Financial planning & analysis.

What it does — 7 skills

✓ Financial reporting — Reproduce the numbers — ARR, margins, KPIs — from the warehouse.✓ Budget vs actual — Track budget vs actual and variances.✓ Cash & margin analysis — Watch cash, margins and KPI drift month to month.✓ Scenario planning — Model scenarios and forecast risk.✓ Forecasting — Forecast the quarter/year and flag risk early.✓ Analyze / read / write management reports — Read, analyse and produce monthly management reports. Learned the hard way, and the most important lesson first: a management report is a GOOGLE DRIVE DOCUMENT / PRESENTATION — the one the Managing Director writes their commentary in — NOT just data-lake numbers. So the real skill is (1) FIND the right report in Google Drive: search by company name + "management report", and verify the document actually belongs to that company and month before reading it — don't grab the most-recently-edited file globally (a classic wrong-doc mistake). (2) READ THE ACTUAL DOCUMENT (open the doc/slides, digest the text and any charts) to capture the MD's narrative — what happened, why, what's at risk, and what actions are planned — because that written analysis is the whole point and appears nowhere in the raw numbers. (3) Use the data lake only to CROSS-CHECK the figures, minding its traps: ARR-type metrics live under source "product" (not "financial"), a company-KPI-period is the SUM of component rows (never dedupe — that reads ARR as ~554K instead of ~56M), and the latest month is PROVISIONAL so trust actuals only through the last closed month. Then report BOTH the MD's narrative and the verified numbers — the story and its evidence together.✓ Data lake — Query the financial data warehouse (the "data lake") for ARR, margins, KPIs and trends. Hard-won learnings baked in: the true grain is (unit, source, KPI, year, period, Gr1, R12, reporting-year, date-key) — always SUM the components, never pick one row. NUMERIC values come back as big.js objects, not plain numbers, so a naive value-handler silently yields null — coerce with String(). Never diagnose a shocking swing as "duplicate double-counting": double-counting inflates, while a collapse toward zero is almost always an un-summed component or the provisional latest month. Verify against pfdw before trusting anything surprising. How to run it (the `lake` CLI, plus the analysis tools built on top): • lake explore --company Acubiz — list the KPIs, sources and periods available for a company • lake series ARR --company Acubiz --source product --chart — the ARR time-series (correctly SUMmed), ready to plot • lake growth ARR --company Acubiz — period-over-period and YoY growth • lake forecast ARR --company Acubiz — projection that prefers the forecast value for the provisional month • lake validate --company Acubiz — flag genuine double-loads and mark the provisional latest month • benchmark ARR — compare one KPI across the whole portfolio • portfolio — the fleet-wide KPI snapshot; waterfall / scenario / explain for bridges, what-ifs and drill-downs • lake manual — the deep guide (grain, provisional months, big.js NUMERIC) baked into the tool itself