Snowflake Documentation

REACHDOG.TAXONOMY | 15 tables + 2 views | ~1.07B rows | ~18 GB compressed | Updated weekly

This dataset is a fully resolved intent-to-product graph spanning the English-language e-commerce landscape. 372 million products from 714,000 online stores, clustered into 29.6 million product neighborhoods, linked to 22.3 million search intent keywords with demand signals, and enriched across five semantic dimensions. Financial signals (pricing, competition, demand, opportunity) are available for 29.6 million clusters via the FINANCIAL_METRICS table, and per-keyword CPC/competition data is available for 53.6 million keywords via the KEYWORD_FINANCIALS table.

Where traditional product taxonomies organize products into a static tree (e.g., Apparel > Women > Dresses), this dataset builds a many-to-many semantic web that connects any search keyword to the products, stores, demographics, occasions, and price trends that surround it. Any node can reach any other node through structured, queryable paths.

Table Inventory

Table	Rows	Size	Description
PRODUCT_BRIDGE	361M	2.2 GB	Product-to-cluster mapping (every product assigned to a neighborhood)
INTENT_PRODUCT_MATCHES	223M	3.9 GB	Keyword-to-cluster relevance scores (top matches per keyword)
WEEKLY_DIMENSIONS	148M	1.0 GB	Weekly breakdowns across 12 dimensions (5 lifestyle + 7 financial)
PRICE_HISTORY_EVENTS	95M	1.3 GB	Individual price change events since Feb 2024
KEYWORD_FINANCIALS	53.6M	0.6 GB	Per-keyword CPC, competition, search volume
PRICE_HISTORY	45M	1.5 GB	Raw price history documents per product
MONTHLY_DIMENSIONS	35M	287 MB	Monthly aggregate breakdowns by dimension
INTENT_LINK_EDGES	32M	85 MB	Intent hierarchy parent-child edges
TITLE_CANONICALS	29.6M	4.7 GB	Cluster metadata: canonical title, size, enrichment tags
FINANCIAL_METRICS	29.6M	0.6 GB	18 financial signals per cluster
INTENT_PRODUCTS	22.3M	786 MB	Master keyword table with search volume and categories
INTENT_LINKS	1.6M	401 MB	Intent hierarchy parent nodes (seed/canonical)
SHOPS	714K	1.0 GB	Active stores with enrichment tags and geographic data
WEEKLY_TOTALS	59	<1 MB	Platform-level weekly totals
MONTHLY_TOTALS	14	<1 MB	Platform-level monthly totals

Views: CLUSTER_FULL joins TITLE_CANONICALS + FINANCIAL_METRICS into a single enriched cluster row. DIM_FINANCIAL_DIMENSIONS aggregates financial dimension tags for faceted browsing.

Entity Relationships

Core join paths
INTENT_PRODUCTS (22.3M keywords)
                        |
                        |-- INTENT_PRODUCT_MATCHES (223M) --> TITLE_CANONICALS (29.6M clusters)
                        |                                          |
                        |                                          |-- PRODUCT_BRIDGE (361M) --> individual products
                        |                                          |-- PRICE_HISTORY_EVENTS (95M)
                        |                                          |-- FINANCIAL_METRICS (29.6M)
                        |
                        |-- INTENT_LINKS (1.6M) / INTENT_LINK_EDGES (32M)  [keyword hierarchy]
                        |
                        v
                        KEYWORD_FINANCIALS (53.6M)  <-- per-keyword CPC/competition/SV
                        SHOPS (714K)  <-- connected via TOP_INTENTS[] and CATEGORIES[]

Join keys:

INTENT_PRODUCTS.KEYWORD_ID = INTENT_PRODUCT_MATCHES.KEYWORD_ID
INTENT_PRODUCT_MATCHES.CLUSTER_ID = TITLE_CANONICALS.CLUSTER_ID
TITLE_CANONICALS.CLUSTER_ID = PRODUCT_BRIDGE.CLUSTER_ID
PRODUCT_BRIDGE.PRODUCT_ID = PRICE_HISTORY_EVENTS.PRODUCT_ID
TITLE_CANONICALS.CLUSTER_ID = FINANCIAL_METRICS.CLUSTER_ID

INTENT_PRODUCTS

22,347,113 rows | The master list of every search intent keyword in the taxonomy.

