All documentsSchema · Version 2.4

Data dictionary

Every table and column, with conventions, identifiers and value lists.

Field guide for the Marketing stack, Hiring, Intent, Announcements-and-scale, Filings-and-records and Market-series products. Method: METHODOLOGY.md. Sourcing: SOURCING.md. Licence notices: LICENCE_NOTICES.md.

Conventions

  • Files are CSV, UTF-8, one header row (JSONL through the API). Lists and objects are JSON in a single cell.
  • Times are UTC, ISO 8601. day is a closed UTC day; week_start is the Monday of a Monday-to-Sunday UTC week.
  • Every file carries company_id (stable), company, and the identifiers of the listing when the company or its parent is listed: ticker, exchange (the venue code), mic (ISO 10383), isin, lei, figi (the share-class FIGI). Join across products on company_id; join to market data on isin, figi or ticker + mic. Identifiers come from public reference data: the global identifier registries, open reference records and the US securities regulator's ticker list (METHODOLOGY.md section 4.1); an ISIN or LEI whose check digit fails is never stored.
  • Daily and weekly rows are written once and never changed. reconstructed=true marks a period written more than seven days after it closed.
  • Only companies in the Fokals company index are delivered, and only those with a website read in English, on its homepage or in the English version it declares: a company with no such website is in no file, and a website with neither is in no file of websites (METHODOLOGY.md section 4.2). listed_securities keeps every listing; a listing of a company that is left out carries no company.
  • The history datasets (the Filings-and-records product, company_funding, company_traffic_history and any other dataset whose section below names provenance) are point in time as published by the source. Every row carries the source's own time (a day filed, a month, a moment observed) and public_from, when the source made the row public: its own release time where it states one, the end of its stated day where it gives a day, otherwise a conservative delay after its own date, never earlier than the source shows and never later than recorded_at. Each dataset below says how. Every row also carries recorded_at (when the row was stored, never changed), provenance (live: stored at the time; backfill: read later from the source's history; each dataset below says within how long) and version (the reader or rule that wrote it). Rows are written once and never revised: a correction, an amendment or a source's revision is a row of its own. To backtest on what anyone could have read on a day, keep the rows whose public_from is before the end of it: history read after the fact keeps the time its source published it, and is marked backfill. To rebuild what Fokals held on a day, keep the rows with recorded_at on or before it. Feeds through the API page on recorded_at, so a row read late from a source's history still reaches a subscriber.
  • A change is only ever worked out within one series: one source read one way. A series' first reading sets its baseline and is never a change, a reading that failed is unknown and changes nothing, and a removal needs two readings in a row without the item.

Marketing stack

company_technologies

One row per company website and technology seen on it in the period.

ColumnMeaning
domainThe company website.
technologyTechnology id (stable).
technology_name, technology_categoryReadable name and category.
technology_cataloguefokals when the signature is one Fokals writes and tests, open when it comes from the open-source catalogue (Methodology 5.1).
first_seen_at, last_seen_atFirst and latest reading that saw it.
missing_sinceSet while absent from recent readings but not yet counted as removed.
seen_viaViews that saw it: html (homepage), container (tag-manager container), network (browser read), header (response headers), dns (DNS records), robots (robots.txt), file (a file at the site's root: llms.txt).
account_idsPublic account ids found with it (pixel ids, analytics properties, containers).

company_tech_events

Every change seen between two readings of a company website. A site's first reading writes none.

ColumnMeaning
domainThe company website the change was seen on.
observed_atReading that saw the change.
categorytechnology, technology_id (a new account id under a technology already there), dns, platform (commerce or content platform replaced), market, language, currency, app, social, page, promo; from 28 October 2026 pricing (the site's pricing page), label (a website label that moved under the same label version) and identity (the homepage's title or description reworded).
keyWhat changed: a technology id, technology:account, a country code, a language code, a currency, ios:<id> or android:<package>, a social platform, a page key, announcement. For pricing: price_raised, price_cut, plan_added, plan_removed, free_plan_added, free_plan_removed, usage_pricing_added, usage_pricing_removed, per_seat_added, per_seat_removed, ai_addon_added, ai_addon_removed, contact_sales_only. For label: the label (sales_motion, target_customer_size, pricing_model, price_tier, offering, business_model, growth_stage, enterprise_focus, compliance_badges, ai_product, has_free_trial, has_api, sells_online). For identity: title or description.
changeadded, removed or changed.
before, afterValues before and after (JSON). A price change carries {plan, amount, currency, period, per} on each side (period: month, year, week, day, one_time or empty; per: seat, unit or empty); a plan added or removed {plan, price} on one side; a pricing flag or a yes-or-no label true or false; a title or description its text, at most 300 characters. All pricing, label and identity changes have change changed.
technology_name, technology_categoryFor technology and DNS events.
technology_catalogueFor technology and DNS events: fokals or open, as in company_technologies.
excludedtrue for a change that was not the company's doing: a vendor's update reaching every website that runs it, a hosted platform's change of CDN or web server, or a change made by a signature since corrected (Methodology 5.4 and 14). Such a change stays in the history and is left out of every count and list of changes, including the changes of the API and the MCP. false for every other change.

company_site_facts

The latest reading of each company website in the period.

ColumnMeaning
last_read_atTime of the reading.
read_urlThe address read when it is not the homepage: the site's English version, declared by the homepage. Empty when the homepage itself was read.
home_languageThe language of the homepage when the site was read in its English version.
title, langPage title and declared language.
marketsCountry codes the site declares alternate versions for.
languagesLanguage codes of the page and its alternates.
currenciesCurrencies of prices marked up on the page.
social_accountsThe company's own accounts: {platform: handle}.
apps{ios: [app ids], android: [package names]}.
key_pages{key: url} for pricing, demo, free_trial, contact_sales, login, careers, investors, newsroom, partners, affiliate_program, store_locator, subscription, wholesale, gift_cards, blog, api_docs, status, and from 28 October 2026 signup (a sign-up for the product), enterprise, trust (a security or trust centre) and checkout (the site's own cart or checkout).
schema_typesStructured-data types marked up on the page.
latest_promotion{text, offers, discount} from the announcement bar.

Website labels

The labels each website carries in the dashboard and the data browser (site_state.labels, with the version of the label definitions in label_version; sites-v1 until the site's next read). They are not yet a column of company_site_facts. Each label has its probability under probabilities; an empty choice or score means the label is left unset.

