World Bank Procurement API Guide with Apify

Query the official procurement service through a maintained Apify Actor with bounded pagination and exports. The World Bank Procurement Scraper turns each public procurement notice into a bounded Apify dataset row and preserves a direct source URL for verification.

Start with a reviewable sample

{"noticeTypes":["Invitation for Bids","Request for Expression of Interest"],"currentOnly":true,"maxResults":100}

Begin with ten to one hundred records, not a full archive. Inspect identifier stability, date formats, nulls, arrays, free-text fields, and source links. Start with ten records, inspect source field coverage, then save the validated input as a Task. Once the sample matches the intended workflow, raise the result limit deliberately and save the input as a named Apify Task.

Design the dataset around source truth

Keep notice ID, type, status, project, country, deadline, procurement method, bid reference, description, and official source URL together in the raw landing table. Do not replace missing values with zero, invent status labels, or silently collapse aliases. Use the publisher's stable identifier as the upsert key where one exists; otherwise combine the canonical source URL and observed identifier. Add a downstream collection timestamp so analysts can distinguish a revised record from a newly published record.

A useful pipeline has three layers. The raw layer stores the Actor output without lossy transformations. The review layer adds normalized dates, currencies, or matching candidates while preserving original fields. The application layer contains only decisions that have passed the organization's validation rules. This separation makes source changes auditable and prevents a convenient spreadsheet formula from becoming an undocumented business rule.

Schedule changes without creating alert noise

Run the Task on a cadence appropriate to how often World Bank changes. Compare stable keys and material fields rather than array order or page position. Send alerts only for additions, removals, deadline changes, status changes, or other changes relevant to the workflow. Store the previous snapshot in a table or object store and make the diff reproducible.

For operational use, log the Actor run ID, dataset ID, input, source URL, and collection time. Use retries for transient network failures, but fail visibly when the publisher returns no records for a query that normally has coverage. Silent empty datasets are more dangerous than explicit failures because they can look like a legitimate all-clear.

Validate before acting

A search result is not the governing solicitation. Confirm dates, eligibility, amendments, submission instructions, and contact details on the official World Bank notice before committing bid resources. The Actor is a collection tool, not a substitute for due diligence, legal interpretation, financial analysis, or publisher methodology. Reviewers should open the source URL, confirm that the record is current, and document the evidence used for any material conclusion.

This workflow is designed for business-development teams, contractors, consultants, and procurement analysts. It is an independent integration and is not affiliated with or endorsed by World Bank.

API behavior to account for

The World Bank search endpoint accepts result size, offset, sort, notice-type facets, country facets, keywords, and deadline bounds. Responses return rows, os, total, and a procnotices array. The Actor caps each request, advances the offset by returned rows, deduplicates notice IDs, retries transient errors, and stops at the user's explicit result limit.

Facet values must match the publisher's labels. “Invitation for Bids” and “Request for Expression of Interest” are not interchangeable with internal procurement method names such as RFB or INDV. Test a filter with ten rows and inspect total before scheduling a large archive. An empty result for an established query is raised as a failure instead of masquerading as a valid empty market.

For downstream APIs, use the Apify dataset endpoint for immutable run output and a Task endpoint for repeatable input. Store the Actor build number with regulated or audited exports. The source service is public and keyless, but that does not remove the need to respect rate limits and official solicitation documents.

Why use an Apify Task

A Task makes the exact input repeatable across schedules, webhooks, the API, Make, Zapier, n8n, and MCP clients. Keep one Task per business question instead of one enormous catch-all export. Small Tasks are easier to test, cheaper to rerun, and clearer when a source field or filter changes. The Thirdwatch Actor on Apify provides JSON, CSV, Excel, RSS, and API access to the resulting dataset.