Amazon launches Part Finder, using technology acquired from Partpic in 2016 • Hacker News Board

Re: Amazon launches Part Finder, using technology acquired from Partpic in 2016

#41

Post by mmt » Sat, Sep 15, 2018, 5:42 AM UTC

anonymous5133 wrote: post #10

inferiorhuman wrote: post #3

I remember when Amazon bought CDNow ages ago. CDNow let me easily find specific CDs by artist and title AND they even had a great recommendation algorithm. I discovered some great new music that way. In 2002 Amazon shut down CDNow with the idea that you'd just use Amazon.com. Unfortunately, search has been one of Amazon's weakest features. Sixteen years later and I still miss CDNow. Amazon rarely shows exactly what I…

Imo most of amazon's search is slopily done. I find it mostly frustrating to use to be honest. It only works good if you know exactly what you want otherwise you just get results of products that are of questionable quality.

> It only works good if you know exactly what you want

It's worse than that, though. I already know exactly what I want (as did the parent commenter, I'm sure).

For their search to work well, you also have to know exactly how the seller happens to have described it.

Except, it's even worse than that, because, even for slightly different versions of the same thing, the description can differ significantly enough to confuse Amazon's search. This is particularly true if there's more than one seller, but I've seen it happen even with the same seller.

Usually, I find their search works, in that, with enough attempts and brute-force, manual perusal of the results, I can find what I wanted to buy (assuming it's there to be found), but I would not characterize it as working well, not for the use case of parts (or even computer/electronic components).

Re: Amazon launches Part Finder, using technology acquired from Partpic in 2016

#42

Post by swsh » Sat, Sep 15, 2018, 5:43 AM UTC

inferiorhuman wrote: post #38

swsh wrote: post #35

Earlier quoted context omitted.

Fastener classification does tend to be a bit vague at best. Generally I find generic terms are used for low grade products, and often for products which you may struggle to find proper material certs for. Fine for general use home use, but no good in industrial engineering applications. Id imagine that the problem is that software engineers don't understand hardware engineering, and it's actually not a space where a…

> Also from a metric country, things like number and wire gauges are odd to me as well as I only occasionally have to deal with them. Well wire gauge is typically not how one would identify a bolt. Number wire gauge would typically be referred to by the units (either AWG or mm^2). > If you are looking for fasteners for specific industrial use then I'd assume you would have the expertise to select the appropriate fast…

Yeah wire gauge fasteners is definitely weird, not something I've used, but from a quick look they exist, although I'm not sure if they get much use.

Castle nuts are a bit of an oddity too. Id assume your friend is possibly using them for trailer bearing retainers or possibly some really old suspension items. Not even sure there is really even a specification the ones available here seem to adhere to.

They are something I've generally gone out of my way to engineer out of a lot of applications as there is better options available.

Re: Amazon launches Part Finder, using technology acquired from Partpic in 2016

#43

Post by craftyguy » Sat, Sep 15, 2018, 5:58 AM UTC

joe_the_user wrote: post #15

tsomctl wrote: post #5

While I don't want to be too presumptuous and say this is going to fail, it is definitely not going to defeat Fastenal or McMaster-Carr. Really, it comes down whether it's a generic part, or specifically engineered for a larger assembly. If it's the former, it's going to be significantly easier to use a website like McMaster-Carr where every item is completely organized into a hierarchy. You can quickly drill down an…

There are two questions: one whether taking a picture and getting a part to buy would be a desirable workflow for customers - and jeesh, it should be obvious that it would be. The other question is whether the process would work. That's a matter of tech, resolution of cameras and the design of the interface. It's easy to imagine the system displaying multiple options for things like the two nearly identical bolts. As…

I'm less worried about the accuracy of the prediction of their algorithm, than I am about their recommendations that involve vendors that make the thing as inaccurately as possible to skimp on cost. Maybe I've just been burned too many times ordering electrical components on amazon.

Re: Amazon launches Part Finder, using technology acquired from Partpic in 2016

#44

Post by craftyguy » Sat, Sep 15, 2018, 5:59 AM UTC

inferiorhuman wrote: post #32

tsomctl wrote: post #5

While I don't want to be too presumptuous and say this is going to fail, it is definitely not going to defeat Fastenal or McMaster-Carr. Really, it comes down whether it's a generic part, or specifically engineered for a larger assembly. If it's the former, it's going to be significantly easier to use a website like McMaster-Carr where every item is completely organized into a hierarchy. You can quickly drill down an…

In theory if you've got distance information and can keep the phone at a known angle in relation to the item differentiating between 27 and 29 mm shouldn't be an issue. What the photo bits will not be able to do without significant user interaction is determine other key properties of the fastener including material strength and likely finish. It's a cute gimmick though.

> determine other key properties of the fastener including material strength

Given a random screw, can anyone determine material strength using household items? If so, I would definitely like to apply this to some personal projects I am working on!

Re: Amazon launches Part Finder, using technology acquired from Partpic in 2016

#45

Post by swsh » Sat, Sep 15, 2018, 6:05 AM UTC

inferiorhuman wrote: post #40

swsh wrote: post #36

Earlier quoted context omitted.

