How it's built

An always-current model — not an opinion.

AI-supported, fully automated, and refreshed on every run: public data flows in, the models update, and the reports are written by code. No analyst hand-picks a number, talks their book, or hands you a view — which is exactly why this is information and forecasts, not advice.

In plain terms

Nobody hand-picks these numbers.

MPLC Research is a fully automated model, not a team of analysts. Public data flows in from official sources; the model processes it the same way every time; and the reports — every figure, chart and sentence of commentary — are generated by code. No human chooses a number to fit a view, writes a personal opinion, or is influenced by who's paying.

That's deliberate. An automated, AI-supported process is consistent (the same inputs always give the same answer), reproducible (anyone can audit how a figure was reached), always up to date (it refreshes from source on every run), and free of the bias that creeps into a human market call. It is also why everything here is information and forecasts only — not financial advice, and not a recommendation to buy, sell or hold anything.

AI-supported · automated · always up to date

The method

A two-layer model, in the open.

No black box. The method is published, the forecasts are tested against history, and every figure carries its source.

1

Top-down

The RBA cash rate and the bond market set the market yield each sector should price at — the macro anchor.

2

Bottom-up

The building-approvals pipeline and a leasing-sentiment signal show what supply and demand are actually doing on the ground.

3

The forecast

The two meet in a tested error-correction model, run thousands of times (a Monte Carlo) to produce a return range and an honest downside.

The data

Depth you can audit.

The edge isn't a louder opinion — it's more ground truth, every point sourced and checked automatically on each run.

149k
Live data points
Macro to precinct, refreshed each run
1,113
Precincts
Supply & demand, building-by-building grain
70+
Sourced series
RBA · ABS · Census · APRA · Treasury
23/25
Pass auto-QA
Flags disclosed, never hidden

Every series traces to a named public source — the exact RBA table, ABS dataflow or Census table — with how it's accessed, how it's derived, how often it updates, and why it moves commercial property. The model resolves from the national economy down through 6 states, 96 city sub-markets and 344 sub-market groups to 1,113 individual precincts. Freshness and plausibility are checked automatically on every run; anything that looks off is flagged, in the open.

Tested against history

A forecast is only worth its track record.

The model is back-tested the honest way — walk-forward and out-of-sample, only ever using data it would have had at the time.

+14%
Skill vs naïve
Lower error than assuming no change
76%
Direction right
Called the next move correctly
94%
Inside the band
Outcomes fell within the 90% range
199
Test quarters
Quarterly out-of-sample checks

A signal is only added to the model if it improves accuracy on this out-of-sample test — never because it sounds clever. The same discipline applies to the honest limits: the model is strong on yields and value, weaker on precise rents, and it is built to recognise regime shifts and shocks rather than to nail a single point forecast.

To be clear

What this is — and isn't.

What it is

  • A transparent, automated, AI-supported model
  • Tested against history, out-of-sample
  • Every figure traceable to a named public source
  • Consistent and reproducible — same inputs, same answer

What it isn't

  • A human analyst's opinion or market call
  • Financial, investment, legal or tax advice
  • A recommendation to buy, sell or hold any asset
  • Tailored to your objectives or circumstances

See it for yourself

Read the model, not a middleman.