Deltabase

Get frequently asked questions about Deltabase and its products. Optionally filter by product: Culture, DARA, Deltabase One, Talent, Workforce. Omit product to get every FAQ entry. Returns a JSON array, each row: question, answer, product (list), category (list), source_url. Use this to answer questions about how Deltabase works, data guidance, or product details rather than guessing. Present results directly.

get_faq

Get a single company by its unique company_id. Returns the full company record as JSON including sectors and modules. Use search_companies first to find the company_id if you only have a name. Present results directly. Coverage varies by company; industry sectors are set according to our own documented rules; the modules field is authoritative.

get_company

Get the corporate group a company belongs to — parent, subsidiaries, divisions and brands — plus which data modules each entity has. Pass any company_id in the group, not only the ultimate parent. The tree is always the full group rooted at the ultimate parent; is_queried_company marks the company you asked about so you can see where it sits. Coverage is per entity and is not inherited. Each node lists that company's own customer-facing modules (the same catalogue as list_modules and get_company) and has_sales_nav (a LinkedIn Sales Navigator company page with enough headcount to capture). A subsidiary with no modules of its own still appears in the tree. Returns JSON: queried_company_id, root_company_id, node_count, and tree (nested nodes with company_id, company_name, company_type, relationship_type, is_queried_company, has_sales_nav, modules, children). Unknown company_id returns {"error": "Company '…' not found"}. Present results directly. Coverage varies by company.

get_company_hierarchy

Get Culture (employee sentiment) data for one or more companies — a single overall figure per company by default, including employee review counts, not just sentiment scores. Data is aggregated primarily from Glassdoor and Indeed reviews, with other specialist sources used where appropriate. Only companies with the Culture module have data. Check the modules field from search_companies before calling this — do not compare companies where one lacks the Culture module. The measure parameter controls which metric is returned, and whether it's a single all-time figure or a year-by-year breakdown (default: "nss"): - "nss": Net Sentiment Score (-100 to +100). Use for "culture", "sentiment", "how employees feel". This is the default — omit measure unless asking for something else. - "prevalence": Coverage ratio (0 to 1). Use for "coverage", "prevalence". - "rating": Star rating (1 to 5). Use for "rating", "stars", "score". - "review_count": Employee Review Count — how many reviews the company has. Use whenever asked for "review count", "number of reviews", "how many reviews", "volume" — there is no separate review-count tool, this is it. Each of the above returns one row per company (year: null) — a single all-time figure, not a breakdown. For a year-by-year trend (e.g. "how has NSS changed over time", line charts, comparing specific years), append "_per_year" instead: "nss_per_year", "prevalence_per_year", "rating_per_year", "review_count_per_year". Those return one row per company per year, and respect the year parameter if given. Each result row: company_id, year (null unless using a _per_year measure), value (the measure), topic (review category or null). The year parameter only applies to _per_year measures — omit it for the default all-time measures, and omit it even for _per_year measures unless the user explicitly names a single year. Present results directly. Coverage varies by company; industry sectors are set according to our own documented rules; the modules field is authoritative.

get_culture_data

Get promoter/detractor/EPS (Employee Positivity Score) data for one or more companies — an NPS-style view of Culture sentiment, derived from review star ratings. IMPORTANT: "eps" here means Employee Positivity Score - an internal culture metric, NOT earnings per share. Never treat it as a financial metric. Data is aggregated primarily from Glassdoor and Indeed review star ratings, with other specialist sources used where appropriate. This is distinct from get_culture_data's NSS (which comes from sentiment analysis of review text, not star ratings) — use this tool specifically when the user asks about promoters, detractors, or EPS. Only companies with the Culture module have data. Check the modules field from search_companies before calling this — do not compare companies where one lacks the Culture module. No topic-level breakdown is available for these measures - a review's star rating is a single whole-review value with no topic dimension. Use get_culture_topic_data for topic-scoped sentiment instead. The measure parameter controls which metric is returned, and whether it's a single all-time figure or a year-by-year breakdown (default: "eps"): - "eps": Employee Positivity Score (promoters minus detractors, -1 to +1). This is the default — omit measure unless asking for something else. - "promoters": Ratio of reviews rating 4-5 stars (0 to 1). - "detractors": Ratio of reviews rating 1-2 stars (0 to 1). - "eps_per_year", "promoters_per_year", "detractors_per_year": year-by-year breakdowns of the above. Each of the non-"_per_year" measures returns one row per company (year: null) — a single all-time figure. The _per_year variants return one row per company per year, and respect the year parameter if given. Each result row: company_id, year (null unless using a _per_year measure), value (the measure), topic (always null - see note above). The year parameter only applies to _per_year measures — omit it for the default all-time measures, and omit it even for _per_year measures unless the user explicitly names a single year. Present results directly. Coverage varies by company; industry sectors are set according to our own documented rules; the modules field is authoritative.

