Suburb Trajectory

Multi-quarter price trajectory for a suburb (optionally per type), expressed as an annualised compound growth rate from a log-linear regression on quarterly medians. Replaces the prior single-year YoY delta which was noisy in low-volume suburbs.

Math ────

1. Bucket sales into quarterly medians over the last 24 months.
2. Need ≥3 quarters with ≥10 sales each to compute a suburb estimate.
3. Linear regression: log(median_t) = a + b·t (t in years from earliest quarter). annualised_growth = exp(b) - 1.
4. R² = 1 - SS_res/SS_tot for confidence.
5. If suburb fails the threshold, fall back to the LGA's trajectory and report fallback_used=True so the caller can surface the degradation honestly.

Why this matters ──────────────── SEQ markets in 2025-26 are running +15-25% YoY. The pricing engine used to time-adjust comparables using a single 12mo-vs-prior-12mo delta — fine for high-volume suburbs, noisy in small ones (a 30-sale suburb's median can swing ±10% from one outlier). Quarterly regression gives 4-5 data points and a confidence measure.