  • Choices: industry (21 industries), business_model (b2b, b2c, both), offering (physical_products, software_saas, services, marketplace, media_content, financial_products, healthcare, travel_hospitality, education, nonprofit_public, other), sales_motion (self_serve, sales_led, both, offline), price_tier (budget, mid_market, premium, luxury, not_stated), target_customer_size (consumers, smb, mid_market, enterprise, mixed), pricing_model (free, freemium, subscription, usage_based, one_time, quote_only, not_stated), geography_scope (local, national, multi_country, global), primary_persona (marketing, sales, engineering_it, finance_ops, hr_people, executives, consumers, clinicians, other), growth_stage (startup, scaleup, established, incumbent, unknown).
  • Scores (0–3, one of four described levels): ecommerce_maturity, tech_sophistication, content_intensity.
  • Flags (true/false): sells_online, physical_locations, subscription, ai_product, sustainability, enterprise_focus, international, has_api, has_mobile_app, has_free_trial, has_demo_request, compliance_badges, has_partner_program, sells_to_government, regulated_industry.

sites-v1 rows carry the first five choices and the first seven flags only, with the same meanings.

Accuracy: website labels are 97–98% accurate, measured field by field on graded random samples (methodology, section 6.5).

vendor_scorecards

One row per technology, calendar month and segment, for the technologies companies take up and drop (a website's stance towards AI crawlers and the files and badges it publishes for machines have none). Written once, in the first days of the next month, and never changed. A period delivers the months whose first day falls in it. It names no company, so it carries no identifiers.

ColumnMeaning
monthThe first day of the month (UTC).
technologyTechnology id, as in company_technologies.
technology_name, technology_category, technology_catalogueReadable name, category and list (fokals or open).
segment_kind, segmentall (segment all: every company); size (the band of stated employees: 1-10, 11-50, 51-200, 201-1000, 1001-5000, 5000+); industry (the website-label industry id); growth_stage (the website label: startup, scaleup, established, incumbent, unknown); country (the head office's country where it is known, otherwise the main market the company's website serves). Industry and growth stage are the companies' labels as they stood when the month was written. A company with no value for a kind is in no segment of it.
vendor_versionThe version of the scorecard definitions the row was written under. A period delivers one version: vendor-v1 until every month is written again under vendor-v2.
baseCompanies running the technology at the start of the month (or, for a website first read during the month, when it was first read). A company counts once, whichever of its websites runs it. From vendor-v2, for a technology a hosting provider serves websites through, the companies that chose it.
served_through_hostFrom vendor-v2, for a technology a hosting provider serves websites through (a CDN, what a CDN adds to the pages, a web server): companies running it at the start of the month only on websites their hosting provider serves through its own network. Counted apart from base, and in no gained, lost or churn_rate. Empty for every other technology, and in vendor-v1 rows.
gained, lost, netCompanies that took it up in the month (none of their websites ran it at the start, one did at the end); companies that dropped it (one ran it at the start, none at the end); gained less lost. A tool added and removed in the same month is neither.
churn_ratelost / base, 0 to 1; empty without a base.
gained_from{technology id: companies}: rivals that the companies taking it up dropped within 30 days of their move.
lost_to{technology id: companies}: rivals that the companies dropping it took up within 30 days of their move. A rival is a technology of the same category that serves the same purpose, from another vendor.
partialtrue for a month not watched from its first day: the first month a technology was watched, or a month near a change of its catalogue entry, when its websites were read with the new entry for the first time.
reconstructed, written_attrue for a month written after the 9th day of the next month; when the row was written.

The API's /technologies/{id}/scorecard adds the last 30 days, counted on request.

company_agent_readiness

One row per company: how ready it is for AI agents, scored weekly from its main website. The row is rewritten only when the score or a part changes. It stands from changed_at on, and a period delivers the rows standing in it (set before the period ended).

ColumnMeaning
domainThe website the score was read from.
ready_versionThe version of the score's definitions the row was written under.
scoreAgent readiness, 0 to 100: the sum of the four parts.
find, price, buy, actThe four parts, 0 to 25 each. find: AI search and assistants can find the company and what it offers. price: an agent can learn its prices. buy: an agent can buy from it. act: an agent can act through it.
evidencePer part, what stands behind it: {find: [...], price: [...], buy: [...], act: [...]}, each item {item, text, first_seen}: item is what was found as a stable id (a technology, a key page, a currency, a structured-data type, a label), text the same in plain words, first_seen when it was first seen on the website (empty where not dated).
industry, countryThe website-label industry id, and the head office's country where it is known, otherwise the main market the website serves.
computed_at, changed_atWhen the row was last written, and when the score or a part last changed.

company_agent_readiness_changes

One row per company each time its agent readiness score moves, dated when the move was seen. Written once and never changed. A company's first score sets its baseline and is not a change, and neither is the first score under a new version.

ColumnMeaning
observed_atWhen the weekly reading saw the move.
ready_versionAs in company_agent_readiness.
score_before, score_afterThe score before and after the move, 0 to 100.
score_changescore_after less score_before.

Domain records

What the public records of each company website's domain say of its mail and its registration (domain_records), delivered through the API (domain_records on /companies/{id}, and email_provider, email_security and domain_registered_on on each /companies row), the MCP (the domain part of a company profile, and the same three fields on each company row) and the dashboard. The mail facts are read at every reading of the website and change only when the records do; the registration is read when the domain is first held, 30 days before its stated expiry, and at least once a year. Never the registrant, a contact or a report address. A company's facts are those of its main website, held for a website on its own registrable domain.

FieldMeaning
email_providerThe mail service the domain receives its mail through: google_workspace, microsoft_365, zoho_mail, other (another service, the company's own servers included) or none (the domain takes no mail).
email_gatewayA security gateway in front of the mail service: Proofpoint, Mimecast or Barracuda; empty where there is none. Behind one, email_provider is the service the domain lets send its mail, else other.
email_securityThe domain's DMARC policy for mail that claims to come from it and fails the checks: reject, quarantine, monitoring (p=none: reports only) or none (no policy).
dmarc_reports_requestedWhether the policy asks for reports. The report addresses are never held.
dmarc_applies_to_pctThe share of failing mail the policy applies to, given only when less than all.
spfThe domain's SPF record by its rule for any other sender: hard_fail (-all), soft_fail (~all), neutral (?all), pass_all (+all), redirect (follows another domain's record), no_all, invalid (more than one record) or none.
email_changedThe day the mail facts last changed.
registered_on, expires_onWhen the domain was first registered and when its current registration ends, from the registry's public record. Some registries state neither.
registry_changed_onWhen the registry's record last changed.
registrarThe company the domain is registered through, as the registry names it.
registration_checkedWhen the registry's record was last read.

company_tech_history

Technology history: the technologies on each company website month by month, from open web measurement crawls that test website home pages each month from desktop and mobile clients. One row per company website and calendar month the crawl tested the website's home page. A series of its own, never compared with company_technologies or company_tech_events. A month the crawl did not test the website has no row: that is no reading, not an empty one. Credit the source as LICENCE_NOTICES.md gives it.

