Rosetta — Semantic Consistency Report

Generated 2026-10-25T04:17:29.997514+00:00
6
Conflicts
1
Critical
2
High
63
Assets at risk
active_user CRITICAL silent contradiction blast radius: 22

'Monthly Active Users' (finance) and 'Monthly Active Users' (marketing) share a name but compute differently (logic overlap 13%).

AI Explanation
Finding'Active User' has 2 incompatible definitions across finance and marketing — same name, different computation.
EvidenceLogic similarity 13%, confidence 93%. Definitions: finance: "Users who completed at least one paid transaction in the trailing 30 day…"; marketing: "Any user with a session or app open in the last 30 days, bots excluded u…". Governance signals: marketing: definition stale (last updated 2024-03-02).
ImpactCRITICAL severity. 22 downstream assets are at risk. A wrong 'Active User' silently feeds 15 decision surface(s) (11 dashboards, 4 models) across 2 teams.
RecommendationEscalate immediately: align finance and marketing leadership on a single canonical definition for 'Active User'. Freeze dependent pipelines until resolved. In Connected Mode, Rosetta can execute a human-approved remediation plan and verify the resulting metadata where supported.
DomainOwnerDefinitionComputation
financeurn:li:corpGroup:finance-analyticsUsers who completed at least one paid transaction in the trailing 30 days.COUNT(DISTINCT user_id) WHERE txn_amount > 0 AND event_date >= CURRENT_DATE - 30
marketingurn:li:corpGroup:growth-marketingAny user with a session or app open in the last 30 days, bots excluded upstream by the safety pipeline.COUNT(DISTINCT user_id) WHERE session_start >= CURRENT_DATE - 30 AND is_bot = false
logic similarity 0.133 · name similarity 1.0 · confidence 0.933 · est. cost $990
arr MEDIUM silent contradiction blast radius: 6

'Annual Recurring Revenue' (finance) and 'Annual Recurring Revenue' (sales) share a name but compute differently (logic overlap 10%).

AI Explanation
Finding'ARR' has 2 incompatible definitions across finance and sales — same name, different computation.
EvidenceLogic similarity 10%, confidence 95%. Definitions: finance: "Sum of all active subscription MRR annualized, excluding one-time fees a…"; sales: "Total contract value of all active annual deals, including upsells and e…". Governance signals: sales: definition stale (last updated 2023-09-27).
ImpactMEDIUM severity. 6 downstream assets are at risk. A wrong 'ARR' silently feeds 3 decision surface(s) (3 dashboards, 0 models) across 2 teams.
RecommendationOpen a data governance ticket for 'ARR'. Draft a canonical definition incorporating both teams' intent and circulate for sign-off in the next governance review cycle.
DomainOwnerDefinitionComputation
financeurn:li:corpGroup:finance-analyticsSum of all active subscription MRR annualized, excluding one-time fees and professional services.SUM(monthly_subscription_amount * 12) WHERE subscription_status = 'active' AND revenue_type = 'recurring'
salesurn:li:corpGroup:sales-opsTotal contract value of all active annual deals, including upsells and expansion revenue booked in period.SUM(contract_value) WHERE deal_type = 'annual' AND stage = 'closed_won' AND contract_status = 'active'
logic similarity 0.097 · name similarity 1.0 · confidence 0.952 · est. cost $270
conversion_rate HIGH silent contradiction blast radius: 11

'Conversion Rate' (marketing) and 'Conversion Rate' (product) share a name but compute differently (logic overlap 16%).

AI Explanation
Finding'Conversion Rate' has 2 incompatible definitions across marketing and product — same name, different computation.
EvidenceLogic similarity 16%, confidence 92%. Definitions: marketing: "Percentage of website visitors who completed any tracked goal (signup, p…"; product: "Percentage of trial users who converted to a paid plan within 14 days of…".
ImpactHIGH severity. 11 downstream assets are at risk. At least one ML model trains on this data — a wrong definition corrupts model signals silently. A wrong 'Conversion Rate' silently feeds 7 decision surface(s) (5 dashboards, 2 models) across 2 teams.
RecommendationSchedule a cross-team definition review for 'Conversion Rate' within this sprint. Propose one canonical term, notify all 11 asset owners, and deprecate the conflicting variant.
DomainOwnerDefinitionComputation
marketingurn:li:corpGroup:growth-marketingPercentage of website visitors who completed any tracked goal (signup, purchase, or trial) in the session.COUNT(DISTINCT goal_completions) / COUNT(DISTINCT sessions) * 100
producturn:li:corpGroup:product-analyticsPercentage of trial users who converted to a paid plan within 14 days of signup.COUNT(DISTINCT user_id WHERE subscription_start <= trial_start + 14) / COUNT(DISTINCT trial_user_id) * 100
logic similarity 0.16 · name similarity 1.0 · confidence 0.92 · est. cost $495
customer_ltv HIGH silent contradiction blast radius: 12

'Customer Lifetime Value' (finance) and 'LTV' (data_science) share a name but compute differently (logic overlap 0%).