Column	Type	Description
KEYWORD_ID	INT	Unique identifier
KW	VARCHAR	The keyword text
TYPE	INT	1=seed (1.0M), 2=canonical (2.1M), 3=intent (19.3M)
SV	INT	Measured search volume (Google/Amazon)
PARENT_SV	INT	Inherited search volume from parent keyword
CATS	VARIANT	Array of category labels

Search volume note: SV is measured directly from Google Ads or Amazon. PARENT_SV is inherited from the parent seed/canonical keyword. 82% of all keywords carry one or both signals. Never sum or average SV and PARENT_SV together.

INTENT_PRODUCT_MATCHES

223,471,130 rows | Links every keyword to its closest product neighborhoods with a relevance score.

Column	Type	Description
KEYWORD_ID	INT	FK to INTENT_PRODUCTS
CLUSTER_ID	INT	FK to TITLE_CANONICALS
SCORE	FLOAT	Semantic relevance score (higher = stronger match)
TITLE	VARCHAR	The matched cluster's canonical title
CLUSTER_SIZE	INT	Number of products in the cluster

TITLE_CANONICALS

29,637,507 rows | Each row represents a product neighborhood with its canonical title and enrichment tags.

Column	Type	Description
CLUSTER_ID	INT	Unique cluster identifier
SIZE	INT	Number of products in this cluster
CANONICAL	VARCHAR	Representative product title (max 125 chars)
TITLES	VARIANT	Array of sample titles from the cluster
ENRICHMENT_TAGS	VARIANT	Enrichment across 5+ dimensions (see below)
HAS_PRICE_HISTORY	BOOLEAN	Whether products in this cluster have price tracking

3.6 million clusters enriched (12.2% of all clusters). Enriched clusters cover the highest-priority neighborhoods and represent 192 million of the 372 million total products. Enrichment is ongoing.

FINANCIAL_METRICS

29,637,507 rows | One row per cluster. Joins 1:1 with TITLE_CANONICALS on CLUSTER_ID. Use the CLUSTER_FULL view for a single enriched row.

Column	Type	Coverage	Description
CLUSTER_ID	INT	100%	FK to TITLE_CANONICALS
PRICE_MEDIAN	FLOAT	99%	Median price across products in cluster
PRICE_MIN	FLOAT	99%	Minimum observed price
PRICE_MAX	FLOAT	99%	Maximum observed price
SELLER_COUNT	INT	100%	Distinct sellers in this cluster
CPC	FLOAT	17%	Cost-per-click estimate (USD)
COMPETITION	FLOAT	17%	Google Ads competition score (0-1)
DEMAND_SCORE	FLOAT	21%	Aggregated demand signal from matched keywords
OPPORTUNITY_SCORE	FLOAT	21%	Composite: demand relative to supply and competition
OPPORTUNITY_TIERS	VARIANT	23%	High / Medium / Low tier labels
INTENT_VOLUME_SCORE	NUMBER	23%	Composite intent volume score
OPPORTUNITY_GAP_SCORE	FLOAT	23%	Demand-supply gap score
PRICE_MEDIAN_13W	FLOAT	4.7%	13-week rolling median price
PRICE_MOMENTUM	FLOAT	4.7%	Price momentum signal
DEMAND_VELOCITY	FLOAT	0.9%	Rate of demand change

KEYWORD_FINANCIALS

53,612,008 rows | Per-keyword financial data. Includes original 22.3M intent keywords plus 30.5M voice-search questions.

Column	Type	Description
KID	INT	Unique keyword ID
KW	VARCHAR	Keyword or question text
TYPE	INT	1=seed, 2=canonical, 3=intent/question
CID	INT	Cluster ID (FK to TITLE_CANONICALS)
COMPETITION_INDEX	FLOAT	Google Ads competition (0-100)
COMPETITION_LEVEL	VARCHAR	Low / Medium / High
CPC	FLOAT	Cost-per-click (USD)
CPC_HIGH	FLOAT	High-range CPC estimate
CPC_LOW	FLOAT	Low-range CPC estimate
SV	INT	Monthly search volume

PRODUCT_BRIDGE

361,388,917 rows | Maps every individual product to its cluster neighborhood.

Column	Type	Description
PRODUCT_ID	INT	Unique product identifier
CLUSTER_ID	INT	FK to TITLE_CANONICALS

SHOPS

714,735 rows | Active, quality-filtered online stores with full enrichment. Screened from 1.5M+ stores; 687K excluded as spam, dropshippers, or failing integrity checks.

