Skip to main content
Impact Score is a 0-100 score Cited assigns to every recommendation, indicating how much visibility improvement the action is likely to drive for a specific brand given its current data maturity. Higher scores mean higher-priority actions.

​
Why we score recommendations

Each monitoring cycle produces 3-15 recommendations per brand — content gaps, competitor positioning shifts, technical fixes. Without prioritization, brands cannot distinguish between a gap that affects one low-volume prompt and a gap that spans dozens of high-intent prompts where competitors dominate. Impact Score answers “which recommendation should I act on first?” by combining four measurable factors into a single priority signal.

​
The four input factors

Impact Score is computed from four inputs, each capturing a different dimension of recommendation urgency.
FactorAbbrev.ScaleWhat it measures
Query Volume WeightQV1-10How many tracked prompts are affected by this gap. More affected prompts = higher weight. Floor: 3.
Competitive Gap SizeCG1-10How large the visibility gap is between the brand and its closest competitor on the affected prompts. See bidirectional logic below.
Fix FeasibilityFF1-5How actionable the fix is with available resources. Technical fixes (robots.txt, schema) score higher than long-term authority campaigns.
Intent ValueIV1-5How commercially valuable the affected prompts are. Commercial and transactional intent prompts score higher than purely informational ones.
QV floor of 3: Brands with very few pipeline runs (zero-state or newly onboarded) receive a QV floor of 3. This prevents recommendations from being structurally capped at LOW priority simply because the brand has not yet accumulated enough data — a new brand’s first recommendations are still meaningfully prioritized.

​
Brand-relative normalization

Impact Score is brand-relative, not universal. A score of 60 means “60% of the maximum achievable impact for THIS brand’s current data maturity” — not a universal benchmark comparable across brands. The formula: 
raw_score = QV × CG × FF × IV
max_possible = QV × 10 × 5 × 5
impact_score = (raw_score / max_possible) × 100
The denominator uses the brand’s own QV value, which means the maximum possible score scales with the brand’s data maturity. This ensures that every brand — from a zero-state brand with 3 pipeline runs to a mature brand with 200 — sees meaningful distribution across the LOW/MEDIUM/HIGH tiers.

​
Worked example: two brands, same recommendation quality

Brand A has 15 pipeline runs (QV = 5). A recommendation has CG = 10, FF = 4, IV = 5. 
raw = 5 × 10 × 4 × 5 = 1000
max_possible = 5 × 10 × 5 × 5 = 1250
impact_score = 1000 / 1250 × 100 = 80 → HIGH
Brand B has 200 pipeline runs (QV = 7). A recommendation has CG = 7, FF = 4, IV = 5. 
raw = 7 × 7 × 4 × 5 = 980
max_possible = 7 × 10 × 5 × 5 = 1750
impact_score = 980 / 1750 × 100 = 56 → MEDIUM
Both recommendations are urgent for their respective brands. The scores are not comparable across brands — that is the point. Brand A’s 80 does not mean it has a “worse” situation than Brand B’s 56.

​
Why brand-relative

A universal 0-100 scale makes HIGH impact scores mathematically impossible for low-data-maturity brands. Their recommendations would all cluster at LOW despite being genuinely urgent. Brand-relative scoring ensures every brand sees meaningful prioritization signal within its own data context.

​
Competitive Gap: bidirectional logic

CG treats offensive and defensive scenarios differently. The question changes depending on whether the brand is trailing or leading a competitor on the affected prompts.

​
Offensive (brand trailing competitor)

When the brand’s visibility is lower than the competitor’s, CG reflects distance-to-catch-up — larger gaps produce higher urgency.
Visibility gapCG scoreInterpretation
< 10 percentage points1Nearly caught up
10-30 pp3Moderate gap
30-50 pp5Significant gap
50-70 pp7Large gap
> 70 pp10Dominant competitor, maximum urgency

​
Defensive (brand leading competitor)

When the brand’s visibility is higher than the competitor’s, the question shifts from “how far behind am I?” to “how exposed am I?” CG reflects proximity-to-vulnerability — how close the nearest competitor is to catching up.
Lead marginCG scoreInterpretation
< 10 pp7Competitor closing in — high urgency
10-20 pp5Moderate exposure
20-30 pp3Reasonable buffer
> 30 pp1Comfortable lead — low urgency
Product consequence for dominant brands: Brands with comfortable leads (> 30 pp margins over all competitors) correctly see mostly LOW-scored recommendations. This is not a bug — it reflects genuinely low competitive urgency. Dominant brands benefit more from competitive watchlist features and category expansion analysis than from gap-closing recommendations.

​
Strategic alignment multiplier

When a recommendation’s gap topic matches the brand’s declared positioning (stored in brand_intent.target_attributes), the Impact Score is multiplied by 1.5. Example: A skincare brand declares “retinol” and “vitamin C serums” as target attributes. A recommendation addressing a retinol content gap receives a 1.5x multiplier because the gap aligns with how the brand wants to be known. A recommendation about “sunscreen SPF” — important but not declared as a positioning priority — does not receive the multiplier. Matching logic: Both the gap’s topic and the brand’s declared target attributes are tokenized into significant words (longer than 3 characters, excluding stop words). Any word-level overlap triggers the multiplier. The match is intentionally broad — strategic alignment is a directional signal, not an exact-match filter.

