Interface mockup from an academic capstone. Not a shipped product. Findings below are inferred network relationships derived from trade and corporate records. They are not verified claims about labour conditions at any facility, and they are not allegations against any named company.

← ClearCart dashboard

Sayari demo: filling the People & Supply Chain pillar

Populating one empty pillar for one product using commercial supply-chain intelligence, and what that exposes about the limits of product-level transparency. Schlage Encode Plus Smart Lock, 7 August 2026.

Tool calls
6
Candidates
554 → 1
Supplier links
165
Risk flags
3 kept
Redacted
2

get_investigation_guidance · search_entities · get_entity_profile · find_beneficial_owners · get_upstream_supply_chain · lookup_risk_factors

Before and after

In the six-product pilot audit, People & Supply Chain was the emptiest pillar: five of six products had no product-level labour evidence from any certification scheme or independent reviewer. The Encode Plus was one of two where the pillar was a total gap — no brand claim to evaluate, no reviewer coverage, nothing to score.

Before

People · no data

No certification, no reviewer coverage, no brand claim. The pillar rendered as a gap because there was nothing to render.

After

People · check this

Ownership resolved and corroborated, supplier network traversed to tier 3, three risk indicators interpreted and confidence-capped.

The Encode Plus was selected for a further reason. Its parent, Allegion plc, is publicly traded, which means findings can be published without the accuracy risk that attaches to supply-chain claims about small private firms.

The resulting pillar card

CLEARCART Schlage Encode Plus

People and supply chain

This lock's parent company buys electronic parts through suppliers that sit near known forced-labour risk, several steps up the chain.

This is a pattern, not a proven case. Nothing here says a worker was harmed making this lock. It says the parts trail runs through companies flagged in forced-labour research.

Confidence
Medium
Applies to
The company
Checked
Aug 2026

How we got here

  1. Schlage → owned by Allegion plc
  2. Ships from an Indonesian assembler
  3. Which buys from parts brokers in Asia
  4. Trail goes dark past this point

Supply-chain relationships retrieved via Sayari, August 2026.


Method

Retrieval used Sayari's Model Context Protocol server, which exposes its corporate and trade knowledge graph as callable tools. Seven steps, each producing a record retained in the source pipeline.

  1. Resolve the brand to a legal entity

    A name search for “Schlage” returned 554 candidate entities. Selection was made on registration identifiers, jurisdiction, and trade-record volume rather than name similarity: SEC CIK 0001604114, United States, active as of 30 June 2026.

    search_entities → entity mYqYdo-toIyWAmlSPqkXrQ

  2. Profile the entity

    14,106 inbound trade records, 2,918 outbound, 165 supplier relationships, 35 customer relationships. Not sanctioned, no politically exposed persons. Three forced-labour risk indicators present.

    get_entity_profile

  3. Verify ownership

    Beneficial-ownership traversal resolved to Allegion plc. This is the only step in the chain rated High confidence, because it is independently corroborated by SEC filings — Allegion's own subsidiary disclosures name the entity.

    find_beneficial_owners, max_depth 5

  4. Traverse upstream

    Nineteen of twenty returned paths route through a single Indonesian contract assembler at tier 2, beyond which the network opens into electronic-component brokers and distributors across Asia. Most returned paths carried a partial_path quality flag, meaning the trace is incomplete by the provider's own assessment.

    get_upstream_supply_chain, max_depth 3

  5. Interpret the risk indicators rather than repeat them

    All three forced-labour flags share a methodology: they denote indirect trade with a third party named in published forced-labour research, assessed across tiers 2 to 4 within a rolling 730-day window, using a proprietary product-component mapping. None denotes an observed condition at a facility. This distinction determines the confidence tier and the wording of the consumer card.

    lookup_risk_factors

  6. Apply the confidence rule

    Under the pipeline's existing tiering, evidence that is cited from a third party and derived by inference is capped at Medium. It cannot reach High without independent corroboration, and the inference chain here is explicitly probabilistic. The card states Medium and states why.

  7. Redact what should not reach a consumer surface

    See below. This step is not optional.

What was excluded, and why

Two Allegion board members carry flags in the ownership data indicating association with sanctioned and forced-labour-linked entities. These attach because the individuals hold directorships at other companies. They say nothing about Allegion, and nothing about this product.

What a naive pipeline renders

This product's owner is linked to a sanctioned entity.

Technically traceable to the provider's output. False in every sense a shopper would understand it, and potentially defamatory.

What this pipeline renders

Nothing.

Person-level and director-level flags are dropped at the adapter boundary. They never enter the record that feeds a consumer surface.

The design decision this forces is worth stating plainly: the correct output is not everything the data provider returns.

Provider data is engineered for a compliance analyst who is trained to read a flag as a lead requiring investigation. A shopper reads it as a verdict. The same field, rendered to the same standard of fidelity, means opposite things to the two audiences.

Limitations

Could this have been done without a commercial provider?

Substantially, yes — for the inputs. Ownership is available free from SEC filings. US bills of lading are public and searchable without cost. Sanctions lists, the UFLPA Entity List, the ILAB list of goods, and the academic reports underlying the forced-labour indicators are all free downloads.

What is not replicable independently is the join: resolving 554 name matches to one legal entity, and recursing from tier 1 to tiers 2, 3, and 4 across jurisdictions. Roughly speaking, the evidence is public and the resolution is not. That is a finding about where the transparency bottleneck actually sits.


Scope of use across the framework

The obvious next question is whether the same source can fill the remaining gaps. It is admitted for two pillars and excluded from the other three.

PillarFitCan doCannot do
People & Supply Chain Strong, direct
  • Resolve a brand to its legal entity and beneficial owner
  • Traverse supplier relationships beyond tier 1
  • Screen entities against sanctions and forced-labour designations
  • Evidence trade lanes and manufacturing jurisdictions from customs records
  • Observe conditions at any facility
  • Attribute a supplier relationship to a specific product
  • Exceed Medium confidence on inferred relationships; only corroborated ownership reaches High
  • See relationships outside the 730-day window or outside published customs data
Transparency Strong, derived
  • Test a stated country of manufacture against shipment records
  • Check whether a named factory or supplier appears in the trade record at all
  • Test whether a claimed short or direct supply chain is consistent with observed intermediaries
  • Distinguish a claim that survives the record from one that is merely repeated
  • Confirm a claim is true — absence from the record is not disproof
  • Evaluate claims with no physical trade footprint, such as wage or governance assertions
  • Assess domestic-only production, where customs records do not exist
  • Distinguish an evasive disclosure from an incomplete dataset

Excluded pillars

Planet, Quality, and Value were evaluated against this source and excluded. In each case a plausible-looking calculation was available and was rejected as unsound.

Two of five is the finding. The pilot audit found that no independent reviewer covered more than two or three pillars for any product. A commercial supply-chain intelligence provider — a far more expensive and specialised source — is admissible for two, and only under stated confidence limits.

No single source spans the framework. That is the argument against a composite score, and it now holds on both the media side and the commercial-data side.

Pipeline integration

This retrieval is treated as one more source adapter alongside creator and reviewer ingestion. It emits the same record shape: per-pillar claims with evidence typing, confidence tier, gap markers, and attribution. Three adapter-specific rules apply.