SpyderBot Documentation

Changes, monitoring and priorities

A change needs both a current result and a comparable baseline. Monitoring uses coherent scheduled cycles. A fresh observation in Explorer does not automatically become historical monitoring evidence, and the latest answer from each prompt must not be assembled into a fictional common cycle.

The time window, scope and method version are part of every comparison. Missing days are not zero-performance days.

On this page

Metric Change · Baseline and History Comparability · Observed Change Contribution · Change Severity · Change Persistence · Observed Leader Change · Prompt Priority · Prompt Demand and Opportunity Estimates · Recommendation Competition Profile · Monitoring Lifecycle and Schedule Health · Change Activity and Opportunity Counts · Context and Segment Differences

Metric Change

Unit: Percentage points for percentage metrics

The difference between a current metric and its comparable baseline.

How it is calculated

For a percentage metric, change in percentage points = current percentage − baseline percentage. Relative percent change is a different calculation and is labelled separately; it is undefined when the baseline is zero. Compare the same prompts, entities, systems, surfaces, locale, cohort, aggregation profile and compatible method versions.

How to interpret it

A larger value does not necessarily mean an improvement: more negative context is adverse. A baseline gap is not a zero baseline.

What to do next

Review the comparison window and contributing observations.

Illustrative example

Visibility moving from 20% to 30% is +10 percentage points and +50% relative growth. It is not “+10% growth”.

Related concepts: Baseline and History Comparability ↗ · Observed Change Contribution ↗

Measurement scope and aggregation ↗ · Evidence and availability ↗

Baseline and History Comparability

Unit: Comparable / Changed scope / No baseline

Whether two results measure the same underlying scope in a way that supports a meaningful comparison.

How it is assessed

Check scope and compatible method versions before calculating change. Keep a first observation distinct from a measured zero. Period summaries use the declared comparable-date policy rather than letting dates with more repeated answers dominate. A changed prompt set or competitor roster starts a new comparison basis unless an explicit common-scope comparison is available.

How to interpret it

“No previous observation” does not mean a newly appearing brand or source. Method changes can create breaks in a time series.

What to do next

Select a common scope or read the visible break in comparability.

Measurement scope and aggregation ↗ · Evidence and availability ↗

Observed Change Contribution

Unit: Component contribution in the displayed unit

Which measured components account for an aggregate difference under a comparable measurement profile.

How it is calculated

Decompose a like-for-like aggregate change using the same declared component weights and observations. Label whether components are prompts, systems, channels or themes. Where a valid decomposition is unavailable, show descriptive changes rather than inventing a contribution percentage.

How to interpret it

This is arithmetic attribution of observed change, not evidence that an edit, source or competitor action caused it.

What to do next

Use it to locate evidence to investigate, then validate any causal hypothesis separately.

Measurement scope and aggregation ↗ · Evidence and availability ↗

Change Severity

Unit: Categorical assessment with reasons

How consequential an observed change appears within the affected scope.

How it is assessed

The category reflects the observed change magnitude, direction, affected scope, evidence support and comparison context. Its explanation identifies the main observations behind the assessment. Severity is a prioritisation aid rather than a publicly reproducible numeric score; internal detection thresholds and calibration remain private.

How to interpret it

Severity is not a confidence level, loss forecast or guarantee that action is required.

What to do next

Review scope and persistence before prioritising remediation.

Measurement scope and aggregation ↗ · Evidence and availability ↗

Change Persistence

Unit: Scheduled-cycle status and count

Whether a qualifying change recurs in comparable scheduled monitoring cycles.

How it is calculated

Track the qualifying observation across the required comparable scheduled cycles. Manual reruns do not automatically create independent persistence confirmation. Missing or incomparable cycles do not count as confirming observations.

How to interpret it

A one-cycle change can be meaningful but is different from a persistent monitored pattern. Do not present an internal threshold as a probability.

What to do next

Inspect the cycle timeline and missing observations.

Measurement scope and aggregation ↗ · Evidence and availability ↗

Observed Leader Change

Unit: Observed transition and basis

A change in the observed leading entity for the named task-aligned metric.

How it is calculated

Compare leaders on the same eligible entity set and metric across comparable cycles. Preserve ties, no-choice outcomes and insufficient evidence. A current leader without a valid prior leader is not automatically a leader switch.

How to interpret it

An observed switch does not establish permanent preference or statistical superiority.

What to do next

Review the leader margin and underlying answers.

