Seoul small multifamily: is 32 days real, or a live-listing illusion?

I disagree slightly. A forced primary reason can hide combinations, such as a long lease being acceptable at a lower price but not at the current ask.
 
Then phrase the reason conditionally: “lease/price combination,” “condition at asking price,” or “location regardless of price.” That captures interaction without creating endless columns.
 
That is closer to how buyers actually filter. Lease length is rarely good or bad in isolation; it matters relative to price and the intended use.
 
Which returns us to Elena’s question: completed deals may not tell the identical story if buyers accepted awkward leases after discounts that active asking prices do not show.
 
So compare the last known asking price before disappearance with the original ask, while acknowledging that the actual deal price may differ or be unavailable.
 
Seller motivation could be approximated only through observable behaviour: quick cuts, repeated relisting, or no adjustment. Avoid assigning personal reasons that are not disclosed.
 
Agreed. Behaviour is evidence; presumed urgency is not. A long listing with no cut may indicate patience, overpricing, or something else entirely.
 
At this point the minimum viable rerun seems clear: deduplicate, define neighbourhoods, split price bands, classify condition and remaining lease, then separate completed, withdrawn, unresolved, and active.
 
I’d also freeze the cohort. New listings arriving during the exercise should go into the next cohort, not continually alter the denominator.
 
Freezing solves one problem but delays the practical view of today’s choices. Maintain a cohort analysis and a separate rolling shortlist.
 
That’s the right compromise. The cohort answers what happened; the rolling list answers what is available. Mixing them caused most of the confusion here.
 
How should listings crossing the ₩921,800,000 threshold after a price cut be treated? Original-price eligibility and current-price eligibility could produce different samples.
 
For the cohort, classify by original observed ask and record later cuts. For the buyer shortlist, use current ask. Again, two questions need two treatments.
 
Same at the ₩1,383,000,000 upper boundary. Otherwise a cut can make a stale property look like a newly relevant listing without showing its prior exposure.
 
I’d now avoid publishing one blended 32-day number. Present it as a starting observation, followed by separate durations for completed cases and current ages for active ones.
 
Would a simple count of properties in each outcome be more honest than averages if the sample is small? It avoids a precise-looking headline built on a few cases.
 
Counts first, then medians where the group is large enough to be meaningful. Readers can see whether one unusual property is driving the story.
 
You must log in or register to reply here.
Back
Top