(Quick disclaimer) I'm a vendor, working with data for essentially the whole retail used market.
I wanted to see if Day's Supply or MDS or however you'd call it is accurate, so I backtested it off 2 million listings in September! Basically I wanted to see if it was worthwhile to switch to a different metric for my own market analysis since I kind of saw that when there was a glut of supply of a certain car locally, MDS would go crazy high, and it didn't give me realistic stability...
So, for results:
for each car, I calculated MDS for its model and age range the week before it was listed and tracked until delisting. If a car moved to another lot, I kept counting it as well. Here's a table covering 1.3 million cars listed in August and followed for 60 days:
Median actual days were a lot smaller than days' supply in every bracket. Same pattern when I reran it over 7 weeks and 2 million cars.
Interestingly, older cars had an even bigger gap. Cars 8-12 years old showed 43 days' supply but delisted by day 16.
I used the industry standard calculation (active listings ÷ (sold in last 45 days ÷ 45)) and accounted for dealer-to-dealer transfer and (at least partially) wholesale/auctions.
If you use vAuto, is this something you think about? Wondering if the variances behind this metric are common knowledge.
I developed my own model to estimate the more realistic "days 'till gone" for a Chrome extension called Corsa - it scans any auction website for vins and shows retail value, turn for that car locally (as well as any spikes/dips in supply), whether car was at a dealer recently, and max bid. No-card 7 day trial: Corsa
I wanted to see if Day's Supply or MDS or however you'd call it is accurate, so I backtested it off 2 million listings in September! Basically I wanted to see if it was worthwhile to switch to a different metric for my own market analysis since I kind of saw that when there was a glut of supply of a certain car locally, MDS would go crazy high, and it didn't give me realistic stability...
So, for results:
for each car, I calculated MDS for its model and age range the week before it was listed and tracked until delisting. If a car moved to another lot, I kept counting it as well. Here's a table covering 1.3 million cars listed in August and followed for 60 days:
| Days' supply when listed | Median days until gone | Gone in 30 days | Gone in 60 days |
|---|---|---|---|
| 30 | 16 | 67% | 86% |
| 41 | 21 | 61% | 83% |
| 51 | 25 | 56% | 78% |
| 63 | 29 | 51% | 74% |
| 85 | 35 | 44% | 68% |
Interestingly, older cars had an even bigger gap. Cars 8-12 years old showed 43 days' supply but delisted by day 16.
I used the industry standard calculation (active listings ÷ (sold in last 45 days ÷ 45)) and accounted for dealer-to-dealer transfer and (at least partially) wholesale/auctions.
If you use vAuto, is this something you think about? Wondering if the variances behind this metric are common knowledge.
I developed my own model to estimate the more realistic "days 'till gone" for a Chrome extension called Corsa - it scans any auction website for vins and shows retail value, turn for that car locally (as well as any spikes/dips in supply), whether car was at a dealer recently, and max bid. No-card 7 day trial: Corsa
