Who this is for, and the moment
The exports tell you what the model does. They never tell you why. The twelve- branch IF chain looks like a mistake until someone explains that a mapping table the business could edit was the thing the first partner was avoiding. That someone has usually left. The notes are where the reasons stay.
The question
“Why is the account-to-driver mapping inside the Forecast Opex formula?”
The result
Forecast Opex: why the mapping is in the formula
The account-to-driver mapping lives in the IF chain because the first partner did not want a mapping table the business could edit. There is an open request to add a thirteenth branch for legal fees. Moving the mapping to a small lookup module is agreed in principle; nobody has had the time.
The action plan reaches the same place from the exports, without the reason:
Consider replacing the long chain of IF tests in the line item CAL03 Opex.Forecast Opex with a lookup table
5.0M cells · 16.0% of the model's effort · found by test 4.6, formulas with more than 10 IFs · done when every cell is equal before and after, the table holds every case the chain held, and the measured effort is no worse.
Three more notes a new architect reads on day one
The FX file stopped in November 2023; the hardcoded rate is the stop-gap
Import FX from Treasury file was always run by hand, outside any process, because Treasury's file is sometimes late and the actuals load must not wait for it. It has not been run since November 2023. The 1.27 USD rate hardcoded in CAL02 Revenue is the stop-gap since then, not a design choice.
Versions in FP&A, and Budget v2
Four versions: Actual, Forecast, Budget and Budget v2 DO NOT USE. Budget v2 was the mid-year re-cut in FY24 and was never approved. It is still selected by OUT02 Board Pack.Old Budget Revenue; ask Dan before removing the version, the board pack comparatives may still read it.
CAL05 Opex OLD is kept on purpose, for now
Replaced by CAL03 Opex in 2021 when driver-based opex went in. Kept until the FY22 audit file is closed, because the auditors' working papers reference its numbers. Nothing reads it; the cells and the calculation time are the price of keeping it. Once the audit is signed off it can go.
One decision, recorded
Caldergate FP&A → Workforce Planning
Owner note n-ff27724850 says Import from Caldergate FP&A - Assumptions into Inputs - Settings has not run since August 2022, HR analytics key the NI and pension rates by hand each year instead, and the import names make FP&A look like a feed into Workforce but it is not unless someone reinstates it.
19 notes on this estate: 5 purpose, 6 conclusion, 4 constraint, 3 decision, 1 procedure. 15 confirmed by the owner, 4 working assumptions. None flagged by the last export.
Evidence drawer · what each chip opens
- Owner noteForecast Opex: why the mapping is in the formulan-f2327b3be0working assumption
- Owner noteThe FX file stopped in November 2023; the hardcoded rate is the stop-gapn-04d0598b25working assumption
- Owner noteVersions in FP&A, and Budget v2n-eaeed7ba09user confirmed
- Owner noteCAL05 Opex OLD is kept on purpose, for nown-e6ddcfdc13user confirmed
- Action planConsider replacing the long chain of IF tests in 'Forecast Opex' with a lookup tableAction plan · #2
A decision note, typed and marked as a working assumption, with the open request and the agreed-in-principle fix. Next to it, the action card the tests produced from the exports alone: the same conclusion, without the reason. Then three more notes a new architect reads on day one, and the one decision recorded so far on this estate, a retired feed with its reason.
The evidence
Notes carry a type (purpose, constraint, decision, conclusion, procedure) and a status (confirmed by the owner, or a working assumption). A note scoped to an object carries that object’s citation on the active snapshot, so when a later export changes the object, the note is flagged for review. A note that is superseded stays, marked, so the old reason is still readable.
What to notice
- A working assumption is labelled as one. The mapping note is the owner’s best understanding, not a confirmed fact, and the chat answer that cites it will say so.
- The notes and the tests agree without talking to each other. Test 4.6 found the chain; the note explains it; the plan ranks the fix second. The next architect gets all three.
- Decisions are recorded on the object. The retired feed shows on the model diagram as a dead arrow, with the note that killed it, so nobody reinstates it by accident.
Related
Use cases: The estate on day one · The fix list · Is it safe to delete.

