Maison Constellation — Map the Narrative Neighborhood
Profound and Peec stop at Share of Voice — they tell you how often you're cited. AUGMA goes further: Maison Constellation maps your narrative neighborhood. For every prompt × model × run, Claude Sonnet extracts the cited entities, resolves them against a curated taxonomy of 125+ luxury Maisons, and classifies each co-citation as alongside / instead-of / in-comparison / as-alternative. The result: a Substitution Leaderboard, a Reciprocal Coverage audit, a Strategic Matrix, and a Co-Citation Network Graph — all in your verticale's vocabulary, refreshed weekly.
125 Maisons
Curated taxonomy across 4 luxury verticales · weekly extraction
What you get with Maison Constellation
Substitution Leaderboard
Top-20 peers by 'instead-of' count, with delta vs prior period and example prompts. The list answers the most painful question: who is replacing your Maison in AI answers, and on which territories.
Reciprocal Coverage Audit
Asymmetry score (-1..+1) per peer: do you outshine them in mutual co-coverage, or do they outshine you? A scatter plot maps each peer on the diagonal — defendable wins above, authority gaps below.
Strategic Matrix — direct rivals vs niche substitutes vs pairs
A 2×2 grid splits your peers into Direct Rivals (red — co-cited AND substituted), Niche Substitutes (orange — substituted on sub-territories), Pairs (green — healthy coexistence), and Below Threshold. An auto-generated Strategic Read above the grid names the lead insight.
Co-Citation Network Graph
A radial SVG graph with your Maison at the centre and the top 10 peers on two concentric rings. Edge thickness = co-citation weight, node colour = quadrant. Hover any peer for the full cooccurrence breakdown + mean sentiment.
How it works
Every Sunday evening, after the monitor pipeline finishes, a fire-and-forget worker calls Claude Sonnet on each AI response. Claude extracts the cited entities, resolves them against a 125-Maison taxonomy seeded across 4 verticales (Parfumerie, Haute Joaillerie, Maroquinerie, Mode & Haute Couture), and classifies each co-citation. Results land in the `co_citations` table with brand_id × monitor_run × prompt × model × cited peer × cooccurrence type. The dashboard reads two RPCs (substitution_leaderboard + reciprocal_coverage) for the leaderboard + scatter, plus a third RPC for the network graph. ~$12-18 per Sovereign brand per weekly run.