Weather and climate, news and events, stocks and commodities, macroeconomics. RootCause reads the entities, locations and time grain in your ontology, then joins external context onto them on the same three keys.
| store_idString | locationCategory | fiscal_weekDateTime | + temp_fNumber99% join coverage | + energy_priceNumber97% join coverage | + cpi_yoyNumber100% geography coverage | + event_riskCategory93% entity coverage |
|---|---|---|---|---|---|---|
| Store 147 | Denver | 2026-W24 | 96 | 41.20 | 3.1 | elevated |
| Store 202 | Austin | 2026-W24 | 112 | 38.75 | 2.8 | low |
| Store 314 | Seattle | 2026-W24 | 92 | 33.10 | 3.4 | low |
Every appended column joins on the keys the ontology already resolved: entity, time and location.
Hourly observations resolve to each site, then roll up to the weekly grain your sales already use.
Weather data is easy to buy. Making it usable takes a quarter. Your sales are weekly by store, the feed is hourly by station, the nearest station is eleven miles from the store, and the two tables share no key.
Every join is a decision. Which station represents this site. How hourly readings roll up to a fiscal week. Which region a national indicator belongs to. Made in a notebook, those decisions stay in that notebook, and the next team makes them again.
RootCause maintains these sources and resolves them against your ontology on entity, time and location. Station to site. Hourly to weekly. Country to region. Ticker to supplier.
Alignment happens once, centrally, and every later analysis inherits it.
One table counts weeks by store. The other counts hours by station. Every question that touches both starts with somebody deciding which station stands in for which store.
Each domain lands as typed columns on entities you already have, resolved on entity, time and location.
Observed history and forward outlook, resolved to the coordinates of each site and rolled to the grain your sales already use.
temp_fNumberprecip_mmNumbersevere_eventCategoryUsed for Demand planning, logistics, energy use, footfall.
Events resolved to the companies and places in your ontology, so a story about a supplier attaches to the supplier record.
event_typeCategoryentity_idStringpublished_atDateTimeUsed for Supplier risk, disruption monitoring, competitive tracking.
Price and volume series aligned to the inputs and counterparties already present in your own tables.
close_priceNumbervolumeNumberfx_rateNumberUsed for Input costs, margin pressure, supplier health.
Standard indicators mapped onto the geographies your ontology already carries, at the hierarchy you report on.
cpi_yoyNumberpolicy_rateNumberregion_codeCategoryUsed for Demand outlook, pricing, market entry.
Buying a feed leaves you with a feed. You still resolve its geography to yours, reconcile its calendar to yours, choose how to move between grains, and rebuild that logic in every downstream model.
Foundational Ontologies land on the entities you already have. The resolution decisions are made, documented and versioned. When an analyst or an agent asks whether last quarter's shortfall tracked with regional temperature, the join already exists.
| Buying a feed | Foundational Ontologies | |
|---|---|---|
| Contracting | Per vendor, per domain | Included |
| Geographic resolution | Your team builds it | Resolved to your sites |
| Temporal alignment | Your team builds it | Resolved to your grain |
| Entity linkage | Usually absent | Resolved to your records |
| Availability to agents | Needs custom tooling | Immediate, through the ontology |
Pick a site. Its weekly sales and the temperature observed at its resolved station are drawn on the same axis, because they are already stored against the same entity, week and place.
KDEN resolved to Store 147
station 11 mi from the site
Hourly observations rolled to fiscal week
the calendar your reports already use
Grid cell resolved to Mountain
the hierarchy your ontology already uses
The band is what a model expected from internal history alone. The dashed line is what store 147 actually sold. Week 24 sits outside the band and stays unexplained until temperature is joined to the same rows.
Every source carries provenance, license terms, update cadence and version history. Answers that draw on external data cite which source and which version produced them, the same way answers over your own data do.
A model built only on internal data can describe what your business did. Add climate, commodity and macro context and the conditions you never controlled enter the same graph as the levers you do control.
Tell us which external driver you think moves your numbers. We will join it to your data and show you the series.