Electronic Components Supply Chain & Risk Dataset

It is a component-level risk panel that answers one question per row: how exposed is this part? Each of the 791 records is keyed by manufacturer part number and stacks four layers on top of that key. An identity layer names the manufacturer and headquarters country. A catalog layer adds an Octopart-style category label, a free-text technical description and datasheet references. A commerce layer lists distributor offers with name, stock, price and SKU. And a risk layer scores the part: a boolean China-origin flag, a confidence score attached to that origin judgment (observed running roughly 0.7 to 0.95), a 0-to-1 risk score, and the reason codes behind it.

The risk layer is what makes the dataset more than another parts catalog. Three reason codes do the explaining: chinese_origin, critical_category and limited_suppliers. Because every flagged row carries both a score and the codes that produced it, the verdicts are auditable - a sourcing team can disagree with a score and see exactly why it was assigned.

The fourth layer is depth. Each record nests a full Nexar/Octopart enrichment block: manufacturer homepage, best datasheet link, and a sellers array of companies with offers carrying SKUs, price points and inventory levels. One flat JSON file holds all four layers, so a single read gets you from part number to priced, stocked, risk-scored reality.

Identity, commerce and risk layers for one record, then two comparison rows from elsewhere in the score range - exactly as they land in a delivery:

mpn=ESP8266EX   mfr=Espressif Systems   country=China   is_chinese=true
category=RF Receivers, Transceivers   confidence=0.95
risk_score=0.60   risk_factors=[chinese_origin]

suppliers=[{name=DigiKey, stock=6433, price=$1.60, sku=1904-1001-1-ND}, ...]
nexar_data={bestDatasheet: url, sellers[] company=DigiKey, offers[] inventoryLevel, ...}

mpn=DFR1063     mfr=DFRobot             risk_score=0.25  risk_factors=[critical_category]
mpn=GPY         mfr=Pycom               risk_score=0.10  risk_factors=[limited_suppliers]

Three rows, three different triggers. The Espressif Wi-Fi SoC flags purely on origin, with the classifier stating 0.95 confidence. The DFRobot part trips critical_category without an origin flag. The Pycom module sits at the low end because its trigger is scarcity - limited_suppliers - not geography. Scores observed in sampled rows run 0.1 to 0.6, with nulls where nothing fired.

Twelve documented fields split across the same four layers: identity (mpn, manufacturer, manufacturer_country), catalog (category, description, datasheets), commerce (suppliers, with the nexar_data block deepening it into per-seller offers) and risk (is_chinese, confidence_score, risk_score, risk_factors).

The dictionary below is the verified core. The nested seller detail inside nexar_data flattens into pipeline-ready columns - those fold out on request rather than being promised blind.

Geography: global distributor coverage - DigiKey, Mouser, TME, Conrad, Win Source and independent brokers all appear in the suppliers arrays - against manufacturers headquartered in the US, EU, Japan and China.

Manufacturer concentration skews analog and embedded. Analog Devices leads with 195 rows, Texas Instruments follows at 184, STMicroelectronics at 100, then Espressif Systems (63), GigaDevice (48), Microchip (41), Seeed Studio (28) and Adafruit Industries (22). Eight manufacturers account for the large majority of the panel, which makes it a focused instrument for analog/embedded sourcing risk rather than a shallow index of everything.

Origin mix: only 14 of 791 rows carry the China-origin flag - roughly 1.8 percent - even though Chinese-headquartered makers such as Espressif and GigaDevice hold over 100 rows between them. That gap between headquarters country and flag rate is itself informative about how the classifier weights evidence.

Temporal: a point-in-time snapshot of 791 records. Prices and stock are as-of-collection values, so treat them as structure rather than live quotes - cadence accrues the time dimension on your side.

Granularity: one row per MPN with nested per-seller offer detail. No rollups, no category averages - every downstream cut aggregates from the atomic unit.

Set against the wider Datadory catalog - 1,744 datasets, average quality score 7.81 - this slice scores 7/10: exceptional structural richness per row, held back mainly by its static snapshot nature and undocumented scoring methodology.

Pick the channel your team already works in: REST endpoints for live lookups against a specific MPN, flat files sized for overnight warehouse loads, or a direct pipe into Snowflake, BigQuery or Redshift. Cadence is yours to set - daily catches design-cycle churn, weekly suffices for quarterly risk reviews, hourly suits monitoring dashboards.

Every delivery ships with the full field dictionary, sample rows for validation, and a schema that holds steady between refreshes. The nested nexar_data block arrives pre-flattened if you want it relational, or intact if your stack prefers documents.

• Supply-chain risk managers screen bills of materials for China-origin exposure part by part. Because every flag ships with reason codes and a confidence score, screening results survive audit - the explanation travels with the verdict.
• Component and sourcing engineers run second-source checks before design freeze: how many distributors carry the part, at what price and stock levels, and whether limited_suppliers fires. A part with one offer object is a different procurement decision than one with six.
• Competitive intelligence teams map manufacturer concentration per category - who owns the supply base, where second sources are missing, and which categories concentrate hardest.
• Data scientists use the flag-plus-confidence structure as labeled training material for origin-classification models, and the nested seller arrays as ground truth for pricing and inventory estimation.
• Investors and analysts quantify component-level sourcing exposure beneath the company-level disclosure layer, particularly for hardware businesses selling into categories flagged critical.

More on how these teams apply the slice lives on the data scientists, competitive intel product teams and investors & quants pages.

Inside this industry slice, three datasets answer three different questions. The Nexar API (Octopart) is the live commerce engine - multi-distributor offers, lifecycle status and specs across the full Octopart database, but it carries no risk scoring of its own. FindChips' Aggregated Price & Inventory Search consolidates current price and stock across authorized and independent distributors - breadth of quote coverage, again without a risk layer. The TME catalog contributes 1.5 million products with documented parameters and files from one major European distributor.

This dataset is the fourth leg: judgment. It is the slice that says not just what a part costs and who stocks it, but whether sourcing it carries geographic or availability risk - and shows its work via reason codes. The practical stack joins this risk layer onto Nexar or FindChips pricing by MPN, which turns three separate feeds into one answerable question: for a given part, what does it cost, who has it, and how risky is depending on it.

Three things, stated up front.

First, this is a snapshot, not a stream. The 791 records reflect one collection pass - prices and stock are as-of-collection values. Teams needing a time dimension start a cadence immediately and let repeated captures build the curve.

Second, the scoring methodology is undocumented upstream. No published methodology explains how risk_score, confidence_score or the reason codes are computed, and the snapshot date of the underlying price data is unstated. Treat the risk layer as a strong prior to be validated against your own sourcing rules, not as a regulatory determination - we say this plainly because a risk feed you cannot interrogate is worse than none.

Third, the verified core dictionary covers twelve fields; the deeper seller-level detail inside nexar_data is confirmed and flattened when we cut your sample, which is also where column naming gets locked down for your pipeline. Name the MPNs, manufacturers or categories you care about and the sample comes back shaped to them.