Predicting per-client payment timing for smarter invoice reminders

Short summary

A statistical approach to predicting when each client actually pays invoices, using median absolute deviation to model per-client payment distributions and trigger reminders only when an invoice deviates from that client's baseline. New clients use a segment-based prior that updates as real payment data arrives. The author's product, Kynth Invoices, automates this with AI-drafted reminders in the user's own voice — but the user always approves before sending.

• •Due dates poorly predict actual payment timing — model each client's personal distribution instead
• •Use MAD (not std dev) so one holiday-delayed payment doesn't widen the nudge window
• •AI drafts reminders in the user's extracted voice profile, but never auto-sends

Generated with AI, which can make mistakes.