Methodology
How OpenTenders turns public procurement data into screening leads, and where the method can fail.
A signal is not an allegation
A signal means an automated rule found a pattern worth checking. Severity prioritizes review; it is not a probability of wrongdoing. A score summarizes published signals and data confidence; it is not a legal or factual conclusion. No signal does not prove a clean record.
From source to publication
- Public notices and available documents are stored durably, parsed, normalized, and measured for coverage.
- Deterministic rules and narrowly scoped document analysis create internal findings with method and version metadata.
- A human reviewer checks the finding and source context. High-risk or person-linked cases require stronger review controls.
- Confirmed findings receive stable public revisions. Later corrections withdraw or supersede a revision instead of erasing history.
Indicator glossary
These are implemented indicator definitions. Public finding pages show the exact code, method, rule/model version, evidence state, review date, and limitations that apply to that revision. Availability depends on data coverage and publication policy.
COMP.ALL_LOTS_ONE_WINNER- Single supplier won every lot of a multi-lot tender
COMP.HIGH_DISQUALIFICATION- Most bidders were disqualified, leaving one winner
COMP.REPEAT_WINNER- Winner captures an outsized share of this authority's awards
COMP.SHORT_DEADLINE- Unusually short bidding deadline
COMP.VAGUE_DISQUALIFICATION- Bidder disqualified with no specific clause, item, or amount cited
SPEC.BRAND_NO_EQUIVALENT- Brand required without an "or equivalent" clause
SPEC.EQUIVALENT_FORBIDDEN- Equivalent products explicitly forbidden
SPEC.GEOGRAPHIC_RESTRICTION- Physical proximity requirement restricts bidders
SPEC.NARROW_PARAMETERS- Specification parameters narrowly match one product
SPEC.ORIGINAL_REQUIRED- Only original manufacturer parts allowed
SPEC.PROPRIETARY_STANDARD- Proprietary standard required instead of an open one
SPEC.SINGLE_SOURCE- Specification insists on a single manufacturer or brand
QUAL.CERT_OVERKILL- Certification requirement exceeds the contract's scope
QUAL.EXPERIENCE_EXCESS- Required reference-contract value far exceeds this contract
QUAL.TURNOVER_EXCESS- Required turnover disproportionate to contract value
PROC.ESTIMATE_GAMING- Sole bid landed suspiciously close to the internal estimate
PROC.CONTRACT_MODIFICATION_RATE- Contract modified after award, at an authority with an unusually high modification rate
NET.REVOLVING_DOOR- Evaluation commission member linked to a bidder
NET.SHARED_OWNERSHIP_BIDDING- Bidders matched on an owner or director name
NET.SHELL_COMPANY_WINNER- Winner was registered shortly before winning
NET.SUPPLIER_VALUE_CONCENTRATION- One supplier dominates this authority's spending
PRICE.IDENTICAL_BIDS- Bidders quoted identical prices
Versions, quality, and limitations
Threshold or logic changes require a new analysis version and recalculation. Indicator quality is assessed on versioned human-labelled samples using precision, confidence bounds, applicability, evidence availability, and disagreement rates. Until a version passes its release gate, it must remain shadowed or disabled.