One reconciled source for Australian macro indicators

A research house was pulling RBA and ABS series from three different places, with inconsistent revisions and missing history. We built a single internal macro database.

The problem

The client's research team produced internal notes on the Australian economy. Each analyst maintained their own spreadsheets for RBA cash-rate history, CPI, unemployment, housing credit and yield curves. Because ABS and RBA revise historical series as methodology changes are introduced, the same economic variable would appear with slightly different values across three spreadsheets. When an analyst wrote "unemployment peaked at 5.8% in 2020", another analyst's file would show 5.9%. Small numbers, but they eroded confidence in every macro chart the team published.

What we did

We extracted the core macro series from the RBA statistical platform and the ABS Time Series Spreadsheets into a single internal database. For each series we stored both the current revised vintage and the vintage as it was published at the time, so the team could choose whether to write history with what we now know, or with what was knowable then. We built a small scheduled job that checks for new releases each week and appends them without overwriting prior vintages. We also normalised series IDs across RBA and ABS so that the same economic concept had one internal label regardless of source.

We then produced a short documentation file listing, for each series: source, frequency, seasonal-adjustment convention, and the known historical revisions. This file is what allows a new analyst to pull a number without asking the team "which spreadsheet is the right one".

Individual analyst spreadsheets retired after the single macro database went live

2 sources

RBA statistical platform and ABS time-series spreadsheets unified under one ID layer

Weekly

Scheduled release check that appends new vintages without overwriting prior ones

Dual

Both current revised and as-published vintages stored, so analysts pick their convention

The outcome

The team retired eleven individual macro spreadsheets. Their published charts now cite one internal series identifier, and disagreements about "what was the actual peak" are resolved by reading the vintage note rather than by arguing. The weekly scheduled job keeps the database current without analyst intervention, and the documentation file has become the onboarding reference for new researchers. The engagement scope was deliberately narrow — we did not build a forecasting model, and we did not add economic commentary.