Rigwise

How retailer offers are matched, checked and ranked

Commerce integrity

Rigwise is an aggregator, never the merchant of record. Merchant & Offer Autopilot turns retailer listings into comparable offers only after product identity, price, shipping, stock and feed-level quality pass deterministic checks.

The short answer

Strong identifiers publish automatically. Conflicting identifiers, ambiguous products, malformed commerce data and implausible prices are quarantined. A bad feed update is rolled back instead of replacing the last known-good offer set.

Current demo audit

These counts come from the same decision ledger that produced the visible sample offers. They are not hand-authored marketing numbers.

Sample feeds only

Last autonomous commerce run

run_20260811180203_3w9em2q3i9pw · Aug 11, 2026, 18:02 UTC

Rows inspected
139
Published
137
Quarantined
2
Publish rate
99%
Identifier matches
136
Title matches
1
Price anomalies
1
Live / sample
0 / 137

One autonomous path from shop feed to product page

  1. 1 · Acquire

    Fetch the authorized JSON feed behind timeouts, size limits, conditional caching, retries and a scheduler lease.

  2. 2 · Resolve

    Try GTIN, then MPN, then one unique brand-and-model title. Every method records its score and evidence.

  3. 3 · Guard

    Validate HTTPS URL, integer money, shipping, currency, stock and the price against its historical median.

  4. 4 · Publish

    Write only passing listings and offers. Keep every rejected row in the audit ledger with exact reason codes.

Product matching hierarchy

Automatic publication requires at least 85 points. There is no manual override in the data path and no language model silently guessing a product.

EvidenceScoreDecisionWhy
Exact GTIN100Auto-publishGlobal identifier resolves to one canonical product.
Exact manufacturer part number98Auto-publishManufacturer identity resolves to one canonical product.
Unique brand + model in title88Auto-publishAllowed only when exactly one canonical product matches.
Conflicting or ambiguous evidence0QuarantineCoverage is rejected before product identity is guessed.

Row-level price guard

With at least three historical observations, a new price must remain between 45% and 225% of its median. A value outside that deliberately broad range is evidence of a likely unit, currency, bundle or parsing error—not automatically a bargain.

Feed-level rollback gate

At least 80% of incoming rows must pass. An established retailer also cannot fall below 60% of its previous published listing count. Failure rolls back the complete transaction and keeps the last known-good offer set.

Offer Autopilot FAQ

Does affiliate commission affect which retailer ranks first?

No. Offer ranking reads stock, data freshness and total delivered price. Affiliate status and commission are disclosure fields and never enter the score.

What happens when a retailer title cannot be matched confidently?

The listing is written to the commerce decision ledger as quarantined and is not shown as an offer. Product coverage is sacrificed before match quality.

Can one bad retailer feed delete all previous prices?

No. A feed must keep at least 80 percent row-level publication quality and cannot collapse below 60 percent of its prior published set. A failed gate rolls back the complete update.

Does AI decide whether two retailer listings are the same product?

No. The current system is deterministic: exact GTIN, exact MPN, then a unique brand-and-model title match above a published threshold. Conflicts fail closed.

Continue with the complete recommendation methodology or read the affiliate disclosure.