Related concepts: Observed Leader ↗

Measurement scope and aggregation ↗ · Evidence and availability ↗

Prompt Priority

Unit: Priority category or ordered list

A planning recommendation for which prompts merit attention or further measurement.

How it is assessed

Discovery Priority uses available planning context to identify useful questions to measure. Observed Priority also uses measured performance and evidence gaps. The reasons explain why a prompt merits attention. Priority can guide collection or action planning, but does not increase the prompt’s weight in the standard balanced headline.

How to interpret it

A high-priority prompt is not necessarily high-volume, high-confidence or high-converting. The internal ranking recipe is not a public performance metric.

What to do next

Use the stated reasons to decide what to investigate next.

Measurement scope and aggregation ↗ · Evidence and availability ↗

Prompt Demand and Opportunity Estimates

Unit: Estimated planning signal

An estimate used to help discover and prioritise relevant questions.

How it is calculated

This planning signal comes from the identified measured, estimated or modelled source and its period and geographic scope. Search-demand proxies remain separate from observed SpyderBot answers and real AI-assistant user activity. Without a supported source value, the estimate is unavailable.

How to interpret it

The estimate is not the number of people who asked the exact prompt in an AI assistant. It is not guaranteed addressable revenue.

What to do next

Combine relevance and observed evidence before committing effort.

Measurement scope and aggregation ↗ · Evidence and availability ↗

Recommendation Competition Profile

Unit: Distribution and counts

How recommendations are distributed across the declared tracked entity set.

How it is calculated

On the common tracked entity set, divide each entity’s balanced recommendation exposure by total tracked recommendation exposure to obtain shares. Concentration (HHI) is the sum of squared fractional shares; Diversity is 1 minus HHI. With no recommendations these are not applicable; with one positive entity HHI is 1 and Diversity is 0. Diversity’s maximum depends on roster size. Mean Recommended Candidates is the balanced mean number of actively recommended tracked entities per answer, including zero for no-choice answers; it can be a decimal.

How to interpret it

Concentrated recommendations do not by themselves measure SEO difficulty, commercial market share or competitor quality.

What to do next

Inspect which entities are recommended for which tasks.

Related concepts: Organic Recommendation Rate ↗ · Observed Leader ↗

Measurement scope and aggregation ↗ · Evidence and availability ↗

Monitoring Lifecycle and Schedule Health

Unit: Lifecycle, scheduled-cycle health and counts

Whether monitoring is active or paused and whether expected observations are arriving on schedule.

How it is assessed

Actual completed scheduled cycles are assessed against the configured cadence, calendar, timezone and deadlines. Lifecycle, schedule health, subject context and metric evidence are separate states. A credit-related pause is a pause reason, not brand performance. Observed Prompt Count describes operationally ready observations, not a universal eligible denominator.

How to interpret it

Monthly schedules follow their calendar configuration, not an assumed fixed 30-day interval. No alert does not prove that nothing changed, particularly when observation is missing.

What to do next

Check the due-cycle timeline, pause reason and metric-specific support.

Measurement scope and aggregation ↗ · Evidence and availability ↗

Change Activity and Opportunity Counts

Unit: Distinct events or eligible prompt units

How many measured changes or actionable prompt opportunities exist in the selected scope.

How it is calculated

Change Activity counts distinct qualifying material events, assigning one primary family for an exclusive event-family breakdown. Operational collection events remain separate. Priority Opportunities counts distinct eligible prompt-matrix units classified in the declared opportunity state; multiple observations of the same unit do not multiply that count.

How to interpret it

Counts do not measure severity, business value or causal impact. A first-observed entity without a comparable prior certified absence is not automatically an emerging entity.

What to do next

Open the change or opportunity evidence and its current status.

Related concepts: Change Severity ↗ · Change Persistence ↗ · Prompt Priority ↗

Measurement scope and aggregation ↗ · Evidence and availability ↗

Context and Segment Differences

Unit: Named metric differences on declared scope

Differences between controlled contexts or separately labelled question segments.

How it is calculated

For a controlled-context comparison, hold other material measurement dimensions compatible and compare the same metric. Intent, funnel or language segment summaries use their own eligible prompt groups and must be labelled as descriptive segment differences. A context change that changes task eligibility creates a different cohort, not a single continuous metric.

How to interpret it

A segment difference does not prove that changing context would cause the measured outcome.

What to do next

Check what was held constant before interpreting sensitivity.

Related concepts: Baseline and History Comparability ↗ · Metric Change ↗

Measurement scope and aggregation ↗ · Evidence and availability ↗