SpyderBot Documentation

Commerce and leadership

Commerce and leadership metrics use explicitly labelled question groups and entity mappings. They must not silently reuse an unrelated organic, risk or product result. Keep brand, product, retailer and founder identities distinct; being connected in a business does not make their metrics interchangeable.

On this page

Commerce Mention Rate · Commerce Share of Voice · Shopping Recommendation Rate · Founder Mention Coverage · Organic Founder Negative Context · Leadership Claim Accuracy · Claim Verification Coverage · Finance and Ownership Context Coverage · Commerce Sentiment, Drivers and Barriers

Commerce Mention Rate

Unit: Percent of eligible commerce answers

How often the target is substantively mentioned in the declared shopping or commerce question group.

How it is calculated

Apply the mention-presence definition and balanced aggregation within eligible commerce questions. Preserve brand versus product and retailer mappings. Question groups that name a target remain separately labelled as aided or comparative.

How to interpret it

This is observed AI representation, not purchases or conversion.

What to do next

Inspect buying-stage questions and the exact entity mentioned.

Related concepts: Organic Visibility ↗

Measurement scope and aggregation ↗ · Evidence and availability ↗

Commerce Share of Voice

Unit: Percent of tracked commerce exposure

The target’s share of balanced exposure among the tracked entities in the commerce group.

How it is calculated

Divide target commerce exposure by total tracked exposure measured on the same commerce scope. A zero total has no defined share.

How to interpret it

This is not sales share or marketplace share.

What to do next

Compare the same tracked set and shopping intent.

Related concepts: AI Share of Voice ↗

Measurement scope and aggregation ↗ · Evidence and availability ↗

Shopping Recommendation Rate

Unit: Percent of eligible shopping answers

How often AI substantively recommends the target in the declared eligible shopping task.

How it is calculated

Count recommendation outcomes across all complete eligible shopping answers, including answers with no recommended target, and apply balanced aggregation. Keep organic discovery, named-brand endorsement and explicit comparisons separate.

How to interpret it

A product appearing in a list is not automatically a recommendation; an affiliate link is not a purchase.

What to do next

Read the reasons and alternatives in the underlying answers.

Related concepts: Organic Recommendation Rate ↗ · Candidate Inclusion Rate ↗

Measurement scope and aggregation ↗ · Evidence and availability ↗

Founder Mention Coverage

Unit: Percent of eligible open leadership answers

How often an answer substantively mentions a verified founder associated with the target company in eligible open leadership questions.

How it is calculated

Apply response-level presence to the verified founder set within neutral, unseeded leadership discovery questions. An answer counts once if it represents any founder in that set. A separately labelled Leadership Mention Coverage uses its declared leadership-role set; a current CEO is not automatically a founder. Deduplicate repeated references and keep individuals’ drilldowns distinct from the set-level result.

How to interpret it

It is not the proportion of all brand discussion about leadership, nor an assessment of a person’s quality.

What to do next

Check entity identity, role and observation time.

Measurement scope and aggregation ↗ · Evidence and availability ↗

Organic Founder Negative Context

Unit: Percent of founder-representing answers in the named cohort

How often answers representing a verified target-associated founder contain negative context linked to at least one represented founder.

How it is calculated

Within eligible open leadership questions, condition on answers that substantively represent at least one verified target-associated founder. Measure whether any represented founder has target-linked negative context, then use conditional balanced aggregation. The denominator is founder-representing answers, never all eligible answers. Keep risk-aided and risk-unseeded cohorts separately labelled. A negative-theme breakdown conditions further on negatively represented answers and must name that narrower denominator.

How to interpret it

No founder representation makes negative context not applicable even when Founder Mention Coverage is a valid 0%. Company criticism alone is not founder criticism, and negative context about one founder is not attributed to every founder. AI repetition does not establish a claim as true. Overlapping themes need not sum to 100%.

What to do next

Review evidence, temporal context and verification before responding publicly.

