How accurate were our 2025 scores? A full audit
We checked every 2025 growth score against the 12-month growth that actually followed, found a 0.61 correlation overall, and used the misses to justify two weighting changes in the January re-fit.
Why we publish this
A growth score is a forecast dressed up as a snapshot. If we don’t check it against what actually happened, it’s just an opinion with a number attached. Every January we re-fit the model’s weights, and before we do, we audit the previous year’s scores against the twelve-month growth that followed, suburb by suburb, across the full index. This is that audit for the 2025 vintage.
The headline number: 0.61
Across the index, the 2025 growth score correlated at 0.61 with the twelve-month growth that subsequently played out. That’s a meaningful relationship for a single number trying to summarise nine separate drivers, but it’s well short of a promise. A correlation of 0.61 means the score explains a substantial share of what happened next and misses plenty of the rest, mostly at the margins between adjacent bands rather than in wild, structural surprises.
Where the model got it right
The scoring held up best at the extremes. In our current published sample, suburbs scoring 80 or above averaged 17.5% twelve-month growth; suburbs scoring 70-79 averaged 11.7%. Both bands cleared the national average growth rate of 8.4% comfortably and consistently.
| Score band | Avg. realised 12-month growth | Hit rate* |
|---|---|---|
| 80-100 | +17.5% | 100% |
| 70-79 | +11.7% | 80% |
| 60-69 | +7.7% | 37% |
| Under 60 | +3.3% | 0% |
*Hit rate: share of suburbs in the band whose realised 12-month growth met or beat the national average of 8.4%.
Where it missed, and what changed in the January re-fit
The middle bands are where the model earns its keep and also where it’s weakest. Suburbs scoring 60-69 delivered a hit rate of just 37%, meaning most suburbs in that band fell short of the national growth benchmark even though the score implied “moderate” strength. That’s a real miss, not noise, and it repeated across enough suburbs to act on.
The clearest pattern in the misses: suburbs with tight listing scarcity but unremarkable income growth kept outperforming their score, while suburbs leaning on income-growth data alone kept underperforming theirs. That’s the direct evidence behind the January 2026 re-fit, in which we raised the weight on listing scarcity and trimmed the weight on income growth. Listing scarcity now sits at 15% of the score, the second-heaviest input after sales momentum; income growth sits at 7%, the second-lightest.
This year’s re-fit hasn’t been through a full twelve-month cycle, so we can’t yet report a 2026 correlation figure with the same honesty we’re applying to 2025 here. What we can say is that, within the current published sample, the relationship between score and growth is tighter than 0.61 on a same-period basis. That’s expected of any in-sample check and isn’t itself proof the re-fit will hold up once real time has passed. That verdict comes next January, publicly, the same way this one did, and we’ll report the misses as plainly as the hits.
The bottom line
A 0.61 correlation is a genuinely useful signal, strongest at the top and bottom of the scoring range and weakest in the middle bands, and the misses we found were specific enough to act on rather than dismiss. We’ve adjusted the weighting accordingly, and we’ll audit this year’s vintage with the same scrutiny in January.