Comparison
Month-end close catches Connect leaks — too late to matter
Concession first: month-end reconciliation works — it finds these leaks reliably. The problem is arithmetic, not competence: by close day, payout windows have closed, and every finding silently downgraded from recoverable to claim to write-off. Latency, not rigour, is the variable worth attacking.
The decay curve
Modelled recoverability falls per payout cycle between refund and discovery. Monthly cadence guarantees worst-case aging on weekly-payout platforms; the maths compounds quietly every single month.
The normalisation trap
Each close books write-offs as routine; routine becomes budgeted expectation; the leak becomes an invisible tax nobody challenges. Breaking normalisation requires upstream visibility rather than better close discipline.
The transition plan
Keep close exactly as-is; add continuous detection feeding it pre-classified causes. Close gets faster because mysteries arrive solved; recoveries improve because findings arrive fresh.
Common questions
Do we still need close adjustments?
Yes — but entries now carry reversal ids and cause codes instead of mysteries.
Does real-time mean constant alerts?
No — thresholds you set govern immediacy; digests absorb noise below them.