The value of a distributor database rests entirely on the quality of its underlying data. This page explains how DistributorIQ data is sourced, verified, maintained, and where its limitations lie.
DistributorIQ is built on a simple conviction: in a market where data quality is rarely disclosed and directories routinely mix verified facts with inferred estimates, explicitness about what you know and how you know it is a competitive advantage, not a weakness.
Every data point in every profile carries a quality tier. Every profile carries a last-verified date. Every shortlist is reviewed by an analyst who is accountable for the output. This page documents the standards behind those commitments.
DistributorIQ profiles are built from five primary source types, applied in sequence. No single source is sufficient — profiles require cross-referencing before data points are accepted.
Sources are only useful once they are machine-readable, deduplicated and normalised into a schema that allows two distributors in different countries to be compared. That pipeline is the part of the methodology that has changed most, and it is what makes maintaining a database viable rather than rebuilding a list per project.
Every registry, licensing body and procurement portal behaves differently. Some publish structured exports, most publish paginated HTML, several publish scanned PDFs. Each source gets its own acquisition routine written against its actual behaviour, with the extraction rules and the retrieval date recorded alongside the output. A single general-purpose scraper would fail on most of them, which is precisely why this data has never been aggregated before.
Regulatory filings, licence records and company disclosures are written for compliance, not for analysis. Large language models read these documents and return structured field values against our schema — legal entity, licence class, registered activities, declared capacity, named officers. The model is instructed to return a null and a reason rather than a guess where the source does not support a value, and every extracted field is stored with the passage it came from so an analyst can check the derivation rather than take it on trust.
The same company appears as an Arabic legal name on a licence, a transliterated trading name on a customs record, and an English marketing name on its own website. Machine translation handles the first pass; the mapping between the three is then held explicitly in the record, so a search in either language reaches the same profile.
One distributor can appear in four registers under three spellings, two legal entities and a trading name. Candidate matches are generated on name, address, licence number and officer overlap, scored, and merged above a confidence threshold. Everything below that threshold is queued for an analyst rather than merged automatically, because a wrongly merged record is far more damaging than a duplicate one.
Values arrive in incompatible units and vocabularies — warehouse capacity in square metres or pallet positions, coverage described as cities, provinces or regions, fleet size as owned, leased or contracted. Normalisation maps all of it onto the same fifty-plus fields, which is what allows scoring to compare like with like rather than comparing how well two companies describe themselves.
Each redeployment is diffed against the previous state of the market. New registrations, lapsed licences, dissolved entities, changed officers and changed activities are surfaced as a change set rather than rediscovered from scratch. This is what turns a quarterly refresh into a manageable review rather than a full rebuild.
On the search side, a written brief is interpreted into weighted criteria — coverage, regulatory standing, channel access, capability, portfolio conflict — which the client can see and adjust before anything is scored. Every distributor in the market is then scored against those weights on the same fields, so the ranking reflects the brief rather than a generic notion of quality.
It is not a scraper pointed at the open web, and it does not decide anything. Automation makes collection affordable enough to keep a database current instead of rebuilding a list every time a project starts. Every field it produces is subject to the quality tier framework below, and every shortlist it feeds is reviewed by an analyst before a client sees it. Where a market or a sector is not yet in the pipeline, the same methodology is run as a consulting engagement instead.
The tier framework is the core of DistributorIQ's transparency commitment. Every data point in every profile is assigned a tier at the point of entry and reviewed at each refresh cycle. Tiers are displayed in profiles and incorporated into scoring confidence calculations.
The database does not attempt to be exhaustive. Completeness is less valuable than accuracy. A distributor with a partially verified profile is more useful than an unverified name from a directory.
A distributor profile that was accurate eighteen months ago may be materially wrong today. Ownership changes, licences lapse, competing brands are picked up. The refresh protocol is designed around the rate of change of different data types.
The AI scores. The analyst is accountable. Every shortlist passes through a specialist analyst who is responsible for the quality and accuracy of the final output. This is not a review checkbox — it is where the analyst layer earns its value.
Stating limitations clearly is not a weakness — it is the only way a client can make an appropriately calibrated decision. The following limitations are inherent to the product at its current stage.
Every profile carries these quality tiers on every field, so you can see what is verified and what is inferred before you rely on it. Where a market is not on Hub, the same methodology runs as a consulting project.