ColumnMeaning
monthThe crawl's month, as its first day. The period of an export filters on it.
domainThe company website.
technology_idsThe technologies detected that month, as catalogue ids (as technology in company_technologies), sorted.
unmapped_namesThe names detected that the catalogue holds no entry for, exactly as detected, sorted. Never dropped.
clientsThe clients that tested the website that month: 1 desktop, 2 mobile, 3 both.
tested_onThe earliest test day of the month's tests of the website; empty when the test gave none.
provenancelive: the month was read within 45 days of its end. backfill: read later, from the crawl's history.
versionThe reader that stored the row (ha-import-v1).
map_version, catalogue_versionThe rule that maps detected names to catalogue ids (ha-map-v1) and the technology catalogue's version, applied when the row is read: the stored names never change, an id can be added.
public_fromWhen the source made this record public: its own release time where it states one, the end of its stated day where it gives a day, otherwise a conservative delay after its own date. Never earlier than the source shows and never later than recorded_at. The time to backtest on. Here: never before the crawl released the month's results, by a conservative delay after the month, since the source states no release time for each crawl.
recorded_atWhen the row was stored. Never changed.

company_tech_history_events

Technology adoption history: each technology a company website took up or dropped in the monthly crawl series. One row per company website, technology and change. A change counts only when two tested months in a row show it, and is placed between the last tested month of the old state and the first of the new. A website's first tested month, its first month after a break in the crawl's method or a long gap in testing, and a month in which a technology's count moves across all websites because the crawl's own rules changed set a baseline and are no change. A website entering or leaving the crawl's sample is never a change. A website added to our list later has its past months read for it in a round once enough have joined (backfill): its first tested month is a baseline as for any website, and it never changes another website's series. Only the current rule version is delivered: a new version works the whole series out again from the stored readings.

ColumnMeaning
monthThe first month of the new state, as its first day. The period of an export filters on it.
domainThe company website.
since_monthThe last tested month of the old state: the change happened between since_month and month.
confirmed_monthThe next tested month after month, which showed the new state again and confirmed the change.
changeadded or removed.
technology_id, technology_nameThe catalogue id and name of the technology; empty when the catalogue holds no entry for the name detected.
source_nameThe technology's name exactly as detected. Always present.
low_confidencetrue for a removal of an analytics, advertising, tag management, marketing automation or retargeting tool in a month a consent tool appeared on the website, and for any change in a month the website was tested from a different set of clients than the month before.
provenanceThat of the reading that confirmed the change (live or backfill).
versionThe change rule that wrote the row (ha-changes-v1).
public_fromWhen the source made this record public: its own release time where it states one, the end of its stated day where it gives a day, otherwise a conservative delay after its own date. Never earlier than the source shows and never later than recorded_at. The time to backtest on. Here: when every tested month from since_month to confirmed_month, which the change rests on, was public.
recorded_atWhen the row was stored, once its confirming month was read. Never changed.

Hiring

job_postings

One row per posting open at any time in the period. Descriptions are not delivered.

ColumnMeaning
posting_idStable id.
board_platformThe job board's system (greenhouse, lever, ashby, workday, recruitee, personio, workable, teamtailor, jsonld for job markup on the company's own careers page).
urlThe public posting.
title, department, teamAs published.
locations[{city, region, country}].
countryFirst location's country (ISO 3166-1 alpha-2).
work_modeonsite, hybrid, remote or empty.
employment_typefull_time, part_time, contract, temporary, internship or empty.
posted_atPosting date as the board gives it.
first_seen_at, last_seen_at, closed_atWhen we first and last saw it, and when it was closed (absent from two complete reads at least 20 hours apart, or not seen on its board for 14 days: then dated the day after it was last seen).
found_on_first_readAlready on the board at our first read of it: first_seen_at is not its opening date.
salary_min, salary_max, salary_currency, salary_periodPay as stated; period year, month, week, day or hour.
salary_usd_annual_min, salary_usd_annual_maxPay on one yearly US-dollar scale at the latest ECB reference rate; empty when outside 5,000–2,000,000.
experience_years_minMinimum years of experience asked for.
degreenone, bachelor, master, doctorate or empty.
visa_sponsorshipoffered, not_offered or empty.
languagesLanguages required (ISO 639-1).
technologiesSoftware named in the posting (catalogue ids); the company's own products are left out.
label_versionThe version of the label definitions the posting was labelled under. jobs-v1 for postings labelled before 23 September 2026 and not yet labelled again.
labels{job_function, seniority, remote_policy, contract_type, technical_depth, people_manager, ai_role, paid_media, selects_tools, international, new_initiative, urgent, team_build, replaces_vendor, budget_owner, tool_*}. An empty choice or score means the label is left unset. jobs-v1 rows carry the first two choices and the first six flags only.
sales_factsFor a posting that states them, what the employer says about the deal (JSON): ote and base as written ({min, max, currency, period}) with ote_usd and base_usd (yearly US-dollar midpoints), base_share (0.5 for a 50/50 split), uncapped, variable_pay, quota ({amount, currency, period}) and quota_usd, ramp_months, lead_source (inbound, outbound, mixed), inbound_share (percent), deal_size and deal_size_usd, sales_cycle_months. Empty when the posting states none. Advertised, not realised.
sales_labels, sales_label_versionFor a posting in a sales function, the sales labels below and the version of their definitions; empty otherwise.

job_function (26): software_engineering, hardware_engineering, data_analytics, data_science_ml, product_management, design, marketing_brand, marketing_performance, marketing_content_social, sales, sales_development, customer_success_support, operations, supply_chain_logistics, manufacturing_production, finance_accounting, legal_compliance, people_hr, it_security, research_science, healthcare_clinical, retail_store, hospitality_food, education_training, executive_leadership, other.

seniority (9): intern, entry, mid, senior, lead_principal, manager, director, vp, c_level.

remote_policy: remote, hybrid, onsite, unspecified. contract_type: permanent, contract, temporary, internship, unspecified.

technical_depth (0–3): 0 no technical skills required; 1 uses technical tools as a user; 2 builds, configures or administers technical systems; 3 designs or leads complex technical systems.

Flags (true/false): people_manager, ai_role, paid_media, selects_tools, international, new_initiative, urgent (immediate start or urgent hire), team_build (one of several hires growing a team), replaces_vendor (brings agency or vendor work in-house), budget_owner.

Tool families (true when the posting names or requires the family's products): tool_salesforce, tool_hubspot, tool_google_ads, tool_meta_ads, tool_ga4, tool_adobe, tool_shopify, tool_sap, tool_workday, tool_snowflake, tool_databricks, tool_aws, tool_azure, tool_gcp, tool_figma.

Accuracy: job posting labels are 98–99% accurate, measured field by field on graded random samples (methodology, section 6.5).

Sales labels

Carried by every posting in a sales function (and by account-management titles filed under customer success and sales leaders filed under executives). Choices: role (sdr_bdr, account_executive, account_manager, sales_engineer, sales_leadership, sales_operations, partnerships_channel, other) and segment (smb, mid_market, enterprise, unspecified). Flags: quota_carrying, inbound_leads, outbound_prospecting, new_business, founding_sales. A posting keeps the version it was labelled under, named in sales_label_version.

company_hiring_daily

Per company and closed UTC day.