Column	Type	Description
ID	INT	Store identifier
NAME / DOMAIN / URL	VARCHAR	Store identification
CITY / PROVINCE / COUNTRY	VARCHAR	Geographic location
PUBLISHED_PRODUCTS_COUNT	INT	Number of listed products
TOP_INTENTS	VARIANT	Array of top search intents for this store (ranked by demand)
CATEGORIES / DEMOGRAPHICS / OCCASIONS / SEASONS / USE_CASES	VARIANT	Aggregated enrichment tags
CATEGORY_SUMMARY	VARIANT	Category distribution summary
SHIPS_TO_COUNTRIES	VARIANT	Detected shipping destinations

PRICE_HISTORY_EVENTS

95,182,652 rows | Individual price changes flattened from raw price history. Covers February 2024 to present.

Column	Type	Description
PRODUCT_ID	INT	FK to PRODUCT_BRIDGE
EVENT_DATE	TIMESTAMP	When the price changed
PRICE	FLOAT	New price at this event
PRICE_CHANGE	FLOAT	Delta from previous price

Price data note: Filter to PRICE BETWEEN 1 AND 10000 for reliable analysis. The $10K+ bucket contains data quality outliers (likely prices stored in minor currency units).

WEEKLY_DIMENSIONS

147,854,202 rows | Weekly snapshots tracking how the product landscape evolves across 12 dimensions.

Column	Type	Description
DIMENSION	VARCHAR	Dimension name (categories, demographics, price_ranges, etc.)
VALUE	VARCHAR	Dimension value (e.g. "graphic tees", "women", "$50-100")
WEDNESDAY	DATE	Week ending date
CLUSTER_COUNT	INT	Clusters with this value that week
PRODUCT_COUNT	INT	Products under this value that week

12 dimensions tracked:

INTENT_LINKS / INTENT_LINK_EDGES

INTENT_LINKS: 1,617,842 rows | Parent nodes in the intent hierarchy.
INTENT_LINK_EDGES: 32,260,516 rows | Parent-child edges. Average fan-out: ~20 intents per parent.

The hierarchy enables roll-up analytics: aggregate long-tail intents under their parent seed to compute total addressable demand for a product space.

Five Enrichment Dimensions

Every enriched cluster carries structured tags accessible via ENRICHMENT_TAGS (VARIANT column on TITLE_CANONICALS):

Dimension	Clusters Tagged	Unique Values	What It Captures
categories	3,472,377	968,390+	Fine-grained product categories far beyond any standard taxonomy
use_cases	3,475,262	521,414+	How the product is used (party dressing, vacation wear, home office)
occasions	3,469,541	111,229+	When/why someone buys it (weddings, date nights, back to school)
demographics	3,475,423	69,842+	Who buys it (women, young adults, pet owners, gamers, brides)
seasons	3,466,272	2,882	Seasonal relevance (summer, winter, holidays, year-round)

Financial Signals

ENRICHMENT_TAGS also contains financial dimension arrays on clusters that have them:

Sample: Market Sizing

How big is the market for a given keyword?

SQL
SELECT ip.KW, ip.SV,
                        SUM(ipm.CLUSTER_SIZE) AS total_products,
                        COUNT(DISTINCT ipm.CLUSTER_ID) AS competing_clusters,
                        AVG(ipm.SCORE) AS avg_relevance
                        FROM INTENT_PRODUCTS ip
                        JOIN INTENT_PRODUCT_MATCHES ipm ON ip.KEYWORD_ID = ipm.KEYWORD_ID
                        WHERE ip.KW = 'wireless earbuds'
                        GROUP BY 1, 2;

Sample: Demand-Supply Gap Analysis

High-demand keywords with few competing products.

SQL
SELECT ip.KW, ip.SV,
                        COUNT(DISTINCT ipm.CLUSTER_ID) AS supply_clusters,
                        SUM(ipm.CLUSTER_SIZE) AS supply_products,
                        ip.SV / NULLIF(SUM(ipm.CLUSTER_SIZE), 0) AS demand_per_product
                        FROM INTENT_PRODUCTS ip
                        JOIN INTENT_PRODUCT_MATCHES ipm ON ip.KEYWORD_ID = ipm.KEYWORD_ID
                        WHERE ip.SV > 1000
                        GROUP BY 1, 2
                        HAVING supply_products < 100
                        ORDER BY demand_per_product DESC
                        LIMIT 50;

Sample: Competitor Research

Which stores dominate a given category?

