Skip to content

Variance Analysis: Priorities & Context

Date: 2026-02-17 Domain: Variance Analysis

Priority Ranking

This domain is tightly focused — one confirmed job with deep validation from many customers. The gap isn't in understanding what's needed but in completing the feature.

Top Priority (Confirmed — 8 sources)

1. Investigate variances between forecast and actuals to create accountability - Sources: ON (5 sessions), Personio (2 sessions), Sonder, Volvo Cars (Confirmed) - Current status: Partially solved — core drill-down works, but key gaps remain - Why it matters: Sonder: "That's the most critical, right? Like just taking the actuals and overwriting your forecast is not smart." Variance analysis is how treasurers learn and improve. Without it, forecasting is just guessing with extra steps. Volvo: "If you see a variance, you pay more attention."

Solved outcomes: - Root cause identification ✅ - Category-level drill-down ✅ - Transaction-level drill-down ✅

Unsolved outcomes: - Show forecast improvement to management ❌ - Automated variance explanations ❌ - Export/sharing for stakeholder communication ❌ - Model performance recommendations ❌ - Permanent vs timing difference tracking ❌

What's Missing

The current Variance Analysis feature answers "what happened?" but not:

  1. "Why did it happen?" — automated explanations (timing shift? categorization error? genuinely unexpected?)
  2. "Is it getting better?" — accuracy trends over time, forecast improvement tracking
  3. "Who needs to know?" — export/sharing workflow for stakeholders (FP&A, CFO)
  4. "What should I change?" — model configuration recommendations based on variance patterns

Next Steps

  1. Accuracy trends over time — Federico at ON is building this in Excel (WMAPE tracking). We should own this workflow.
  2. Stakeholder communication — export/sharing so treasurers can send variance reports to CFO/FP&A without screenshots
  3. Model recommendations — "Based on this variance, consider switching account X to manual forecasting"