ColumnMeaning
open_postings, new_postings, closed_postingsOpen at the end of the day; seen opening that day (not first-read postings); closed that day.
by_function, by_seniority, by_country, by_work_modeOpen postings by label or field: {value: count}.
ai_postings, paid_media_postingsOpen postings with those flags.
tech_mentionsOpen postings naming each tool: {tool: count}.
median_salary_usdMedian of open postings' yearly US-dollar pay (midpoint of range).

company_sales_weekly

The sales organisation per company and closed week (Monday to Sunday UTC), from its own postings. Written once after the week closes. A sales posting is one whose sales labels give it any role but other; a company has a row for a week in which it had one open, opened or closed.

ColumnMeaning
week_startThe Monday.
open_sales, new_sales, closed_salesSales postings open at the week's end; opened in the week (not first-read ones); closed in it.
open_allEvery posting open at the week's end, for the sales share.
by_role, by_segment, by_countryOpen sales postings by sales-label role and segment, and by country: {value: count}.
new_countriesCountries with the company's first sales posting this week, none before.
upmarketThe company's first enterprise-segment sales posting opened this week.
median_base_usd, median_ote_usdMedian advertised base pay and on-target earnings of open sales postings, yearly US dollars, by role. Base is the stated base, or a plain salary range when no OTE is stated.
toolsSales tools named in open sales postings: {tool: count}.
quota_statedOpen sales postings that state a quota or target.
inbound_shareMedian stated share of inbound pipeline, percent, when stated.

sales_pay_benchmarks

What employers advertise for sales roles, across companies, per closed week, role and country (ALL for every country together). Only groups of at least five postings.

ColumnMeaning
week_start, role, countryThe week, the sales-label role, the posting's country or ALL.
postingsOpen sales postings in the group with a stated base or OTE.
base_p25_usd, base_median_usd, base_p75_usdQuartiles of advertised base pay, yearly US dollars.
ote_p25_usd, ote_median_usd, ote_p75_usdQuartiles of advertised on-target earnings.

posting_projects

One row per company and project its job postings state, over the postings seen on its board in the period: a project stated by several postings is one row. Facts only: the postings' text is never delivered. An expansion into the company's own country is not a project here.

ColumnMeaning
projectThe project's stable key: migration:<from>><to> (either side may be empty), first_team:<function>, compliance:<programme>, expansion:<country or region>, ai_build:<focus>.
kindmigration (a move from one system to another), first_team (the first team of a function), compliance (a programme the company is working towards), expansion (a country or region it is entering), ai_build (an AI product it is building).
detailsThe project's details with readable names (JSON): for a migration from, from_name, to, to_name (software ids as in technologies of job_postings); for a first team function, function_name (a job_function id); for compliance programme, programme_name (soc2, iso27001, hipaa, pci, fedramp, gdpr); for an expansion country, country_name (ISO 3166-1 alpha-2) or region, region_name (europe, north_america, latin_america, asia_pacific, southeast_asia, middle_east, africa, emea, mena, dach, nordics, benelux, anz, americas); for an AI build focus (agents, llm, ml, ai).
summaryThe project in plain words: "moving from Salesforce to HubSpot", "building its first data and analytics team".
postings, open_postingsPostings seen in the period that state it, and how many of them are open now.
first_seen_at, last_seen_atWhen the first of those postings was first seen, and when the last of them was last seen on its board.
posting_title, posting_idsThe title of the latest posting that states it, and the ids of every posting that does (posting_id of job_postings), latest first.
project_versionThe version of the project definitions the postings were read under.

Intent

company_intent_weekly

Per company, topic and closed week, for topics scoring at least 5.

ColumnMeaning
topic, topic_label, topic_groupOne of 69 topics in nine groups.
intent_versionThe version of the topic and score definitions the row was written under.
scoreThe company's buying intent for the topic in the week, 0–100.
surgeTrue in a week the company's score for the topic rises sharply against its own recent weeks.
signalsNumber of signals behind the score.
evidenceThe dated evidence behind the score: [{date, kind, source, weight, detail}].

company_signals

Every dated signal behind the scores.

ColumnMeaning
observed_atWhen it was seen (a posting's own date where it has one; a filing's date).
sourceWhere it was seen: site, dns, careers, filing or news (an announcement).
kindWhat was seen, as a stable id: for example tech_added, market_added, funding, news_event.
topicsThe topics the signal is tied to.
weightIts strength, 0–5.
detailWhat was seen (technology, market, posting title, amount raised, an announcement's title and link), as JSON.
excludedtrue for a signal raised by a website change that was not the company's doing (excluded in company_tech_events). Such a signal stays in the history; scores written from then on do not count it, and it is left out of every list of signals, including the API's and the MCP's. false for every other signal.

company_funding

Form D notices filed with the US securities regulator by operating companies (pooled investment funds left out). Point in time like the Filings-and-records datasets: notices of earlier years are read back from the regulator's daily indexes into the same rows, marked backfill, and no row is ever revised.

ColumnMeaning
accessionThe filing's accession number.
formD (new notice) or D/A (amendment).
filed_atFiling date.
issuer, cikIssuer's legal name and its id with the US securities regulator (CIK).
company_id, company, tickerSet when the issuer matches exactly one company in the index.
industryIndustry group as filed.
city, stateIssuer's principal place of business.
amount_offered, amount_soldUS dollars; empty when indefinite.
first_saleDate of first sale.
is_amendmentAn amendment to an earlier notice.
detailsEntity type, jurisdiction, year of incorporation, revenue range, exemptions, securities, whether the offering runs more than a year, business combination, minimum investment, amount remaining, number of investors, non-accredited investors, sales commissions and finders' fees.
public_fromWhen the source made this record public: its own release time where it states one, the end of its stated day where it gives a day, otherwise a conservative delay after its own date. Never earlier than the source shows and never later than recorded_at. The time to backtest on. Here: the end of the day filed, US Eastern time, since the regulator publishes a filing by the end of its filing day.
recorded_atWhen the notice was stored. Never changed.
provenancelive when the notice was stored within 14 days of being filed, backfill when it was read later from the filings of earlier days.
versionThe readers that stored the notice: sec-history-v1, or sec-v0 for a notice stored before versions were recorded.

company_next_weeks

The technologies each company is likely to add next. One row per company, week and technology the company did not run that week, for the companies that changed in the recent weeks. Written once a week and never changed. A prediction carries its support counts: how many companies made each change behind it and then added the technology, of how many made the change.

ColumnMeaning
week_startThe Monday of the week the prediction is as of (UTC).
next_versionThe version of the prediction definitions the row was written under.
technology, technology_name, technology_category, technology_catalogueThe technology, as in company_technologies.
probabilityThe estimated likelihood (0 to 1) that the company adds the technology in the weeks after week_start, with the support counts behind it in because.
becauseThe changes the prediction stands on: [{token, text, companies, of}], token a stable id of a change the company made (a technology added or removed, hiring in a function, a page added, a surge on a topic), text the same in plain words, companies how many companies made that change and then added the technology, of how many made the change while not running it.
reconstructedtrue for a week written more than seven days after it closed.

adoption_sequences

The changes that come before companies add a technology. One row per week, change, technology and window: as of the week, the companies that made the change (the antecedent) in one week and added the technology in a later week within the window. Written once a week and never changed; it names no company, so it carries no identifiers.

