisms not found in existing branding literature, including structural absence as strategy, five distinct coherence types, and the independence of brand power from brand health. Cross-model replication (Claude and Gemini) confirms the findings are framework-driven, not model-specific. The toolkit is open-source.

Education

Links