get_culture_promoter_detractor_data

Get Culture (employee sentiment) data broken down by review topic for one or more companies. Data is aggregated primarily from Glassdoor and Indeed reviews, with other specialist sources used where appropriate. Use this — not get_culture_data — whenever the user asks about specific topics within culture, such as management, work-life balance, pay, benefits, or any other aspect of employee experience. Also use this when the user asks to compare how companies perform across different culture topics. The measure parameter controls which metric is returned per topic, and whether it's a single all-time figure or a year-by-year breakdown (default: "nss_per_year"): - "nss_per_year": Net Sentiment Score per topic, per year. This is the default — omit measure unless asking for something else. - "prevalence_per_year": How often the topic comes up in reviews, per year — a 0-1 ratio. Use for "topic frequency", "how common a topic is", "how often mentioned", or any Driver/High-Risk/Enabler/Low-Risk style topic matrix that needs both NSS and frequency per topic. - "review_count_per_year": Review count per topic, per year. - "nss": Net Sentiment Score per topic, single all-time figure (year: null). - "prevalence": Topic frequency per topic, all-time (year: null). - "rating": Star rating per topic, all-time (year: null). - "review_count": Review count per topic, all-time (year: null). Only returns rows where a topic is present — overall (null-topic) scores are excluded. Each result row: company_id, year (null for the all-time measures above), value (the measure), topic (the review category). Results are sorted by company, year, then topic — suitable for side-by-side topic comparison. The year parameter only applies to _per_year measures — omit it for the all-time measures, and omit it even for _per_year measures unless the user explicitly names a single year. Present results directly. Coverage varies by company; industry sectors are set according to our own documented rules; the modules field is authoritative.

get_culture_topic_data

Get employee turnover (hiring and departures) data for one or more companies. Data is derived from LinkedIn Sales Navigator hiring and headcount insights. Only companies with the Employee Turnover module have data. Check the modules field from search_companies before calling this. Each company's figures are rolled up from its own data plus every company below it in Deltabase's company hierarchy (subsidiaries, divisions, brands) - this is the group total, not just the parent entity in isolation. To get a single subsidiary/division/brand's own figures rather than the whole parent group, pass that entity's own company_id instead of its parent's - the roll-up always starts from whichever company_id you give it, so a child entity with no further subsidiaries below it effectively returns just its own data. Returns one row per company per month, with fields: - headcount: total employee headcount at the end of the month. Use for "headcount", "employee count", "how many people work there", "size". - new_hires: new hires that month. Use for "new hires", "hires", "joiners", "recruits", "hiring", "people joining". - senior_new_hires: of which, senior-level hires. Use for "senior hires", "senior new hires", "senior-level hiring". - departures: departures that month (new hires minus net headcount change). Use for "departures", "leavers", "leaves", "exits", "attrition count", "people who left", "how many people quit/were let go" - this tool does not distinguish voluntary resignations from involuntary departures. - turnover: departures divided by the previous month's headcount, as a decimal ratio (e.g. 0.05 means 5% turnover). Use for "turnover", "turnover rate", "attrition rate", "churn rate", "% who left". Not summed from company-level rates - recomputed from the group's combined departures and headcount, so it stays a true rate at every level of the hierarchy. - median_tenure_years: current median employee tenure across the group. Use for "tenure", "median tenure", "how long people stay", "average time at company". A single figure, not a monthly one, so it repeats on every row for that company. A field is null for a given month if that measure could not be computed (e.g. no prior month available to compute turnover for). Results are sorted by company, then chronologically. Present results directly. Coverage varies by company.