ColumnMeaning
as_ofThe Monday of the week the counts are as of.
next_versionAs in company_next_weeks.
antecedent, antecedent_textThe change, as a stable id and in plain words.
technology, technology_name, technology_category, technology_catalogueThe technology added after it.
window_weeksHow many weeks after the change the addition counts in, from 1.
supportCompanies that made the change and then added the technology within the window.
antecedent_companiesCompanies that made the change while not running the technology.
confidencesupport / antecedent_companies, 0 to 1.
base_rateThe same share for comparable companies: those that changed in a week, whatever the change.
liftconfidence / base_rate: how many times as often the companies that made the change added the technology.
reconstructedtrue for a week written more than seven days after it closed.

A change that names a website's traffic tier moving is derived from the Chrome UX Report (Google), CC BY 4.0: keep the attribution with it.

Announcements and scale

company_news

One row per announcement a company made about itself: an item on its newsroom or press page, an entry in a feed its homepage declares, or (listed companies) a Form 8-K current report.

ColumnMeaning
atThe date the company published it (published_at); when the source gives none, when we first saw it.
company_id, company, tickerThe company that published it.
sourcerss, atom, page (the company's newsroom page) or sec_8k.
urlThe original announcement or filing document.
titleAs published (for an 8-K: the form and its item labels).
event_typesZero or more of product_launch, partnership, acquisition_made, acquired_or_merged, funding_round, leadership_change, layoffs_restructuring, expansion, incident, financial_results, award_or_recognition, regulatory_or_legal, other. For an 8-K, from the item numbers it was filed under.
itemsForm 8-K item numbers filed under (for example 2.01, 5.02, 9.01); empty otherwise.
accessionThe filing's accession number, for 8-Ks.
excerptAt most 1,200 characters of the announcement's own text (for an 8-K, the text under its items).
label_versionThe version of the event-type definitions: sec-items-v1 for an 8-K (mapped from its item numbers), a news- version for other items.
roles, role_labels, role_versionFor a leadership change, the role concerned (below) and the version of its definitions; never the person.

News labels: one choice, event_type (the list above), and five flags: about_this_company (the item is about the company itself), executive_appointment, executive_departure, names_amount (a money amount is stated), mentions_hiring. An item that is not about the company itself carries no event type.

Leadership roles: an item whose event type is a leadership change carries the role concerned, one of ceo, cfo, coo, cro_cso (revenue, sales or commercial leader), cmo, cto_cio_cpo, chro, general_counsel, board, other_executive, none, and the direction (appointed, departed, succession), with a multiple flag when more than one role changes. The person is not stored; roles holds the role id. An item keeps the version it was labelled under, named in role_version.

company_headcounts

Employee counts over time, from public statements. One row per company, date and source.

ColumnMeaning
as_ofThe date the count refers to: the annual report's period end, or the date the open reference record gives.
company_id, company, tickerThe company.
employeesThe count as stated.
sourcesec_10k (the annual report) or wikidata (an open reference record).
refThe filing's accession number, or the open reference record's item id.

traffic_ranks

The website's monthly traffic tier from the Chrome UX Report (Google), licensed CC BY 4.0; delivered with attribution.

ColumnMeaning
monthFirst day of the month the rank is for.
domain, company_id, companyThe website and its company.
rank_bucketThe smallest bucket the origin is in: 1000, 5000, 10000, 50000, 100000, 500000, 1000000, 5000000, 10000000, 50000000. Smaller is more traffic.
originThe origin the bucket came from (https://www.example.com), the best among the domain's variants.
attributionChrome UX Report (Google), CC BY 4.0.

company_traffic_history

Traffic tier history: the monthly traffic tier of each company website from the Chrome UX Report (Google), licensed CC BY 4.0, worldwide and within each country, as a series of its own that is never compared with traffic_ranks. One row per company website, month and scope. A month is delivered once it has been read in full. A website on our list when the month was read (watched_from on or before list_as_of) that has no row for it was under the source's threshold, not a drop; a website added later has no reading for that month until the month is read again for it, in a round once enough websites have joined (its rows then backfill, recorded then). Rows are written once and never changed. No visitor data.

ColumnMeaning
monthFirst day of the month the rank is for. The period of an export filters on it.
scopeglobal, or a country by its two-letter code in lower case (us, gb ...): the website's rank within that country, never compared with the global one.
domainThe company website.
rank_bucketThe smallest popularity bucket among the website's origins that month: 1000, 5000, 10000, 50000, 100000, 500000, 1000000, 5000000, 10000000, 50000000, 100000000. Smaller is more traffic.
tier_decaderank_bucket rounded up to a power of ten (5000 reads 10000): 1000, 10000, 100000, 1000000, 10000000 or 100000000. The value to compare from month to month, whatever the scheme.
schemeThe buckets the month used: decade (powers of ten, months to September 2022) or half (the half steps as well, from October 2022). Compare across the change with tier_decade.
originThe origin the bucket came from (https://www.example.com): of the website's four (https or http, bare or www), the best ranked; among equals https before http, and bare before www.
known_gaptrue for a month the source itself reports gaps in. Compare across it with care.
provenancelive when the month was stored within 45 days of its end, backfill when it was read later.
versionThe rule that wrote the row (crux-history-v1).
public_fromWhen the source made this record public: its own release time where it states one, the end of its stated day where it gives a day, otherwise a conservative delay after its own date. Never earlier than the source shows and never later than recorded_at. The time to backtest on. Here: the end of the day, Pacific time, on which the source published the month's ranks, as its release notes state; recorded_at for a month whose release day is not stated.
recorded_atWhen the row was stored. Never changed.
watched_fromWhen the website joined our list of company websites.
list_as_ofWhen the month's read began: the time of the list of websites it was read against. A website with watched_from on or before it and no row for the month was under the threshold; one added later has no reading for that month (it may gain a row, recorded then, when the month is read again).
attributionChrome UX Report (Google), CC BY 4.0, on every row.

Referring companies

How many other companies' websites link to a company's website, its rank and its change. Fields of the company row (GET /api/v1/companies, GET /api/v1/companies/{id}, find_companies, enrich_companies, company_profile), kept up to date as each website is read and ranked once a day. The three are empty until the counts are whole, about four weeks after they are first held, once every website has been read again. Counts only: the companies behind a count are not part of the data.

ColumnMeaning
referring_companiesHow many distinct other companies' websites link to the company's website now. A company's own websites never count, nor do shared platforms (social networks, app stores) or pages that link very many companies.
referring_rank0 to 100: the share of companies linked from at least one other company's website that are linked from fewer. null with no company linking, or before the day's ranking.
referring_change_90dLinks begun less links ended over the last 90 days, seen on websites read before. null with no company linking and no change.

company_health_weeks

One row per company and closed week, for every company followed that week, steady ones included, so each company has a full weekly series. Written once after the week closes and never changed. Steady is the common band: most companies, most weeks.

