Metrics¶

Every number Brandvelia reports is a deterministic SQL aggregate. AI features, dashboards, and finance pages all read the same definitions — nobody recalculates by hand in the UI.

Core metric definitions¶

Metric	Formula
`revenue`	Gross order value (USD).
`refunds`	Money returned (USD).
`chargebacks`	Disputed charges (USD).
`fees`	Platform / network fees (USD).
`ad_spend`	Spend (USD).
`opex`	Operational expenses (USD).
`conversions`	Successful orders.
`net_revenue`	`revenue − refunds − chargebacks − fees`
`profit`	`net_revenue − ad_spend − opex`
`gross_roas`	`revenue / ad_spend`
`net_roas`	`net_revenue / ad_spend`

All monetary values are USD. Periods are scoped by calendar date: today (last 24h) and range (the selected window).

Deterministic snapshot¶

getDashboardSummary() (used by Finance, Copilot, and Experiments) runs the bv_get_dashboard_summary RPC over a month window and returns:

interface DashboardSummary {
  today: PeriodMetrics;
  range: PeriodMetrics;
  funnel: { visits: number; leads: number; conversions: number };
  top_offers: RankedEntity[];
  top_campaigns: RankedEntity[];
}

Ranked entities order by net_revenue − ad_spend.

Radar / Opportunity Score signals¶

For marketplace products the Radar feeds a composite Opportunity Score (0–100) built from:

Gravity — marketplace demand (sweet spot: 25–120).
Initial payout (USD).
Quality tier.
Payout milestones.
Percentage breakdown across signals.

The dashboard KPIs are pure SQL: total snapshots, mean score, sweet-spot count, and max initial payout.

Rule of law¶

Metrics and financial calculations are deterministic in SQL — never calculated by an LLM. AI assistants read the precomputed snapshot and may only report what's present.