How Pakki interprets buyer evidence
A useful buyer brief shows what the sources support, how it matches your factory, and what remains unknown. These are the rules behind that distinction.
Research scope
This page describes a research method. It does not establish any company’s purchasing intent and is not legal or compliance advice.
Packaging findings retain source references, excerpts when available, evidence labels, and dates. Public brand directories are dated research snapshots. Their observations should be checked against the linked brand pages before use.
Observed, derived, inferred, and unknown
Observed: An attribute supported by source text or qualifying visual evidence. The packaging extractor currently emits observed claims. For example, a source that explicitly says “stand-up pouch” supports a format observation; a supported product image can establish a visible format, but cannot establish its material or certification.
Derived: A value computed from underlying evidence, rather than quoted directly from a source. The interface supports this label separately from observed claims. Fit and coverage scores are examples of computed outputs; they are not statements made by a buyer.
Inferred: An interpretation that goes beyond a directly supported observation. The interface provides a separate inferred label. For example, seeing a coffee bag does not by itself confirm its barrier structure or the timing of a future order.
Unknown: A missing requirement or factory capability, shown as an unknown or insufficient-evidence state. It is not a fourth claim label and is not treated as a confirmed mismatch.
Claim cards show the evidence type and a separate low, medium, or high confidence label, with expandable references. A trusted primary or official text reference can support high confidence. Visual format evidence alone receives at most medium confidence. These labels describe support for the claim, not a guarantee of a sale.
How the factory Fit Score works
The capability match uses six dimensions. Within each comparable dimension, the proportion of buyer requirement terms matched by factory capabilities earns that dimension’s weighted points. Scores are rounded to a 0–100 scale.
| Dimension | Weight | Comparison |
|---|---|---|
| Packaging format | 35% | Buyer format claims compared with the factory’s formats. |
| Material | 25% | Named packaging materials compared with the factory’s materials. |
| Process and features | 15% | Printing and packaging features compared with printing, converting, and equipment capabilities. |
| Industry | 15% | Buyer product categories compared with the factory’s served industries. |
| Market | 5% | Recorded target markets compared with the factory’s markets. |
| Certification | 5% | Explicit buyer certification requirements compared with listed factory certifications. |
Missing information contributes no points and appears as unknown. A known buyer requirement that is absent from the supplied factory capabilities appears as a gap. Scores are not rescaled to 100 when evidence is sparse.
Example: if format is the only comparable dimension and it matches fully, both Fit Score and match coverage are 35. The result is still insufficient evidence. A high match requires both a score of at least 75 and match coverage of at least 75%; coverage below 40% is insufficient evidence. No comparable dimensions means no score.
Size ranges, order quantities, and lead times can be recorded in the factory profile, but are not currently included in this score. Certification name matching does not independently verify a certificate or establish compliance.
Two different coverage measures
Evidence coverage counts how many of eight packaging categories have at least one trusted source reference: format, material, feature, product category, printing process, required certification, size range, and target market. It is the number of covered categories divided by eight, rounded to a percentage. Four supported categories give 50%; ten references in one category still cover only one category.
Match coverage adds the weights of the scoring dimensions where both buyer requirements and factory capabilities are available. It measures how much of the comparison is possible, not how much matches. The buyer brief marks low coverage as insufficient evidence and lists unknowns separately from gaps.
There is no fixed number of source links that makes a buyer fully qualified. Several links repeating one fact cannot fill missing materials, requirements, or commercial context.
Dates, refreshes, and corrections
An observation date records when the evidence was observed or collected. A source’s publication date, when available, is stored separately. An observation date is not the date a brand began using a package, chose a supplier, or decided to buy.
Active company website watches are scheduled no more frequently than every 168 hours (one week). Actual successful refresh time depends on source availability and processing. A scheduled scan is not proof that new evidence was retrieved.
Public directories retain their own review dates and are separate from workspace watches. Recheck the original page when freshness matters. Send corrections with the affected page and an official source.
Submit a research correction · Explore public brand directories