ColumnMeaning
week_startThe Monday of the week (UTC).
health_versionThe version of the health definitions the row was written under.
distress, breakoutTwo scores of 0 to 100: how strongly the week's evidence shows the company under strain, and how strongly it shows it growing fast.
bandWhere the two scores put the company: distress (signs of distress), watch (worth watching), steady, growing, breakout (breaking out). band_text gives the words.
reasonsThe reasons that count most, strongest first: [{reason, text}], reason a stable id of what was seen (a change on the website, in hiring, in an announcement or a filing, a price or a label), text the same in plain words ("a hiring freeze", "cut a price", "website traffic rose").
shiftsThe shifts in how the company positions itself seen in the 13 weeks to the week: [{shift, text}], shift one of upmarket (moving upmarket), downmarket, ai_pivot (turning to AI), new_buyer (selling to a new buyer), price_up, price_down. Empty for most companies.
inputsHow many dated inputs the week's scores stand on.
industry, countryThe company's industry label, and the head office's country where it is known, otherwise the main market its website serves.
reconstructedtrue for a week written more than seven days after it closed.

A reason that names a website's traffic tier is derived from the Chrome UX Report (Google), CC BY 4.0: keep the attribution with it.

company_groups

A company's place in a group of companies: one record per company with a parent or a subsidiary known, served by the API (group on /companies and /companies/{id}) and the MCP (company_profile section group, and the brief fields on every company row). Read from three kinds of public record: the legal entity register's relationship records for the LEIs we hold (each entity's direct and ultimate parent as it reports them, read monthly), the subsidiary list a listed company files with its latest annual report, read when it files a new one, and the ownership of a brand (a brand's website filed under the company that owns it).

ColumnMeaning
parentThe one parent: { id, name, lei, country, source, basis, confirmed_by }. id is the company in the index, null when the parent is not held (then name and country are the register's). source is register, annual_report or brand; basis says it in words (registered, filed, brand); confirmed_by lists the others that name the same company.
ultimate_parentThe top of the group, in the same form; null when it is the parent itself.
subsidiariesThe largest of the three counts below.
subsidiaries_filedThe subsidiaries the company's latest annual report lists (a listed company lists its significant ones), less the holdings of 50% or less it lists beside them; null when no list was read.
subsidiaries_registeredThe entities the register shows the company consolidating, directly or through others; null when it shows none.
subsidiaries_heldIts subsidiaries held as companies in the index, from any of the three, its brands among them; each counted once.
annual_report{ form, filed_at, accession } of the report the list was read from.
register_as_ofThe day of the register file its parents or count last changed with.
filed, filed_totalOn /api/v1/companies/{id} and in the profile: the list as the annual report files it, at most 100 rows of filed_total, each with its name, jurisdiction, country, the stake the report states (stake, percent; 50 or less is a holding, never counted nor linked) and the company it points at.

A name in a subsidiary list points at a company only when exactly one company carries it, its country agrees with the jurisdiction the list states, it is not a holding of 50% or less, and, for a company known by its website alone, the subsidiary bears the filer's own name; every other name is kept as filed, never guessed. A subsidiary that files one annual report with its parent takes no list of its own from it. An LEI points at the one company holding it on its own record, and only when the record is the company's own (its countries or, with no country to compare, its name agree). One parent is given however it is known: the register first, then the annual report (a company in a listed company's list and in no register record takes that company as its parent), then the brand's owner. A brand's owner that is the recorded parent or ultimate parent, or sits above or below the recorded parent in one chain of ownership, agrees with it; one that does not is not shown, and is listed for review.

listed_securities

Listed securities: one row per equity listing worldwide, keyed by FIGI, with the company it belongs to when one matches. Served by the API (/listed) and available as a file on request; the identifier columns of every other dataset come from it.

ColumnMeaning
figiThe listing's FIGI (stable, one per listing). A listing an open reference record asserts before its FIGI is assigned carries a wd: key in its place until a pass brings the FIGI.
share_class_figiThe FIGI shared by every listing of the same share class; the value carried as figi on the other datasets.
nameThe issuer's name as the source gives it.
ticker, exchangeThe ticker and the venue code (US, LN, GR, JP ...).
micThe venue's ISO 10383 market identifier code, when known.
isin, leiThe security's ISIN and the issuer's LEI (through the LEI register's ISIN mapping), each only when its check digit is valid.
countryThe issuer's country, from its LEI record or open reference records.
security_typeThe security type: Common Stock, Depositary Receipt, REIT, Partnership Shares and the national certificate forms; never funds, ETPs, rights, warrants or preferred stock.
statusactive, or delisted once a listing has not been seen for 45 days after a full pass, or (a wd: listing) once no source lists its ticker today (the row is kept, so history joins stay intact).
company_id, companyThe company in the Fokals index this listing belongs to; empty until a match is made. The match is by ISIN, LEI, ticker on the same market, or name and country, in that order.

Filings and records

What companies and their insiders file with the US securities regulator, and the public records about them. Every row carries the day the source gives it, public_from, recorded_at, provenance and version (Conventions), and is never revised: an amendment, a restatement or a revision by the source is a row of its own, so the rows public by a day give the record as it was known that day. A filing belongs to the company that carries the filer's CIK, matched when the row is read (the oldest company in the index that carries it); a filing whose filer is no company in the index is delivered with the company fields empty. A person is never named: an insider or a holder who is a person is a role and a reference that holds within one issuer only. Every row also carries company_id, company and the listing's identifiers (Conventions).

company_insider_filings

Insider filings: one row per report filed by a company's insiders about its shares (Forms 3, 4 and 5, amendments included), by the day it was filed.

ColumnMeaning
accessionThe report's accession number, its unique key.
formThe form as filed: 3, 3/A, 4, 4/A, 5, 5/A.
amendmenttrue for an amendment (/A).
filed_atThe day the report was filed. The period of an export filters on it.
public_fromWhen the source made this record public: its own release time where it states one, the end of its stated day where it gives a day, otherwise a conservative delay after its own date. Never earlier than the source shows and never later than recorded_at. The time to backtest on. Here: the end of the day filed, US Eastern time, since the regulator publishes a filing by the end of its filing day.
recorded_atWhen the row was stored. Never changed.
provenancelive when stored within 14 days of being filed, backfill when read later from the filings of earlier days.
versionThe readers that stored the row: sec-history-v1, or sec-v0 for a row stored before versions were recorded.
issuer_cikThe issuer's CIK, ten digits, as filed.
issuerThe issuer's name as filed.
issuer_tickerThe trading symbol the report gives, as filed (may be out of date or empty); ticker is the company's listing.
event_dateThe report's own date (its period).
owner_rolesEach reporting owner's role, in filing order: director, officer, ten_percent, other.
owner_refsEach reporting owner's reference: the same for one owner within one issuer, meaningless at any other.
organisationsThe names of the reporting owners that are organisations. A person is never named.
plannedtrue for a report made under a trading plan set in advance, false when the form says it was not, empty where the form did not ask (every Form 3, and every report before the form asked).

company_insider_trades

Insider transactions: one row per line of an insider report (a trade, a grant, an exercise or conversion, or a holding reported), by the day the report was filed.