​
Low-query-count safeguard

If the strategic multiplier fires AND the recommendation is supported by fewer than 5 queries, the final score is capped at 59 (just below the HIGH threshold). This prevents single-query edge cases from surfacing as HIGH priority based on strategic boost alone. The recommendation still appears — it is scored MEDIUM rather than HIGH until more query data supports the signal. When this fires in practice: Approximately one recommendation per brand cycle, typically on niche positioning topics where the brand has declared a target attribute but the prompt library contains only 1-3 relevant prompts. Once the prompt library grows to cover the topic with 5+ prompts, the cap no longer applies.

​
Tier thresholds

Impact Scores map to three priority tiers displayed in the dashboard.
TierScore rangeMeaning
HIGH60-100Act on this recommendation first. High query volume, significant competitive gap, actionable fix, valuable intent.
MEDIUM30-59Important but not urgent. Address after HIGH-tier items, or when HIGH-tier items are already in progress.
LOW0-29Low urgency. The gap exists but is small, affects few prompts, or involves a difficult fix. Monitor rather than act immediately.

​
Production tier distribution

The following table shows the tier distribution for four production brands after the April 2026 recalibration. Each row reflects the brand’s current data maturity and competitive position.
BrandBefore recalibrationAfter recalibrationNotes
Zero-state brand1 HIGH / 4 MED / 2 LOW3 HIGH / 3 MED / 1 LOWGrowing brand, more recommendations surfaced as HIGH
Dominant brand0 HIGH / 5 MED / 0 LOW0 HIGH / 1 MED / 4 LOWDominant brand — defensive logic correctly produces LOW scores
Scaling brand1 HIGH / 3 MED / 9 LOW5 HIGH / 5 MED / 3 LOWLow-query-count cap fired once (strategic multiplier capped at 59)
Mid-range brand0 HIGH / 7 MED / 0 LOW0 HIGH / 4 MED / 3 LOWMid-range brand, improved differentiation between MED and LOW
The dominant brand’s shift to mostly LOW is the defensive CG logic working as designed — a dominant brand with 30+ point margins over competitors has genuinely low competitive urgency.

​
What the Impact Score does not capture

Impact Score is a prioritization tool, not a prediction engine. Specific boundaries:
  • Does not predict exact visibility lift. The estimated_improvement field on each recommendation is a separate, rougher estimate. Impact Score ranks recommendations against each other; it does not quantify the outcome of acting on them.
  • Does not account for implementation cost in time or money. Fix Feasibility (FF) is a 1-5 proxy based on fix type, not a project plan. A technical fix scoring FF = 5 might still require engineering scheduling.
  • Does not incorporate historical success rate. Whether past recommendations of the same type produced results is not yet factored in. This is planned future work.
  • Not comparable across brands. A score of 60 for Brand A is not equivalent to a score of 60 for Brand B. Each brand’s score is relative to its own maximum possible score. This is the defining property of brand-relative scoring.

​
Frequently asked questions

Impact Scores are recomputed on each narrative refresh cycle. Scores change when the underlying inputs change — new pipeline data affecting QV, competitor visibility shifts affecting CG, or prompt library updates changing the query count. A score shift of 5-10 points between cycles is normal.
Two common causes. First, the brand may be dominant in its category — defensive CG logic assigns low urgency scores when the brand leads competitors by 30+ percentage points. This is correct behavior, not a bug. Second, the brand may have very few pipeline runs — but the QV floor of 3 should prevent extreme clustering. If all recommendations are LOW and the brand is not dominant, contact support for a diagnostic review.
On every narrative refresh cycle: twice weekly for Pro plan (Tuesday and Friday at 02:00 UTC), weekly for Starter plan (Sunday at 02:00 UTC). See refresh cadence for the full pipeline schedule.
Not currently. The scoring formula is applied uniformly. If you believe a recommendation is more or less urgent than the score suggests, use it as a starting point and apply your own business context — the score does not know your team’s capacity, current sprint priorities, or strategic timing.
Impact Score is a relative prioritization ranking (0-100, brand-relative). It answers “which recommendation should I work on first?” The estimated_improvement field is a rough prediction of how much mention rate or share of voice might change if the recommendation is implemented. They measure different things — one is priority, the other is predicted outcome.
The strategic multiplier requires a word-level match between the recommendation’s gap topic and the brand’s declared target_attributes. If target attributes are not populated, or if the gap topic does not overlap with any declared attribute, the multiplier does not fire. Brands that populate their target attributes thoroughly see strategic alignment on approximately 75% of recommendations.
Plan tier does not directly affect the scoring formula. However, plan tier determines the narrative refresh cadence (Pro: twice-weekly, Starter: weekly), which affects how quickly Impact Scores reflect new pipeline data. Pro brands see updated scores 2-3 days sooner than Starter brands after each pipeline run.
Last updated: 2026-04-21