Skip to content

Variance Analysis - Fundamentals

What It Is (General)

Variance analysis is the process of comparing predicted values to actual outcomes to measure forecast accuracy and understand the drivers of any differences. In cash forecasting, this means comparing forecasted cash positions, flows, or balances against what actually occurred.

What It Means for Our ICP

How Treasury Teams Think About It

Treasury teams distinguish between two types of variance checks:

  1. Daily/Short-term checks - Did payments go through? Is the balance where it should be?
  2. More operational, anomaly-focused
  3. "My assumption is everything works" - only flag exceptions

  4. Primarily about confirming treasury-initiated payments executed

  5. Weekly/Long-term analysis - Was the forecast accurate over time?

  6. More analytical, performance-focused
  7. Looking at 13-week historical accuracy
  8. Drives model configuration decisions

Typical Processes & Timing

ON's Approach (Source: 2025-11-11) 1. Daily: Check if funded accounts have correct balances 2. Weekly: Review forecast vs actuals for the week 3. 13-week: Historical accuracy review for trend analysis

Key insight: Treasury only verifies their own payments (from Kadiba); AP payments are tracked by accounting

Personio's Approach (Source: 2024-10-03) - Variance analysis between actuals and FP&A budget - Key challenge: Data validation takes too long before analysis can start - Currently delayed to T+11/T+12 when Workday reconciliation completes - Goal: Get to variance analysis as the starting point, not the end point

Sonder's Approach (Source: 2024-10-03) - Variance analysis is "the most critical" part of reporting - Keep forecast versions separate from actuals - don't just overwrite - Compare accuracy over time: "How accurate were we forecasting end of October at start of October vs on the 15th?" - Philosophy: "Every forecast I build, I assume I'm gonna be wrong. So it's just a matter of what did I get wrong?"

Tools They Use Today

  • Mental/manual analysis - No formal tools for variance analysis (Mentioned by: ON)
  • Dashboards - Custom views comparing forecast to actuals (Mentioned by: ON)

How They Talk About It

  • "Can I trust it?" - The fundamental question variance analysis answers
  • "Variance threshold" - The level of deviation that triggers investigation (e.g., 20%)
  • "Anomaly" - An unexpected variance requiring investigation
  • "Feedback loop" - Using variance data to improve model configuration
  • "Starting point should be variance analysis" - Ideal state vs current reality (Personio)
  • "Data validation" - The tedious prep work before analysis can begin (Personio)
  • "Keep them separate" - Don't overwrite forecast with actuals, track both (Sonder)
  • "Learn from it" - Use variances to improve future forecasts (Sonder)
  • "Timing difference" - Variance explanation for delayed payments/receipts (Sonder)

Granular Variance Analysis Value

Palm Philosophy (Source: Treasury Dragons 2025-12-09) Gurjit Panu emphasized proactive, granular variance analysis: "In order to have a very proactive variance analysis, the opportunities to go really granular is where you unlock quite a bit of value. In Palm, we alert you for any variances across any category across all of your accounts."

The netting problem: "A lot of cases, treasury teams are doing the variance analysis at a company or group level across balances. But then you have a problem when you have conflicting flows - one subsidiary overpaying by a million and one receiving a million more than expected. That nets the balance out to zero. So, as a team, you think your forecast is correct, but in reality, there might be some underlying issues under the surface."

This highlights why category-level and entity-level variance visibility matters, not just consolidated balance views.

Sources: view all