Gravity is a decision system for CPG concepts. It structures fragmented market evidence drawn from nine public evidence streams, cross-references each signal against independent sources, weights what survives, and converts recurring patterns into a scored diagnosis and one clear next move. As of August 2026 it holds 21,170 classified signals across 18 CPG categories.
Nine independent evidence streams. One source is a signal; repetition across sources that do not depend on each other is a pattern. Every stream below lists what it reveals, what is actually read, and what the repetition tells you.
What people repeatedly want, question, reject, complain about, and integrate into their routines.
Where attention, demand, and category activity appear to be accumulating.
Publicly available market evidence from research firms, consultancies, institutions, and industry organizations.
How products are positioned, priced, formulated, packaged, and presented.
What the market looks like where products are actually sold.
What founders, operators, investors, buyers, and category experts are discussing before it becomes formal market consensus.
What is being repeatedly discussed, questioned, or prioritized across industry events.
Patterns learned from how real CPG brands launch, grow, reposition, scale, stall, or fail.
What ingredients may credibly support, where interest is growing, and where formulation or claims risk begins.
A point-in-time snapshot of the classified corpus, August 2026. Streams refresh on different cadences, so these counts move.
| Stream | What it holds | Signals |
|---|---|---|
| Live Intel Feed | Competitive moves, funding rounds, retail expansions and category launches, tracked twice weekly. | 5,898 |
| Research Findings | Structured findings extracted from category reports, market studies and industry research. | 4,667 |
| PDF Library Signals | Signals classified out of 85 full-text research documents held in the library. | 4,013 |
| Ingredient Signals | Social velocity, formulation trends, clinical signals and adoption patterns, by ingredient. | 2,694 |
| Market Trends | Directional category movement — where attention and demand are accumulating. | 1,599 |
| Consumer Intel | Review mining, consumer language, friction points and stated unmet needs. | 1,464 |
| Brand Intel | Deep profiles across tracked competitors — GTM, velocity, repeat, and live competitive signal. | 835 |
| Total | Across 18 CPG categories, including 85 full-text research documents | 21,170 |
Seven steps. The evidence enters as scattered public artefacts and leaves as one recommendation with its reasoning visible.
Evidence arrives in incompatible shapes — a review, a podcast transcript, a retailer assortment, a research chart. Step one makes them comparable.
A signal is checked against the other streams. One source saying something is an anecdote; several independent sources saying it is evidence.
Not all evidence carries equal weight. Source independence, recency, specificity and corroboration determine how much a signal counts.
Repetition across independent streams is what separates a pattern from noise. This is the step that reads a market before it becomes a trend.
The pattern is compared against brand case precedent — what happened to products that met similar conditions.
Market, product and business model are read together, surfacing what holds, what conflicts, and what is still unproven.
The diagnosis converts into a recommendation a founder can act on, with the reasoning visible.
Ask Gravity helps you understand the market. Gravity Scan evaluates the concept. Aterra helps make the decision.
Focused market questions, supporting evidence, and one clear next move.
Ask Gravity →Complete concept diagnosis through the Gravity decision framework, returned as a Gravity Intelligence Brief.
Run a Gravity Scan →Senior interpretation, prioritization, and strategic guidance from an operator.
Work with Aterra →A methodology is only as credible as the boundaries it states. These are the limits, stated plainly.
Where public excerpts from firms such as Euromonitor, SPINS, NielsenIQ, Mintel or Circana are read, they are public excerpts. Gravity does not claim subscription-level access to those datasets, and does not resell them.
It does not run surveys, recruit respondents, or simulate synthetic consumers. It reads behaviour that already happened in public — reviews, search, assortment, launches, conversation.
The corpus is built for U.S. category conditions. A read on another market would be inference, not evidence, and is out of scope.
It reports what the available evidence supports and where it is thin. Markets move, and evidence can be absent rather than negative — the brief says which.
The live intel feed is tracked twice weekly; research and library signals update as documents are added; some slower-moving streams refresh on a longer cycle. Counts on this page are a point-in-time snapshot.
Gravity reads nine public evidence streams: consumer behaviour, market momentum, public industry research, product and market activity, retail and channel signals, podcast and expert intelligence, industry event signals, brand case studies, and ingredient and formulation evidence. As of August 2026 that amounts to 21,170 classified signals across 18 CPG categories.
No. Gravity reads publicly available evidence, including public excerpts and summaries from major research firms. It does not hold subscription-level access to proprietary databases and does not resell licensed data. Where a benchmark comes from a public excerpt, that is what it is.
A panel asks people what they would do. Gravity reads what people already did — what they bought, reviewed, searched for, complained about and repurchased — alongside what the shelf, the trade press and operators are doing. It is evidence of revealed behaviour rather than stated intent, and it runs in minutes rather than weeks.
One source is a signal. Repetition across independent sources becomes a pattern. Gravity weights evidence by source independence, recency, specificity and corroboration, and only treats something as a pattern when it recurs across streams that do not depend on each other.
No. Every signal traces to a real, public artefact — a review, a listing, a transcript, a document, an assortment. AI is used to classify and cross-reference that evidence, not to invent respondents.
The live intel feed is tracked twice weekly. Research and library signals update as documents are added. Counts published here are a snapshot taken in August 2026; slower-moving streams refresh on a longer cycle, so individual stream ages vary.
Yes — a complete worked example is published as a sample Gravity Intelligence Brief, showing the scoring, the evidence cited, and the recommendation.
Gustavo Nader, founder of Aterra Studio and co-founder of AMAZ, a functional beverage brand distributed through Whole Foods, Erewhon and Sprouts. The interpretation layer is built from operating the exact decisions Gravity reads.
Written and maintained by Gustavo Nader, founder of Aterra Studio and co-founder of AMAZ — a functional beverage brand distributed through Whole Foods, Erewhon and Sprouts, and an Expo West Top 5 pick. Gravity's interpretation layer is built from operating the exact decisions it reads.
Published 2026-08-05 · Last reviewed 2026-08-05 · Corpus snapshot August 2026 · See a worked example →