get_employee_turnover_data

Get Workforce (headcount) data for one or more companies. Data is derived from LinkedIn profile data. Only companies with the Workforce module have data. Check the modules field from search_companies before calling this. The measure parameter selects ONE breakdown (default: "allocation_by_function"). Each call returns one breakdown only — call again with a different measure for another. Every breakdown is job function crossed with at most one other axis; there are no three-way splits. Allocation measures — value is that group's share of the workforce (0 to 1): - "allocation_by_function": share of workforce in each function. The default; use for "headcount breakdown", "workforce split", "which functions". - "allocation_by_function_and_tenure": function split further by tenure band. - "allocation_by_function_and_continent": function split further by continent. - "allocation_by_function_and_seniority": function split further by seniority. Headcount measures — value is a profile count (a number of people), not a share: - "headcount_total": total headcount, one row per company (function is null). Use for "how many people work there", "total headcount", "company size". - "headcount_by_function": headcount in each function. - "headcount_by_function_and_tenure": headcount by function and tenure band. - "headcount_by_function_and_continent": headcount by function and continent. - "headcount_by_function_and_seniority": headcount by function and seniority. An unrecognised measure falls back to "allocation_by_function". Each result row: company_id, function (null for "headcount_total"), function_description, value, and the dimension column for the chosen breakdown (tenure / continent / seniority) — the others are null. Results are sorted by company then function. Present results directly. Coverage varies by company; industry sectors are set according to our own documented rules; the modules field is authoritative.

get_headcount_data

Poll for the result of a summary-style search_reviews question that returned a "QUEUED" status. Pass the review_search_id from that response. Returns a JSON object. Its "status" field is one of "QUEUED", "IN_PROGRESS", "COMPLETED", or "FAILED". While "QUEUED" or "IN_PROGRESS", there is no "results" field yet — wait a few seconds and call this again with the same review_search_id. Once "COMPLETED", the object also has "results" and "rationale" fields in exactly the shape search_reviews documents for a summary-style question (a list of theme summaries: theme, summary, reviews — each review carries its content_id). If "FAILED", "rationale" explains what went wrong; there is no "results". A review_search_id that doesn't exist, or belongs to a different user, returns {"error": ...} instead of a status object.

get_review_search_result

Get Talent (skills) data for one or more companies. Data is derived from LinkedIn profile data. Only companies with the Talent module have data. Check the modules field from search_companies before calling this. This NEVER includes Skills Deep Dive data, even when the company's modules field also lists a "Talent Deep Dive: ..." entry — Deep Dive is a separate, additional dataset with its own tool, get_talent_deepdive_data, not a superset of this one. If the modules field lists both, and the user's question is specifically about a Deep Dive vertical (e.g. Gen AI, HR, Finance, Design & Engineering skills), call get_talent_deepdive_data instead of, or in addition to, this one. Returns skill allocation per company — the share of the workforce with each skill. Each result row: company_id, skill, skill_tags (categories), skill_description, tenure (null for overall allocation, or a tenure band for tenure-split allocation), value (allocation as a decimal fraction, e.g. 0.15 means 15% of workforce). Do NOT sum value across skills for a company, and do not describe such a sum as a coverage/total figure. Skills are not mutually exclusive — one person can hold several, so allocations overlap and summing them produces a meaningless number (rows will not add up to 1.0 / 100%, unlike Workforce's function allocation, which does partition the workforce and can legitimately be summed). Results are sorted by company then skill. Present results directly. Coverage varies by company; industry sectors are set according to our own documented rules; the modules field is authoritative.

get_talent_data