Forecast accuracy: a layer-by-layer diagnostic

When the CRO's forecast misses by 20% three quarters in a row, everyone wants to fix the forecast model.

That's almost always the wrong fix.

The forecast model is Layer 4. The bug is almost always in Layer 2 (data model) or Layer 1 (source stack). Common patterns:

• ARR is double-counted because two custom objects both store it
• Stripe and HubSpot don't agree because the integration drops fields silently
• Lifecycle stages don't have entry criteria, so deals jump backwards

Fix the layer below first. The forecast model is downstream.