Variance Analysis - Current Solutions & Workarounds¶

Tools & Systems¶

What it is: No formal tools - variance analysis done through manual inspection
How they use it: Check account balances, compare to what was expected
Limitations: Time-consuming, not systematic, no historical tracking
Used by: ON

What it is: Structured variance analysis framework using WMAPE as primary KPI, with two dimensions: snapshot accuracy (per 13-week forecast) and horizon accuracy (performance by weeks-ahead)
How they use it: Weekly Tuesday captures of forecast vs actuals; WMAPE heatmap comparing Palm vs Kyriba across snapshots and time horizons
Results: 7-week evaluation validated Palm at ~8% vs Kyriba ~16-20%. Methodology being maintained going forward.
Limitations: Manual in Google Sheets; not yet in Palm platform. Most recent weeks have fewer data points.
Used by: ON (Federico, Giannis)
Source: ON (2026-03-10) — FC Validation deck

What they do: Manually check if funded accounts have correct balances each morning
Why: No automated alerts for payment failures or unexpected balances
Source: ON (2025-11-11)

What they do: Assume everything works; only investigate when told or when seeing huge variance
Why: Can't verify every transaction; focus on treasury-initiated payments only
Source: ON (2025-11-11)

What they do: Save forecast versions quarterly, keep actuals separate, compare over time
Why: Need to track forecast accuracy; learn from past variances
Source: Sonder (2024-10-03) - "We can go back and say, well, how accurate were we forecasting?"

What they do: Google local holidays when unexplained variances occur (zero collections, etc.)
Why: Holiday schedules across countries are hard to track; cause timing variances
Source: Sonder (2024-10-03) - "We sometimes just Google it, right?"

What they do: Reach out to payroll, FP&A, or other teams to explain variances
Why: Treasury sees the result but doesn't know the "why" behind it
Source: Sonder (2024-10-03) - "We have to go ask people that are the experts there"