ColumnMeaning
accessionThe report's accession number. With line, the row's unique key.
lineThe line's number within the report.
formThe report's form as filed (3, 4, 5, with /A for an amendment).
filed_atThe day the report was filed. The period of an export filters on it.
public_fromWhen the source made this record public: its own release time where it states one, the end of its stated day where it gives a day, otherwise a conservative delay after its own date. Never earlier than the source shows and never later than recorded_at. The time to backtest on. Here: its report's, since a line is public with its report.
recorded_atWhen the report was stored; its lines are stored with it. Never changed.
provenanceAs in company_insider_filings.
versionAs in company_insider_filings.
issuer_cikThe issuer's CIK, ten digits.
traded_onThe transaction day; empty for a holding.
owner_refThe reporting owner within the issuer: the same reference for one owner within one issuer, meaningless elsewhere.
roleThe strongest role among the report's owners: officer, director, ten_percent, other.
titleThe officer's title as filed.
securityThe security's title as filed.
derivativetrue for a derivative security (an option, a unit).
codeThe form's transaction code (P, S, A, D, F, G, M ...), or HOLD for a holding a Form 3 or 5 reports.
code_meaningThe code in plain words (Bought on the market, Sold on the market, Granted or awarded, Withheld for tax ...).
acquiredtrue for shares acquired, false for shares disposed of, empty when the form does not say.
sharesShares in the line, as filed.
pricePrice per share, as filed.
owned_afterShares owned after the line, as filed.
directtrue for direct ownership, false for indirect.
value_usdshares times price for a buy or a sale on the market (P, S); empty otherwise.

company_holder_filings

Large holder filings: one row per statement filed by a holder of more than five per cent of a company's shares (Schedules 13D and 13G, amendments included), by the day it was filed.

ColumnMeaning
accessionThe statement's accession number, its unique key.
formThe form as filed: SCHEDULE 13D, SCHEDULE 13G (with /A) in the structured form; SC 13D, SC 13G (with /A) before it.
amendmenttrue for an amendment.
filed_atThe day the statement was filed. The period of an export filters on it.
public_from, recorded_at, provenance, versionAs in company_insider_filings.
issuer_cikThe issuer's CIK, ten digits.
issuerThe issuer's name as filed.
event_dateThe date of the event that required the statement; empty for a statement filed as text.
holdersEach filer, as JSON: ref (within the issuer), role (holder), entity (true for an organisation) and name (an organisation's only). A person is never named.
percentThe largest share of the class any filer reports; empty for a statement filed as text, never estimated.
sharesThe most shares any filer reports; empty for a statement filed as text.
activetrue for a 13D, a holder that may seek to influence the company; false for a 13G, a passive holder.
formatxml for a statement filed in the structured form (from 18 December 2024), text for one filed as text before it: its filers are read from its header only.
cusipThe class's CUSIP as the structured form gives it; empty for text.

company_financial_facts

Financial statements as filed: one row per figure as filed in a company's structured financial statements: a filer, a metric, the concept and unit it was filed under, the period it covers and the filing that stated it. The same figure stated again in a later filing (a comparative or a restatement) is a row of its own, so the rows filed by a day give the figures as known that day. Only figures the filer tagged with a standard concept and no dimension are held; no figure is worked out from others.

ColumnMeaning
filedThe day the filing was filed. The period of an export filters on it.
cikThe filer's CIK, ten digits.
metricOne of twenty: revenue, cost_of_revenue, gross_profit, research_development, selling_general_admin, operating_income, interest_expense, income_tax, net_income, eps_basic, eps_diluted, shares_diluted_average, operating_cash_flow, capital_expenditure, cash, total_assets, total_liabilities, total_equity, long_term_debt, shares_outstanding.
taxonomyus-gaap, ifrs-full or dei (the cover page).
conceptThe concept as filed, with its taxonomy (us-gaap:Revenues, ifrs-full:Revenue).
unitA currency (USD), a currency per share (USD/shares) or shares.
period_startThe first day of the period; for a balance, the day itself.
period_endThe last day of the period, or the balance's day.
months3, 6, 9 or 12 for a flow (52 and 53 week years are 12); 0 for a balance on one day.
fiscal_yearThe fiscal year of the filing, as the filer states it.
fiscal_periodThe fiscal period of the filing (Q1, Q2, Q3, FY, H1).
formThe filing's form: 10-K, 10-Q, 20-F, 40-F, 10-KT, 10-QT, 6-K, 8-K, or an amendment of one (/A).
accessionThe filing's accession number.
valueThe figure as filed, in unit.
frameThe calendar frame the regulator gave the figure when the row was stored (CY2025Q3, CY2025Q3I), when it gave one. A stored row keeps the frame it was read with, so two rows can carry one frame: read periods by period_start, period_end and filed.
public_fromWhen the source made this record public: its own release time where it states one, the end of its stated day where it gives a day, otherwise a conservative delay after its own date. Never earlier than the source shows and never later than recorded_at. The time to backtest on. Here: the end of the day filed, US Eastern time, since the regulator publishes a filing by the end of its filing day. as_of reads it.
recorded_atWhen the row was stored. Never changed.
provenancelive when stored within 14 days of filed, backfill when read later from the filer's history.
versionThe parser that wrote the row (sec-facts-v2).

Read as of a day through the API, a metric and period give one figure: from the latest filing that stated the period by that day, in the unit most of the metric's figures are in, and the preferred concept within that filing, with first_filed, filings (how many filings stated it by then) and restated (whether a later filing changed it).

company_filing_events

Company events from filings: one row per current report (8-K or 8-K/A) a filer filed, by the filer's CIK and the report's accession; a report filed for two registrants is two rows. The event types are worked out from the item numbers when a row is read, in the same thirteen types as company_news, and are never mixed with the types announcements carry.

ColumnMeaning
filedThe day the report was filed. The period of an export filters on it.
cikThe filer's CIK, ten digits.
accessionThe report's accession number (0001234567-26-000123). With cik, the row's key.
form8-K or 8-K/A (an amendment).
public_fromWhen the source made this record public: its own release time where it states one, the end of its stated day where it gives a day, otherwise a conservative delay after its own date. Never earlier than the source shows and never later than recorded_at. The time to backtest on. Here: the end of the day filed, US Eastern time, since the regulator publishes a filing by the end of its filing day.
recorded_atWhen the row was stored. Never changed.
provenancelive when stored within 14 days of filed, backfill when read later from the filer's history.
versionThe reader that stored the row: eight-k-history-v1 (a filer's history, read once) or eight-k-days-v1 (the daily pass). Both read the same file the same way.
itemsThe item numbers the report was filed under, as the regulator lists them (2.02, 5.02, 9.01).
event_typesThe event types the items make, in the thirteen types of company_news. Empty for a report filed under item 9.01 alone.
rule_versionThe items-to-types rule applied when the row was read (sec-items-v3).
leadership_from_items_onlytrue when event_types holds leadership_change from item 5.02 alone: without the report's text, a report of an officer's pay alone also reads as a leadership change.

