Add Finance Concepts

2026-03-21 07:30:08 +00:00
parent 7ce79148f8
commit 227f3bce6d

83
Finance-Concepts.md Normal file

@@ -0,0 +1,83 @@
# Finance Concepts
This page explains the FP&A domain concepts implemented in the engine. It's useful background for developers integrating with the API or extending the service layer.
---
## Variance Analysis
A variance is the difference between a budgeted amount and an actual amount for a given GL account, department, and fiscal period.
```
variance_abs = actual budget
variance_pct = (actual budget) / budget × 100
```
Whether a variance is good or bad depends on the **account type** — see Favourability below.
---
## Favourability
A cost going over budget is bad. Revenue coming in above budget is good. The engine handles both correctly using a `favour_high` flag on each GL account.
| Account type | `favour_high` | Favourable when... |
|---|---|---|
| Revenue | `true` | actual > budget (more revenue than planned) |
| COGS | `false` | actual < budget (cost of goods came in under) |
| OPEX | `false` | actual < budget (operating expenses under control) |
| CAPEX | `false` | actual < budget (capital spend under plan) |
| Headcount | `false` | actual < budget (salary/contractor cost under plan) |
The `status` field in variance responses is `"favourable"` or `"unfavourable"` based on this logic. It is not simply positive or negative variance — a cost account coming in -5% is favourable, while a revenue account coming in -5% is unfavourable.
---
## Budget Versioning
The engine stores the original budget and up to three forecast revisions as separate version records. This enables three-way comparison (original vs. forecast vs. actual) without overwriting any history.
| Version | Meaning |
|---|---|
| `original` | The approved annual budget set at the start of the fiscal year |
| `forecast1` | First intra-year revision (e.g. Q1 reforecast) |
| `forecast2` | Second revision (e.g. Q2 reforecast) |
| `forecast3` | Third revision (e.g. Q3 reforecast) |
All variance endpoints accept a `version` query parameter. Comparing `forecast3` actuals against `original` shows how far the year moved from plan; comparing against `forecast3` shows execution accuracy against the latest expectation.
---
## GL Account Types
Accounts are typed to drive both P&L rollup structure and favourability logic.
| Type | Description | `favour_high` |
|---|---|---|
| `revenue` | Top-line income | `true` |
| `cogs` | Cost of Goods Sold — direct costs of delivering revenue | `false` |
| `opex` | Operating expenses — Sales, Marketing, G&A, R&D | `false` |
| `capex` | Capital expenditure — depreciable assets | `false` |
| `headcount` | Salaries, benefits, and contractor costs | `false` |
---
## Fiscal Periods
Periods are stored as integers 112, decoupled from calendar months. This means the schema supports fiscal years that start in any calendar month (e.g. a company with a July fiscal year start uses period 1 = July, period 12 = June).
When loading actuals from an ERP, map your ERP's period numbering to the engine's period integers at ingest time.
---
## Actuals Idempotency
The actuals ingest endpoint is idempotent by the combination of `(fiscal_year, fiscal_period, department_id, gl_account_id)`. Posting the same combination twice updates the amount — it does not create a duplicate row. This makes it safe to re-run ERP export feeds without needing to truncate and reload.
---
## Alert Threshold
The `/variance/alerts` endpoint returns only GL lines where the absolute variance percentage exceeds a configurable threshold (default 10%). This is useful for building exception-based monitoring: rather than reviewing every line, finance teams see only the ones that materially deviated from plan.
The threshold applies to `abs(variance_pct)`, so it flags both over-budget costs and under-budget revenue.