Though a weak excuse, my break from blogging can be explained by a poor choice in reading. Apparently, most of my ideas are inspired by what I read and my current choice, The Decline and Fall of the Roman Empire, while interesting and enjoyable, does not lend itself to the auto market. Today I started a new book, or books (not quite sure yet), and this has, for some reason, led me to a thought about search features on classified sites.
There are millions of cars for sale on Autotrader and finding the right one to buy is actually quite a pain. Assuming that one knows the exact year, make, model, price, color, engine, transmission, zip code, radius, and mileage of a vehicle that one would like to purchase, searching on autotrader might be relatively easy. If one is just starting a car search though, this is unlikely to be the case.
Short of a design-based purchase, like the Pontiac Solstice would be for me (I like the taillights on the Solstice more than the Sky), most decisions are likely to be driven by a mix of research, emotions and desire for features. A good classified search would return the most relevant results – yes, a relevancy algorithm. People broadly starting out a search most likely have key criteria they are using to purchase a car, like, “I want a German luxury car” or “I need to be able to drive up to five people in some cases, usually just 2 or 3, it has to be comfortable on my commute, reliable, have enough power to merge into traffic, black exterior, with a sunroof, leather would be great, factory warranty please and my budget is….”
This might be akin to the needs discovery phase of selling with a lot up, and a well informed sales person would think through available inventory and come up with the most relevant vehicle(s) for the consumer to check out. Instead of functioning in an intelligent way, most classifieds return results with simple sorts (mileage, price, et cetera) and paid results are first. This makes sense for classified engines as they derive revenue not from sales but instead from dealers, deriving additional revenues from query results is just good business. A car dealer, on the other hand, might have different objectives. The goal of his site should be to return the most relevant cars for his customer.
Any algorithm, however simple compared to Google, would still need constant tweaking to achieve good results. What attributes are most important to a consumer, and hence should be weighted the most, are likely to vary across consumer but general patterns are likely to hold constant. Maybe one of the larger dealer site providers could give dealers this enhanced functionality… I know I would like it.