On - FC Analytics Overview - 2026-02-04¶

Date: 2026-02-04
Company: On
Attendees: Federico Morando, Amanda Mitt, Simon, Jennifer Pearson, Giannis
Type: Customer Call
Source: Notion Customer Success Meetings

Metadata¶

V2 limitations: only predicts at category level, global balance as aggregate, limited human interaction
V3 targets (by summer): total information system on all levels, human-informed adjustments, bidirectional information flow, enhanced seasonality detection
Key improvements: better "when" predictions, IC modeling as covariates, scenario planning with commitments, exogenous factor integration

Scientific approach — all work proprietary, acting as research + implementation team
Ecos dashboard for experiment comparison vs current models
Performance metrics: bias levels improved; quantile magnitudes from 30M to 13M deviation; VMAP from 300k to 6k
7 behavioral clusters based on trend volatility, seasonality, correlation patterns
Cluster 2 heavily problematic but showing improvement
Categories performing better: tax, payables (seasonal patterns)
Financing category showing degraded performance

IC transactions cleanup most impactful for model performance — will transition from rule-based to model-based in V3
APARS data integration planned — will systematize information into actionable signals
Categorization improvements needed: better descriptions, standardization, address sparse data, system-determined clusters preferred

Synced from Notion: https://www.notion.so/2fdc47335ffb81528f99d0054eab35fe