Measurement scope and aggregation ↗ · Evidence and availability ↗

Leadership Claim Accuracy

Unit: Percent of adjudicated eligible claims, balanced per answer

The supported-true share of fact-checkable leadership claims that have an adequate true-or-false adjudication.

How it is calculated

Within each applicable answer, deduplicate equivalent verifiable claims and calculate supported true / (supported true + contradicted false). Combine these per-answer ratios using conditional balanced aggregation. Unchecked or unresolved claims are excluded from accuracy but included in verification coverage. Opinions and non-verifiable statements are identified separately.

How to interpret it

This is not a pooled count of true claims divided by all claims. Low verification coverage can prevent publication. With no eligible claims the metric is not applicable; with claims but none adjudicated accuracy is unavailable. All adjudicated claims being false can yield a valid 0%.

What to do next

Read the cited verification evidence and its time alignment.

Illustrative example

In two equally weighted applicable cells with per-answer ratios 1/1 and 1/3, balanced accuracy is (100% + 33.3%) / 2 = 66.7%; the pooled 2/4 = 50% is a different diagnostic.

Related concepts: Claim Verification Coverage ↗

Measurement scope and aggregation ↗ · Evidence and availability ↗

Claim Verification Coverage

Unit: Percent of eligible verifiable claims, balanced per answer

How much of the eligible verifiable claim inventory has a supported true-or-false adjudication.

How it is calculated

Within each applicable answer, calculate (supported true + contradicted false) / (supported true + contradicted false + unresolved or unchecked eligible claims). Combine these per-answer ratios using conditional balanced aggregation. Unknown extraction prevents claiming a complete inventory.

How to interpret it

Verified does not mean true: both supported and contradicted claims count as adjudicated. Historical truth is not automatically current truth. Coverage is not accuracy.

What to do next

Resolve unverified claims before relying on an accuracy headline.

Illustrative example

An answer with 2 supported, 1 contradicted and 2 unchecked claims has 60% verification coverage and 66.7% accuracy among adjudicated claims.

Measurement scope and aggregation ↗ · Evidence and availability ↗

Finance and Ownership Context Coverage

Unit: Percent of eligible neutral finance answers

How often eligible neutral finance questions produce target-linked finance or ownership representation.

How it is calculated

Apply response-level presence across all usable answers in the predeclared eligible neutral finance group, then balance the results. Do not define eligibility by whether the answer happens to mention finance. Claim tables preserve supported, contradicted, unresolved, opinion and non-verifiable statuses and their temporal context; row-level accuracy and verification coverage use their own declared claim scope.

How to interpret it

Complete evidence without finance representation can yield 0%. Finance discussion is not an investment recommendation or verification that every quoted financial fact is current.

What to do next

Inspect claim evidence and time alignment before relying on the representation.

Related concepts: Leadership Claim Accuracy ↗ · Claim Verification Coverage ↗

Measurement scope and aggregation ↗ · Evidence and availability ↗

Commerce Sentiment, Drivers and Barriers

Unit: Cohort-specific distributions, theme rates and observed referral units

How shopping answers portray the target, which reasons they give, and separately which connected referral outcomes were observed.

How it is calculated

Commerce sentiment uses the four-label distribution in the declared commerce cohort. Driver/barrier coverage counts each attributed theme once per represented response–entity and uses conditional balanced aggregation; overlapping themes need not sum to 100%. AI Commerce Referrals consumes the published connected AI Traffic result with its exact commerce definition, source, window, attribution and unit; it is not inferred from shopping mentions.

How to interpret it

A positive shopping theme is not a purchase. An unavailable referral integration is not zero and must not be replaced with Quick Overview estimates.

What to do next

Compare stated buying reasons with supported referral outcomes while preserving the measurement distinction.

Related concepts: Sentiment Distribution ↗ · Theme and Aspect Coverage ↗ · AI Referral Sessions ↗

Measurement scope and aggregation ↗ · Evidence and availability ↗