I had an idea that I can't implement, but I think is good. It would be cool if we could have a place where people could enter their used parts sale prices and condition. We could then get the mean value biased by condition to give a market value for used parts. There is a lot more complexity that could be added to the algorithm, but what do you think of the core idea?

too many other variables such as location that dictate the price of the parts...in my opinion

and people tend to *hype* their product. you would need a independend inspector to come up with some remotely useful data ...

Andy

Yeah, I suppose. I still think it would be worthwhile to collect the data. To be taken with a grain of salt perhaps, but I think there's some value. I have no idea how much Targa tops are going for nowadays. Would be nice to know a ballpark figure when I'm thinking about buying one.

How about something that scans ebay and records only the prices of parts that actually sell.. then you'll know the high end of things anyways. I do that often by looking at completed auctions, but that only goes back 30 days, so rare stuff is never there. If you made a dtatabase, after a year it would probably be fairly complete. Condition is another factor though, but you might be able to do something like olympic judging and throw out the highest and the lowest.

I've looked at recording results of ebay auctions before. The problem lies in that there are no standards for descriptions and condition.

Someone may list "2.0 SS Heat Exchangers". Someone else may list "Exhaust Thingies". Without significant manual cleansing of the data, it wouldn't work.

There used to be a small handful of companies that did just that. One Ebay sued out of business, the rest made nice with Ebay and essentially stopped doing it. I used to work for one of these companies (AuctionWatch, which no longer exists). A real pity, too, as the data was excellent. An honest, unbiased report on what everything from beanie babes to water pumps were worth on the open market.

Now, Ebay helps you do much of this yourself. Set up a My Ebay account (free), and hit the "add to watch list" link on items you'd like to track the prices on. The price stays visible on your tracking page after it closes. 2-3 items are often enough to get a representative price range. I can tell you, for example, that a clean set of four four-lug Fuchs will go for $400-500 (no tires), whereas a single wheel is about $80 (so buy four singles and get the fourth one free!). Takes a long time to get the data this way, but prices don't fluctuate all that rapidly.

People who list things with bad titles also either don't sell the item (so they don't count), or they sell for a very low price (lucky buyer). The way this problem is solved is buy looking at a lot of items over time, and just averaging out the prices. You'll find that the prices tend to cluster around the average, and the outliers can simply be ignored. There's always some group of idiots who get into a bidding war and pay 2x the going price for something, and some idiot who lists a $200 item with no reserve and a $1 starting price and misspells the name so badly only one person finds it, and gets it for a buck. But that's not the norm.

Using common search terms as the "key" (like exhaust, not heat exchangers) will get good results, for the seller, the buyer, and the scanner. Sellers know this, which is why you'll see titles like:

"914 912 911 VW 2.0 Fuchs wheels, set of four Porsche Ghia"

Trying to hit the widest possible audience. For the buyer, using the smallest possible set of the shortest possible search terms will also gain the best results. Search for "914 wheel", not "Porsche 914 wheels", and you'll get all of the auctions where someone misspelled Porsche, or left it off. The search engine does a pattern match on sub-words (wheel matches wheels, but wheels only matches wheels).

This thought has crossed my mind several times before; in fact, it's part of what I do as a day job. One essential enabling technology in order to spider web sites and aggregate/analyze market pricing data, is a structured cataloging and taxonomy system. Otherwise you have no chance of comparing apples-to-apples pricing data, or of matching "want-to-buy" ads to "want-to-sell" ads, on an automated high-volume basis.

We all know that there are lots of sites trying provide price-comparison data for consumer electronics/computer products, but the results are very spotty. At most they help provide links to dealer sites to check out that you may not have been aware of beforehand. They are very poor at providing real-time availability, condition (for used items) or location data. In many cases they are "skewed" because they rely on partners or even paid sponsors to provide the content, i.e. sellers have to willingly agree to show their prices alongside their competitors' prices.

Pure auction sites like eBay are great for buyers and sellers of "one off" items, but are completely reliant on seller labor to classify, describe and track items for sale, which is highly inconsistent but also highly scaleable and cheap. Also, there is substantial effort and risk on the buyers' end since it is essentially an unregulated "swap meet" environment.

Unfortunately the kind of software, content-management, and sales/customer-service labor that is involved in doing this the "ultimate way", probably is not feasible for markets such as 30-year old p-car parts. It is however feasible for certain markets such as ET&M instruments, aircraft subsystems, machine tools, high-end computer servers/workstations, etc. Or possibly used cars/trucks! A few years ago when I studied this issue in detail, the ROI required an average transaction size around $5K, give or take.

The plus side is, for most of our "hobbyist" purposes, crude keyword-based searches and plain old brute force browsing (i.e. the eBay or BBS Classifieds model) accomplishes decent results. We just check the listings as often as we can, sometimes having to sift thru the same ones over and over, and pray that the item we need will show up! If you are looking to buy/sell something a bit rare and in-demand (i.e. a few 0's behind the price), however, this manual process becomes very non-optimal.

I mostly concerned (personally) with getting prices for non rare items. How much are rear control arms going for? How much are cigarette lighters going for? That sort of thing. I think with the community we have going here we could just do a user input system instead of a spider system. I think it's worth a try (even a crude demo) just to see what the data looks like after 6 months.