The Product Information Chain Is a Purchasing Problem

Between a manufacturer's specification and a buyer's decision sit four different kinds of authority. A manufacturer publishes a data sheet. A representative defines territory. A distributor holds availability. A dealer accepts the quote request. Each knows something the others do not.

The problem is not one missing field

A buyer sourcing a commercial window or door needs dimensions, glazing options, approvals, lead times, local availability and the name of the party authorized to quote. A searchable product name covers none of that. Some of these facts are stable and technical. Others are commercial and change daily, and they should never be inferred from a catalog.

Why one product record is hard to keep

A single item can carry a manufacturer's model, a distributor SKU, a configuration code and a project-specific quote line. The link between those identifiers has to be written down. A technical PDF can be the authority on a test result and say nothing about stock or installation. A dealer locator can be the authority on who sells and not know what is on the shelf. A distributor can know what it can quote and not own the certification.

This matters when an assistant does the research. It can identify the right product and still name the wrong seller, confuse supply with installation, or present an option as available everywhere. We treat these as failures to test for. We have not measured how often they happen.

The authority map

The useful investigation traces a sample of real products across manufacturer literature, technical documents, dealer locators, distributor pages and quote workflows. For each fact it asks who created it, who can change it, when it took effect, which geography and configuration it covers, and what happens when sources disagree.

The result is a map of who is authoritative for what, not a new master spreadsheet that claims to own everything. A manufacturer can own the performance specification while a local supplier owns the quoting instructions for its territory. Both can be right at once.

What AI can and cannot do

An assistant can help a buyer find documents, summarize known options and prepare questions. It should not turn incomplete documentation into an engineering approval, current inventory, a price or a binding commitment. Where a specification or substitution affects compliance, a person reviews it.

The next step is a repeatable audit across channels: a defined product category, a dated sample of product documents and sales-channel pages, completeness scored field by field, and disagreements recorded with screenshots or archives. Until that exists, this stays an analysis, not a measurement.

What to do first

Begin with the five questions buyers ask most, not a catalog rebuild. Name the authoritative owner of each answer. Separate stable specifications from changing commercial facts. Publish effective dates and territory limits, and send uncertain questions to the responsible person. Measure whether a buyer can find a valid next step, not only whether an AI mentions the brand.

Method and limits

This is a framework drawn from Lehvel's own work in building products, not a survey of suppliers. It claims no market-wide error rate, cause or measured return. Further work needs product-level audits, interviews and independent industry data. Reliability and transparency follow the NIST AI Risk Management Framework (2023), which does not measure building-material catalogs.