What the recommendation history got wrong in Q1

The whole point of the recommendation history is that we grade ourselves on our own dates, publicly. In Q1 we closed four horizons. One held up cleanly, two were partial, one missed. That's not a great run. It's also why this section exists.

The miss: we forecast a 4-6% 12-month median uplift on a north-shore suburb in Jan 2025. Actual outturn was 1.9%. The flaw wasn't the upside case — the macro tailwind was real — it was that we under-weighted a specific supply signal (an approved DA on a large adjacent site that added competing stock mid-horizon). Good post-mortem question: how do we fold DA pipelines into the forecast earlier?

The partials (yield calls on two mid-ring QLD suburbs): we had the direction right but overshot magnitude. Both suburbs moved, just less than we said. The fix is tighter — the confidence intervals we were publishing were too narrow. We've widened the default band on yield claims from ±1.0% to ±1.5% as a result.

The win: a 2025 rate-stress call on a Sydney-ring suburb held up exactly. That's one call out of four, so we're not writing home about it. The value in the ledger isn't the wins — it's that the misses are visible, which is the only way they ever get learned from.