Pluk Evaluation Standard

The Pluk Evaluation Standard defines how an interaction becomes part of a PlukScore. This page is the detailed specification. For a short version, read how PlukScore works.

The four trust dimensions

Every interaction is rated on the same four dimensions, in every category.

• Reliability
  Showed up, followed through, kept commitments on time.
• Competence
  Delivered work that met a reasonable standard for the field.
• Honesty
  Communicated accurately about scope, pricing, and outcomes.
• Commitment Fulfillment
  Fulfilled the specific agreement made with the customer.

Evidence levels

Every interaction carries an evidence level. Stronger evidence carries more weight and raises Confidence. Uploaded documentation is stored privately and is never published.

1. Level 1
   Unverified Interaction
   An interaction reported without a verified account or supporting material.
2. Level 2
   Account Verified
   Submitted by a contributor with a verified Pluk account.
3. Level 3
   Interaction Supported
   Backed by private documentation such as receipts, invoices, or correspondence.
4. Level 4
   Moderator Verified
   A Pluk moderator has reviewed the supporting documentation directly.

Interaction weighting

Interactions are not counted equally. The weight of an interaction is determined by four factors: the strength of its evidence, the reliability of the contributor, the severity classification of what occurred, and how recent it is. Completion status is recorded as context for moderators and carries no weight of its own. The exact coefficients are held privately so the system cannot be gamed.

Recency

Older interactions lose influence gradually. A business that improved is not held to its past indefinitely, and a business that declined does not coast on old results.

Contributor Reliability Index

Pluk maintains an internal Contributor Reliability Index built from corroboration by independent contributors, evidence quality, and moderator decisions. Helpfulness votes are capped at roughly a tenth of its influence so popularity cannot buy credibility. Duplicate submissions and coordinated posting patterns reduce it. The index is never published, is never shown to businesses, and no individual figure is disclosed.

Severity and misconduct reports

Contributors do not assign a severity rating. The engine derives a routine classification from the ratings and context. If something serious occurred, a contributor may file a Report Professional Misconduct claim describing what happened, with documentation and witnesses. A Pluk moderator reviews the report and makes the final classification, which overrides the engine.

• Routine dissatisfaction
• Material failure
• Serious breach
• Alleged severe misconduct
• Substantiated severe misconduct

Moderation and disputes

Businesses can claim their profile and respond publicly to any interaction. They can also dispute an interaction and submit counter documentation. Moderators review disputes, adjust evidence levels where warranted, and can remove content that breaks the integrity rules. Pluk does not independently verify every submission, and the evidence level on each interaction states exactly how far verification went.

Confidence methodology

Confidence reflects the quantity of interactions, the diversity of independent contributors, the level of verification, and the time span the interactions cover. It is reported as a tier.

• Low confidence
• Moderate confidence
• High confidence
• Very high confidence

Building PlukScore

A PlukScore publishes once an entity has at least 5 interactions from at least 3 independent contributors. Until then the profile shows Building PlukScore. Evidence is not a publication gate. It raises Confidence.

Category attributes

Different fields have different texture, such as bedside manner for a physician, estimate accuracy for a contractor, or ambience for a restaurant. Pluk collects these as informational attributes. They help you compare entities within a category and they never affect the PlukScore.

Fraud resistance

Pluk is designed to identify suspicious patterns, including duplicate submissions, clusters of activity from related accounts, and contributors whose history does not match the interactions they report. Detected patterns reduce contributor reliability and can trigger moderator review. The specific signals are not published, because publishing them would make them easier to evade.

Auditing and versioning

Every entity page carries a plain language explanation of what its score means, what drives its Confidence, and what is still missing. Behind that, Pluk moderators hold a complete audit record: every snapshot, every per interaction factor, the engine version in effect, and the full change history. Contributor identity and internal reliability weighting stay in that private layer. Changes to the standard are versioned, and every snapshot records the version that produced it.