AI Explanation
Finding'Customer LTV' has 2 incompatible definitions across finance and data_science — same name, different computation.
EvidenceLogic similarity 0%, confidence 100%. Definitions: finance: "Predicted gross margin a single account will generate over a five-year h…"; data_science: "Historical dollars collected from each customer since signup. No project…". Governance signals: finance: handles sensitive/PII data; data_science: no owner assigned — stewardship unclear; data_science: not registered in the business glossary; data_science: handles sensitive/PII data; data_science: definition stale (last updated 2023-11-02).
ImpactHIGH severity. 12 downstream assets are at risk. At least one ML model trains on this data — a wrong definition corrupts model signals silently. A wrong 'Customer LTV' silently feeds 8 decision surface(s) (5 dashboards, 3 models) across 2 teams.
RecommendationSchedule a cross-team definition review for 'Customer LTV' within this sprint. Propose one canonical term, notify all 12 asset owners, and deprecate the conflicting variant.
DomainOwnerDefinitionComputation
financeurn:li:corpGroup:finance-analyticsPredicted gross margin a single account will generate over a five-year horizon, discounted to present value.SUM(projected_margin * discount_factor) OVER five_year_horizon
data_science⚠ unassignedHistorical dollars collected from each customer since signup. No projection, no discounting.SUM(invoice_amount) GROUP BY customer_id
logic similarity 0.0 · name similarity 1.0 · confidence 1.0 · est. cost $540
revenue MEDIUM silent contradiction blast radius: 6

'Net Revenue' (finance) and 'Revenue' (sales) share a name but compute differently (logic overlap 16%).

AI Explanation
Finding'Revenue' has 2 incompatible definitions across finance and sales — same name, different computation.
EvidenceLogic similarity 16%, confidence 92%. Definitions: finance: "Gross bookings minus refunds and chargebacks, recognized in the period."; sales: "Total value of closed-won deals booked in the period, before refunds.". Governance signals: sales: definition stale (last updated 2025-01-19).
ImpactMEDIUM severity. 6 downstream assets are at risk. A wrong 'Revenue' silently feeds 3 decision surface(s) (3 dashboards, 0 models) across 2 teams.
RecommendationOpen a data governance ticket for 'Revenue'. Draft a canonical definition incorporating both teams' intent and circulate for sign-off in the next governance review cycle.
DomainOwnerDefinitionComputation
financeurn:li:corpGroup:finance-analyticsGross bookings minus refunds and chargebacks, recognized in the period.SUM(booking_amount) - SUM(refund_amount) - SUM(chargeback_amount)
salesurn:li:corpGroup:sales-opsTotal value of closed-won deals booked in the period, before refunds.SUM(booking_amount) WHERE stage = 'closed_won'
logic similarity 0.158 · name similarity 1.0 · confidence 0.921 · est. cost $270
customer_churn~attrition MEDIUM hidden synonym blast radius: 6

'Churn Rate' and 'Customer Attrition' appear to be the same metric under different names (logic overlap 77%).

AI Explanation
Finding'Customer Churn' and 'Attrition' are duplicate metrics across product and customer_success — same logic, different names.
EvidenceLogic similarity 77%, confidence 81%. Definitions: product: "Share of paying customers who cancelled their subscription in the period…"; customer_success: "Fraction of paying customers who cancelled their subscription during the…".
ImpactMEDIUM severity. 6 downstream assets are at risk. A wrong 'Customer Churn vs Attrition' silently feeds 4 decision surface(s) (4 dashboards, 0 models) across 2 teams.
RecommendationOpen a data governance ticket for 'Customer Churn vs Attrition'. Draft a canonical definition incorporating both teams' intent and circulate for sign-off in the next governance review cycle.
DomainOwnerDefinitionComputation
producturn:li:corpGroup:product-analyticsShare of paying customers who cancelled their subscription in the period.COUNT(DISTINCT customer_id) WHERE subscription_status = 'cancelled' / COUNT(DISTINCT customer_id)
customer_successurn:li:corpGroup:customer-successFraction of paying customers who cancelled their subscription during the period.COUNT(DISTINCT customer_id) WHERE subscription_status = 'cancelled' / COUNT(DISTINCT customer_id)
logic similarity 0.769 · name similarity 0.0 · confidence 0.808 · est. cost $270