Echo Signal compares two snapshots of aggregated public text from the same source, taken at different times, and surfaces what has shifted. Rising terms, declining terms, tone changes, topic emergence, cultural-driver migrations. Built for the question every strategist eventually asks: is the conversation moving — and where?
What this tool is. Echo Signal performs a differential analysis between two text corpora that represent the same source type at two different times. It surfaces frequency-shifted terms, tonal changes across Plutchik's emotional dimensions, and migrations across Schwartz value drivers. The math is straightforward: log-odds ratio for term frequency comparison (Monroe, Colaresi & Quinn, 2008), cosine distance for emotional vectors, percentage-point delta for value categories.
What this tool is not. Not a forecasting engine. Not a predictor of what will happen next. Not a substitute for proper market research or longitudinal studies. The output is descriptive — what changed between two specific text samples you provided — not predictive. A rising term in your sample does not guarantee that term will continue rising in the wider world.
How to read the scores. Treat differential signals as hypotheses to investigate further, not as findings. A 200% increase in a low-frequency term may be noise; a 30% increase in a high-frequency term is usually substance. The velocity score reflects how much of the conversation's vocabulary has changed — high velocity means more flux, not necessarily more importance.
Source matching is critical. The single biggest source of error is comparing apples to oranges — for instance, press articles from one publication in T1 against forum posts in T2. Always match: same publication type, same kind of public source, same approximate audience. Otherwise the "trend" you detect is just a source artifact.
Ethical use. Echo Signal is designed for marketers, strategists, journalists, and researchers analyzing changes in public discourse on topics, brands, or industries. It is not designed for, and should not be used for, tracking changes in the speech of identifiable individuals over time, monitoring private communications, or any application targeting specific persons. Both snapshots must be public aggregate texts.