Rebuild the inventory model monthly, with a fourth method for stock ageing

[anthomakr]

Follow-up to #3, which shipped the three-method yearly model. Two changes on top of it.

1. One tab per method, and grouped assumptions

Assumptions were a flat list of 18 drivers and the three methods shared one section, so reading a single recipe meant filtering the other two out by eye.

Assumptions are now grouped in four sub-sections (revenue and margin, working capital terms, capital and financing and tax, inventory method drivers). Sub-sections group inside the tab rather than creating new ones, verified against the Excel export: still one Assumptions sheet. Each method gets a root section, so one tab per method, and pulls the drivers it reads through callups. The Assumptions tab stays the only source of truth: a callup mirrors a value, it does not own it.

2. Monthly, and a fourth method

The yearly model could not express stock ageing at all. On flat annual flows, a "last 180 days of purchases" rule is just a DIO of 180 in disguise: I checked, and the engine spreads a yearly flow evenly across months, so the mechanism would have been inert.

Rebuilt on a monthly grain with a real seasonal profile over 2024-2029. 2024 is a ramp-in year, not a forecast: it gives the trailing-12-month windows a full history so 2025 onward is clean, and it absorbs the working capital build.

COGS12 (trailing twelve months) is what every ratio method reads. A monthly COGS times a ratio would make inventory swing with the season instead of with the business.

Method D forecasts the flows and lets the stock follow, instead of inferring the stock from COGS. Under FIFO the remaining stock is always the most recent purchases, so age is a subtraction rather than a layer-by-layer simulation:

Stock older than 60 days  = MAX(0, inventory − purchases of the last 2 months)
Write-off (over 180 days) = MAX(0, stock before write-off − purchases of the last 6 months)

The finding

	Dec 2025	Dec 2026	Dec 2027	Dec 2028	Dec 2029
Share of stock older than 60 days	16%	16%	17%	20%	22%
Write-off	0	0	0	0	0

At normal purchasing the 180-day write-off never fires, and the stock ages anyway. A business turning stock in 60 days never holds anything for 180, so the accounting alarm stays silent for years while a quarter of the warehouse quietly gets old. Raise Purchase policy factor from 1.02 to 1.08 and two thirds of the stock is over 60 days with the write-off still at zero.

No ratio method can produce that column: A, B and C derive the stock from COGS, so by construction they cannot know how old it is.

Continuity

A, B and C reproduce the yearly model to the euro at every December, so nothing in the earlier framing breaks:

Method	Dec 2025	Dec 2029
A. Percent of COGS	4,315,200	6,849,537
B. Inventory days	4,322,192	6,255,285
C. Unit build	4,330,667	5,984,138
D. Ageing from flows	3,509,000	6,051,723

Spreads unchanged: 594,252 A vs B and 865,398 A vs C in 2029. Annual P&L totals identical (net income 2,475,000 in 2025 through 4,874,093 in 2029).

Verification

Balance check at 0 on all 72 periods and all four methods, re-run with the switch on 1, 2, 3 and 4. Cash never goes negative under any method (floor 2.1M under D, the most demanding). layerz_validate_model clean. The model is promoted to a Layerz template at user scope, not system.

The superseded yearly model is no longer referenced anywhere in the repo and can be deleted once this merges.

🤖 Generated with Claude Code