4 ms·

What would be best paths through product space? My first thought was “doesn’t Amazon already organize things independent of vendor?” But what we’re really look

by GavinB 17y ago

What would be best paths through product space? My first thought was “doesn’t Amazon already organize things independent of vendor?” But what we’re really looking for here is a recommendation engine, and Amazon really isn’t very good at that. As I see it there are three main steps which should be done in order. The more you know about ahead of time, the further you can skip ahead. Here are the steps: Step 1. I don’t know what I’m looking for. Maybe I’m looking for a good science fiction book. Or a gift for a 14-year-old niece. A deep gift recommendation engine alone could probably make a good startup. The engine should ask some probing questions and give recommendations—either specific products or product categories, depending on the type of question. This experience mimic the experience of having an experienced salesperson ask questions and point you towards good options. A beautiful browsing option would be a great way to conclude this step. Step 2. I know what I want, but I haven’t decided which brand or model to get. Here we need a deep comparison. The ability to rank priorities would be really helpful. Let’s say I’m buying a laptop. The site should ask me how much I care about weight, screen size, processing power, etc. It can then recommend models that fit within my budget. There’s a huge amount of domain-specific knowledge required to get this right for a variety of products. Step 3. I know what product I want. Here we need to give comparisons among vendors. Price, shipping speed, return policies, and other general customer feedback are important. This is where it helps to be a trusted third party. This problem is a huge undertaking, which is why no one has come close to solving it yet. Frankly, unless the domain is very restricted, I don’t think a three person team will be able to do enough specific programming and testing to make it work. The principles and mechanics would best be discovered in a very narrow marketplace, and one full of confusion. Electronics would be the natural stating place, but a less crowded market might also work.

OmarIsmail 17y ago
The problem with steps 2 and 3 is that most of the time people aren't really sure they know what they want. They may have narrowed things down to a category, such as laptops as you mentioned, but even then you'd be hard pressed to find someone that absolutely KNOWS they want a laptop. Maybe a netbook + desktop would be a better combo? Or an iMac-style nettop. Furthermore, at any given time people are usually in the market for more than one thing, all at various stages of the buying cycle. So now if we can simplify and treat all cases as "I don't know what I'm looking for" then follow-up to that is "I'm looking". In that case the experience can be optimized around the looking and searching. I think the key concept in searching is, unsurprisingly, speed. Let me look at a lot of products quickly, let me research an individual product quickly, jump back out quickly, jump between categories, jump between different lists, sorts, filters, etc. And I think pg is missing something, as the path doesn't end when a product purchase is made. You have all the nitty gritty of the actual transaction + shipping, but then you have post-purchase life as well. Most stores have you buy the product and then have you go somewhere else for support. Sure you can review the product on the store, but what if you want to hack it? or you have a problem? or you want to upgrade? etc etc. Is it up to the manufacturer at that point? What if you can't find the manufacturer's support resources? We're doing some interesting things here at ProductWiki, but our goal is to create a great base of content and data. I imagine that one day a company will take the open data to create a really nice and innovative shopping experience.
- GavinB 17y ago
  You're totally right. Guiding people in the choice between even just the various computing options is a major task. This is just a big, big nut to try to crack.. To anyone attempting it, I'd suggest again starting in a very tightly focused domain. My pick would be cameras -- the classic "whattheheckdoIwanttobuyIhavenoideawhatthesenumbersmean" product.
extension 17y ago
If there's too much structure for the developers alone, then we need to enable the community to contribute that structure, in addition to content. But can the application be genericized enough to cover all of the shopping domains without being completely washed out? Think about the different kinds of web content that you read when making buying decisions. Off the top of my head, there's recommendations/searches (to discover things), consumer reviews/ratings (to evaluate them) and expert "buying guide" type articles (to gain a deeper understanding of a non-trivial decision). Currently, using all of these sources involves a lot of Googling and jumping from site to site. What if a single site could encapsulate all of them? This site would aggregate prices and host reviews of individual products. Products could also be tagged as belonging to any number of "aisles" of varying degrees of specificity like "Sporting Gear", "Wall Sconces", "Gifts for Mom", "2 Inch Tapered Drywall Screws", "Eco Friendly", and so on. The aisles would be user created and could be just about anything. Individual users can have their own personal aisle for things they own or want. Stores can have their own aisle too (which would be a likely area for monetization). Each aisle would have a hub page showing the highest rated products, sub-aisles and related aisles, external links, and user written guides, faqs and whatever else they want to write. The idea would be to make the content model flexible enough to adapt to the needs of any special domain and to the different situations consumers find themselves in.

by GavinB 17y ago

OmarIsmail 17y ago

The problem with steps 2 and 3 is that most of the time people aren't really sure they know what they want. They may have narrowed things down to a category, such as laptops as you mentioned, but even then you'd be hard pressed to find someone that absolutely KNOWS they want a laptop. Maybe a netbook + desktop would be a better combo? Or an iMac-style nettop. Furthermore, at any given time people are usually in the market for more than one thing, all at various stages of the buying cycle. So now if we can simplify and treat all cases as "I don't know what I'm looking for" then follow-up to that is "I'm looking". In that case the experience can be optimized around the looking and searching. I think the key concept in searching is, unsurprisingly, speed. Let me look at a lot of products quickly, let me research an individual product quickly, jump back out quickly, jump between categories, jump between different lists, sorts, filters, etc. And I think pg is missing something, as the path doesn't end when a product purchase is made. You have all the nitty gritty of the actual transaction + shipping, but then you have post-purchase life as well. Most stores have you buy the product and then have you go somewhere else for support. Sure you can review the product on the store, but what if you want to hack it? or you have a problem? or you want to upgrade? etc etc. Is it up to the manufacturer at that point? What if you can't find the manufacturer's support resources? We're doing some interesting things here at ProductWiki, but our goal is to create a great base of content and data. I imagine that one day a company will take the open data to create a really nice and innovative shopping experience.

GavinB 17y ago
You're totally right. Guiding people in the choice between even just the various computing options is a major task. This is just a big, big nut to try to crack.. To anyone attempting it, I'd suggest again starting in a very tightly focused domain. My pick would be cameras -- the classic "whattheheckdoIwanttobuyIhavenoideawhatthesenumbersmean" product.

extension 17y ago

If there's too much structure for the developers alone, then we need to enable the community to contribute that structure, in addition to content. But can the application be genericized enough to cover all of the shopping domains without being completely washed out? Think about the different kinds of web content that you read when making buying decisions. Off the top of my head, there's recommendations/searches (to discover things), consumer reviews/ratings (to evaluate them) and expert "buying guide" type articles (to gain a deeper understanding of a non-trivial decision). Currently, using all of these sources involves a lot of Googling and jumping from site to site. What if a single site could encapsulate all of them? This site would aggregate prices and host reviews of individual products. Products could also be tagged as belonging to any number of "aisles" of varying degrees of specificity like "Sporting Gear", "Wall Sconces", "Gifts for Mom", "2 Inch Tapered Drywall Screws", "Eco Friendly", and so on. The aisles would be user created and could be just about anything. Individual users can have their own personal aisle for things they own or want. Stores can have their own aisle too (which would be a likely area for monetization). Each aisle would have a hub page showing the highest rated products, sub-aisles and related aisles, external links, and user written guides, faqs and whatever else they want to write. The idea would be to make the content model flexible enough to adapt to the needs of any special domain and to the different situations consumers find themselves in.