SQL
SELECT s.DOMAIN, s.COUNTRY, s.PUBLISHED_PRODUCTS_COUNT,
                        s.ENRICHED_CLUSTERS,
                        ARRAY_SIZE(s.TOP_INTENTS) AS intent_breadth
                        FROM SHOPS s,
                        LATERAL FLATTEN(INPUT => s.CATEGORIES) c
                        WHERE c.VALUE::STRING = 'pet supplies'
                        ORDER BY s.ENRICHED_CLUSTERS DESC
                        LIMIT 25;

Sample: Price Trend Analysis

How have prices moved in a category over 2 years?

SQL
SELECT DATE_TRUNC('MONTH', phe.EVENT_DATE) AS month,
                        AVG(phe.PRICE) AS avg_price,
                        MEDIAN(phe.PRICE) AS median_price,
                        COUNT(DISTINCT phe.PRODUCT_ID) AS products_tracked
                        FROM TITLE_CANONICALS tc
                        JOIN PRODUCT_BRIDGE pb ON tc.CLUSTER_ID = pb.CLUSTER_ID
                        JOIN PRICE_HISTORY_EVENTS phe ON pb.PRODUCT_ID = phe.PRODUCT_ID
                        WHERE tc.ENRICHMENT_TAGS:categories[0]::STRING = 'wireless earbuds'
                        AND phe.PRICE BETWEEN 1 AND 10000
                        GROUP BY 1
                        ORDER BY 1;

Sample: Seasonal Demand Mapping

Which product categories are seasonal?

SQL
SELECT cat.VALUE::STRING AS category,
                        sea.VALUE::STRING AS season,
                        COUNT(*) AS cluster_count
                        FROM TITLE_CANONICALS,
                        LATERAL FLATTEN(INPUT => ENRICHMENT_TAGS:categories) cat,
                        LATERAL FLATTEN(INPUT => ENRICHMENT_TAGS:seasons) sea
                        WHERE ENRICHMENT_TAGS IS NOT NULL
                        GROUP BY 1, 2
                        HAVING cluster_count > 50
                        ORDER BY 1, 3 DESC;

VARIANT Field Reference

Several columns use Snowflake's VARIANT type. Common access patterns:

VARIANT access patterns
-- Array access
                        ENRICHMENT_TAGS:categories[0]::STRING          -- first category
                        ARRAY_SIZE(ENRICHMENT_TAGS:categories)         -- category count
                        TOP_INTENTS[0]:kw::STRING                      -- first intent keyword
                        TOP_INTENTS[0]:sv::INT                         -- first intent's search volume

                        -- Flatten for aggregation
                        LATERAL FLATTEN(INPUT => ENRICHMENT_TAGS:categories)
                        LATERAL FLATTEN(INPUT => TOP_INTENTS)
                        LATERAL FLATTEN(INPUT => SHIPS_TO_COUNTRIES)

                        -- Filtering
                        WHERE ARRAY_CONTAINS('summer'::VARIANT, ENRICHMENT_TAGS:seasons)
                        WHERE ENRICHMENT_TAGS:categories[0]::STRING = 'graphic tees'

Data Quality Notes

Price history outliers: The $10K+ bucket (32M events) contains data quality issues. Always filter PRICE BETWEEN 1 AND 10000 for reliable analysis.

Enrichment coverage: 12.2% of clusters are enriched (3.6M of 29.6M). These cover the highest-priority neighborhoods representing 192M of 372M total products. Enrichment is ongoing and expanding weekly.

Search volume: 88.6% of keywords have SV=0 (expected for long-tail intents). Use PARENT_SV for ranking those keywords. Combined, 82% of all keywords carry some demand signal.

Store quality: 687K stores were excluded as spam, dropshippers, or inactive. The 714K in the SHOPS table passed all quality filters. We screen over 1.5 million stores to produce this filtered set.

Tag normalization: Enrichment tags are free text, not from a controlled vocabulary. Some dimensions have synonyms (e.g., "all year", "year-round", "all seasons"). Group or normalize before aggregating.

Refresh Cadence

The dataset refreshes on a weekly cadence. Dimension snapshots, price tracking, financial metrics, and cluster enrichment are all updated weekly. New products, stores, and keywords enter the graph as they are discovered and processed.