Business verification

Whether a company is registered and active, with its registry numbers, from November 2026. The records come from the company's LEI record (read monthly), its record as a US securities filer (read weekly), the UK company number its own website states (read at every website read) and the UK company register's record of that number (read on first finding, then monthly). No record names a person or gives an address. Served on the company page, by GET /api/v1/companies (registration, and the registered, regstatus and incorporated filters), GET /api/v1/companies/{id} and the MCP (the profile's registration section, registry_status and registered_with on every row).

company_verification

One row per company that any registry holds, or whose UK company number is known.

ColumnMeaning
company_idThe company.
verified_byThe registries that verify the company: lei (an LEI not annulled or found a duplicate), sec (a US securities filer), uk_register (the UK register holds the number under the company's own name, or the number came from its LEI).
statusactive, inactive, dissolved, insolvency (in liquidation, administration, receivership or a voluntary arrangement) or unknown. The UK register's status where it verifies the company; else active while the company files periodic reports as a US securities filer; else the LEI record's (active only while the LEI's registration is kept current; an inactive entity is dissolved when its record says so).
incorporated_onThe day the company was incorporated: the UK register's, else the LEI record's day of establishment.
uk_numberThe UK company number (eight characters: 01234567, SC123456).
uk_number_basisHow it is known: register (confirmed by the register), lei (from the LEI record), stated (on the company's own website, not checked) or not_confirmed (checked: the register holds no company under it, or holds it under another name).
lei, cikThe LEI and the CIK (US securities regulator id) of the records held.
checked_onThe day a registry's record was last read.

company_registrations

One row per company and registry record (source), rewritten only when the record changes.

ColumnMeaning
sourcelei, sec, uk_site (the number the website states), uk_lei (the UK number the LEI record gives, its registration authority being the UK register: RA000585, RA000586 or RA000587) or uk_register.
registry_idThe LEI, the CIK or the UK company number.
statusThe record's status in the words above.
factsThe record's own fields. LEI: entity and registration status, legal form code (ISO 20275) and its name, registration authority and number, jurisdiction, day established, end and reason, next renewal. sec: filer categories, SIC, state or country of incorporation, fiscal year end, exchanges, latest periodic report, notice of deregistration, reporting state. uk_register: company status and finer status, type, incorporated and ceased, SIC codes (UK SIC 2007), jurisdiction.
read_at, changed_atWhen the record was last read, and last changed.

Market series

market_series

How an industry, a country, a size band or a market moved, computed weekly from the company tables (METHODOLOGY.md section 9). One row per metric, dimension, window and as-of date, written once. Rows over fewer than 20 companies (20 postings for pay and share metrics) are not written.

ColumnMeaning
as_ofThe Sunday the windows end (UTC, inclusive).
metricA metric family, or family:subject for families with subjects (tech_added:salesforce, seniority_share:senior, news_event:funding_round, intent_surge:crm).
family, subject, metric_name, unit, sourceThe catalogue's description of the metric, added on export and by the API.
dimension_kind, dimensionall; industry (the website-label industry id); sector (the company's sector text); country (ISO 3166-1 alpha-2 headquarters country); size (1-10, 11-50, 51-200, 201-1000, 1001-5000, 5000+); market (an exchange code, for listings).
window_days7, 30, 180; 365 for headcount; traffic has 30 and 180 only.
countThe numerator: events, postings or companies in the window; for a median metric, the median itself.
cohortWhat the rate is over: companies or websites covered since before the window began (same-store; website changes: added before the window before began), or postings for share and pay metrics. 0 for listings.
ratecount in the metric's unit: per_100_companies, per_100_sites, share_of_postings (0 to 1), share_of_companies (0 to 1), median_usd, median_pct; empty for count metrics.
previousThe same count over the window just before, on the same cohort. Announcements, from the first as-of written after 16 November 2026: that window as it stood as long after it closed as this one when written; empty before announcements were read.
growth(count − previous) / previous; empty when there was nothing before.
index100 at the series' first as-of date, then the rate relative to it; for a series counted a new way, 100 at the first as-of counted that way (method_since).
written_atWhen the row was written.
reconstructedtrue for an as-of written more than seven days after it closed, after a run that failed (METHODOLOGY.md section 9).
method_since, method_changedWhere a metric began to be counted a new way (METHODOLOGY.md section 9): the first as-of counted the way this row was, and true on that as-of. Earlier rows were counted the earlier way. Empty and false for a series counted one way throughout.

Metric families (GET /api/v1/series/metrics gives the live catalogue with subject lists): hiring_new, hiring_closed, hiring_open, hiring_net, pay_median, seniority_share, work_mode_share, function_share, ai_role_share, sales_share, sales_segment_share, sales_lead_source_share, sales_ote_median, sales_base_median, tech_added, tech_removed, tech_prevalence, gtm_page_added, market_added, language_added, replatformed, news_event, intent_surge, traffic_up, traffic_down, headcount_change_median, listings_new, listings_delisted.

Online stores

store_catalogue

What a company's own online store offers, read from the store itself once a week, for every company that runs its own online store and sells online; never a marketplace. One row per store with the figures of its latest weekly read; no row is kept for a product. Given as store on GET /api/v1/companies and GET /api/v1/companies/{id} (with its price rises and cuts of the last year) and in the MCP's company rows and profile.

ColumnMeaning
platformshopify or woocommerce.
productsProducts the store offers at its latest weekly read.
products_added_30dProducts added to the store in the last 30 days; empty where the store does not say.
median_priceThe median price of its priced products, in currency.
currencyISO 4217 code of the store's prices.
discount_shareShare of its priced products sold below their regular price, 0 to 1.
out_of_stock_shareShare of its products that cannot be bought now, 0 to 1.
read_atThe latest weekly read (UTC). A store not read in the last 60 days is not given.
price_moveThe latest rise or cut of the store's regular prices in the last 90 days (store_price_moves), or empty.

store_price_moves

A rise or a cut of a store's regular prices, measured on the same products from one weekly read to a later one. One row per store and read that saw it, written once and never changed. A store shows none before its second read.

ColumnMeaning
directionraised or cut.
changeThe move of the store's regular prices across the products compared: 0.052 is 5.2% up.
matchedProducts of the store compared at both reads.
observed_atThe weekly read that saw the move (UTC).
since_atThe read before it.
currencyISO 4217 code of the store's prices.