Job Postings

SoftwareIQ captures openings directly from covered employers and maintains daily and weekly hiring histories.

For new postings, zero means a valid comparison found no additions. Null means the comparison is unavailable. Active-posting counts can remain available while a new-posting comparison restarts.

The purpose-built SI tracker assembles hiring across major global systems integrators and Big Four firms, with a focus on selected priority onshore and nearshore markets. Vendor-practice attribution is conservative; unsubstantiated associations remain unclassified.

Purpose-built classifications organize postings into functions, commercial roles, geographies and investment themes. Explicit not_applicable, unallocated and unknown states preserve the distinction between classification coverage and observed activity.

Company-specific allocation states

Versioned methodology and historical consistency

Use a compatible classification_version_id with its taxonomy references for reproducible analysis.

When a methodology release changes historical classifications, SoftwareIQ may restate the supported classified history under the new version. The underlying direct-source posting observations and observed_total series remain unchanged. Prior classified versions are kept distinct, allowing research teams to pin a version, reproduce an earlier result, and adopt a restatement on their own schedule. Classified observations from different versions should never be spliced into one time series.

Every posting in a classified row belongs to one function, role-state tuple, company-allocation tuple, and mutually exclusive geography leaf. Therefore filtering one or several dimensions is valid, and supported classified observations reconcile to directly observed company totals over compatible coverage. Filled bulk company totals do not imply reconstructed dimensional breakdowns. It is not a collection of independent marginal totals.

• Geography leaves are united_states, canada, latin_america, europe, apac, mea, other_international, and unknown.
• international is a taxonomy grouping node and never appears in a fact row; expand it to its leaf members.
• new is based on posting-identity change only when adjacent observations are comparable. At a baseline, the value is null.
• Unavailable measures remain null, except eligible filled bulk company active totals. Collection availability can remain unavailable when that total is estimated. A legitimate observed count of zero remains zero.
• Daily facts are the primary observation grain. Weekly facts represent the same population; active stock is not the sum of daily stocks and must not be summed with daily rows.
• period_start and period_end describe the observation interval. Collection and publication timestamps describe when SoftwareIQ processed the data, not when an employer created the underlying role.

get_job_posting_series returns one count per period after applying the selected dimensions. Each row contains period_start, period_end, count, availability_status, continuity_status, and is_comparable_to_previous. API/MCP also returns estimated_count, estimate_method, estimate_anchor_dates, collection_event_ids and continuity_segment_id. Estimate values are separate from count; the bulk projection has one count field. Metadata retains company, measure, grain, classifier selection, freshness, and pagination.

Date bounds select period_start. Observed totals remain separate from classified intersections; missing or outage totals remain null. Retain the compatible classifier and taxonomy versions for classified queries.

MCP setup and agent access · Discuss S3 or Snowflake bulk delivery.

Bulk company active totals can include linear interpolation across eligible closed collection gaps of at most 14 calendar days with compatible reliable anchors, rounded to whole postings with halves up. These rows retain the legacy observed_total selector and use continuity_status=estimated. Collection availability and comparison flags remain authoritative. New-posting flows and unsupported dimensional breakdowns are never filled; long/open outages and unsupported migration gaps remain null. Collection events explain affected dates and anchors.

JOB_POSTINGS_DAILY

Grain: one observed total or nonzero classified dimension intersection per company, date, measure, and applicable version. Stable key: job_posting_row_id.

JOB_POSTINGS_WEEKLY

Grain: one observed total or nonzero classified dimension intersection per company, week, measure, and applicable version. Use the availability and comparability fields when comparing periods.

JOB_TAXONOMY_MEMBERS

Stable key: (taxonomy_version_id, dimension_id, member_id).

CLASSIFIER_VERSIONS

Stable key: classification_version_id.

JOB_COLLECTION_EVENTS

Stable key: event_id. Join to facts by company and overlapping period.

DATASET_CATALOG

Stable key: dataset_id. Discover the active classifier here; do not hard-code it for current-view applications.

• Pin classification_version_id for reproducible research and adopt restated classifications as a complete versioned history.
• Use observed_total for the collected population and classified_count for dimensional analysis; these are two views of the same postings.
• Apply continuity and comparability fields before calculating changes across collection events or new baselines. An ATS migration can change observed coverage and requires a continuity check before comparing counts.
• Interpret postings as employer demand and investment intent rather than realized hires or financial performance.