13% of NSW suburbs showed fake price cliffs in published data

If you charted recent sales in a NSW suburb straight from the Valuer General's bulk data, there's a good chance the line fell off a cliff in the last month or two. Across the ~4,000 suburbs we track, about 13% showed one of these cliffs in the raw feed. Almost none of them are real.

What's actually happening

NSW Valuer General sales records are published after settlement, not at exchange. Settlement typically runs four to six weeks behind the agreed sale, and the bulk dataset lags further still. So at any moment, the most recent month or two is radically under-counted: only the fastest-settling, already-recorded sales have shown up.

When you compute a median from that thin, biased set of recent sales, you get a number built from a handful of transactions — often skewed low or simply unrepresentative. Plot it next to a full month and it looks like prices collapsed overnight. They didn't. The month just hasn't finished arriving.

Why a naive median makes it worse

A median is only as good as its sample. With three or four recorded sales in a partial month, a single cheap unit or one delayed prestige settlement swings the figure wildly. The result is a chart that screams "crash" or "spike" at the right-hand edge — exactly where a reader's eye lands when they're trying to judge the current market.

This isn't a NSW-specific quirk. Any settlement-lagged sales feed has it. It's just rarely corrected, because correcting it means admitting your most recent data point is the least trustworthy one.

The fix

Ordus handles it in two parts:

1. Null the unreliable tail. For each suburb series, we drop the trailing months whenever the recorded sale count falls below a noise floor — roughly the lower of three sales or 30% of that suburb's trailing-12-month average volume. If the month is too thin to trust, we don't plot a median for it at all.

1. Mark what's still settling. On the price charts, the most recent stretch is drawn under a "data settling" band, so it's visually obvious that those points will firm up as more sales are recorded.

The headline "current median" shown across the terminal is a trailing-12-month rolling median, not a single calendar month — which sidesteps the thin-month problem entirely for the number most people actually read.

Why it matters

A buyer or investor glancing at a suburb chart shouldn't have to know about settlement lag to avoid being misled by it. The fake cliff is the single most common way raw government sales data lies to a casual reader. Filtering it is unglamorous, but it's the difference between a chart you can trust at a glance and one that quietly punishes you for trusting it.

Full detail on how every Ordus number is derived is on the methodology page. You can see the corrected charts on any suburb — try Manly or Parramatta.