Stainless I would argue belongs in material not in grade and class as these are different attributes. You can get grade 10.9 bolts in stainless, black, zinc plated and possibly in other materials and finishes. If it doesn't specify the type of stainless steel the safest thing to assume it's a low quality material which doesn't conform to any specific specification or at least lacks the appropriate accompanying paperw…

> Stainless I would argue belongs in material not in grade and class as these are different attributes. You can get grade 10.9 bolts in stainless, black, zinc plated and possibly in other materials and finishes. If it doesn't specify the type of stainless steel the safest thing to assume it's a low quality material which doesn't conform to any specific specification or at least lacks the appropriate accompanying pape…

We don't have the luxury of McMaster Carr here, but a quick search suggests that they have history and expertise in supplying engineering equipment which explains why they are so much better than Amazon.

It does seem like large tech companies tend to take less than stellar attempts into these sort of areas because they don't hire people with the appropriate expertise.

That said I appreciate the fact they don't for a number of reasons as it's definitely something I wouldn't want Amazon to step into and stomp the competitors in the field.

Re: Amazon launches Part Finder, using technology acquired from Partpic in 2016

#46

Post by DanBC » Sat, Sep 15, 2018, 6:47 AM UTC

joe_the_user wrote: post #15

tsomctl wrote: post #5

While I don't want to be too presumptuous and say this is going to fail, it is definitely not going to defeat Fastenal or McMaster-Carr. Really, it comes down whether it's a generic part, or specifically engineered for a larger assembly. If it's the former, it's going to be significantly easier to use a website like McMaster-Carr where every item is completely organized into a hierarchy. You can quickly drill down an…

There are two questions: one whether taking a picture and getting a part to buy would be a desirable workflow for customers - and jeesh, it should be obvious that it would be. The other question is whether the process would work. That's a matter of tech, resolution of cameras and the design of the interface. It's easy to imagine the system displaying multiple options for things like the two nearly identical bolts. As…

> one whether taking a picture and getting a part to buy would be a desirable workflow for customers - and jeesh, it should be obvious that it would be.

It's useful for people buying one or two parts. Why chase that market when most of these parts are sold by the thousand?

Re: Amazon launches Part Finder, using technology acquired from Partpic in 2016

#47

Post by dsnuh » Sat, Sep 15, 2018, 7:37 AM UTC

inferiorhuman wrote: post #3

I remember when Amazon bought CDNow ages ago. CDNow let me easily find specific CDs by artist and title AND they even had a great recommendation algorithm. I discovered some great new music that way. In 2002 Amazon shut down CDNow with the idea that you'd just use Amazon.com. Unfortunately, search has been one of Amazon's weakest features. Sixteen years later and I still miss CDNow. Amazon rarely shows exactly what I…

I used to work for Music Trader, which was a local San Diego used media chain that was purchased by CDNow. It's been forever since I've seen that name! Crazy to think they ended up being bought by Amazon. I was only there for like 8 months of their ownership, but it always felt like they were on the edge of going under. I recall the FoxPro based computer system they brought in was life changing for us.

Re: Amazon launches Part Finder, using technology acquired from Partpic in 2016

#48

Post by markholmes » Sat, Sep 15, 2018, 8:49 AM UTC

dsnuh wrote: post #47

inferiorhuman wrote: post #3

I remember when Amazon bought CDNow ages ago. CDNow let me easily find specific CDs by artist and title AND they even had a great recommendation algorithm. I discovered some great new music that way. In 2002 Amazon shut down CDNow with the idea that you'd just use Amazon.com. Unfortunately, search has been one of Amazon's weakest features. Sixteen years later and I still miss CDNow. Amazon rarely shows exactly what I…

I used to work for Music Trader, which was a local San Diego used media chain that was purchased by CDNow. It's been forever since I've seen that name! Crazy to think they ended up being bought by Amazon. I was only there for like 8 months of their ownership, but it always felt like they were on the edge of going under. I recall the FoxPro based computer system they brought in was life changing for us.

FoxPro. Now that’s something I haven’t heard in at least a decade.

Re: Amazon launches Part Finder, using technology acquired from Partpic in 2016

#49

Post by tcmb » Sat, Sep 15, 2018, 9:20 AM UTC

inferiorhuman wrote: post #3

I remember when Amazon bought CDNow ages ago. CDNow let me easily find specific CDs by artist and title AND they even had a great recommendation algorithm. I discovered some great new music that way. In 2002 Amazon shut down CDNow with the idea that you'd just use Amazon.com. Unfortunately, search has been one of Amazon's weakest features. Sixteen years later and I still miss CDNow. Amazon rarely shows exactly what I…

What you're describing is called "faceted search". For that to work properly, two things are needed: The product data needs to be structured properly, so that you can actually find the attributes on each product variant. This has to be consistent for every vendor, of which Amazon has many, of course. So they would have to define a common data format for every vendor selling M8 bolts, or convert their formats into a common standard, making sure all the attributes are there and can be located in the imports.

And second, someone needs to sit down and define which product attributes should be represented as facets for the search and displayed accordingly in the UI, and what input elements should be used to select the range of parameters (checkboxes, sliders, text box etc). This is not easily solved by algorithms which have no knowledge of the semantics of the attributes..

I suppose that a general-purpose, consumer-oriented site like Amazon doesn't put enough priority on something like M8 steel bolts. The facets for, say, computer monitors are much better, for example.

Re: Amazon launches Part Finder, using technology acquired from Partpic in 2016

#50

Post by snaky » Sat, Sep 15, 2018, 12:57 PM UTC

tcmb wrote: post #49

inferiorhuman wrote: post #3

I remember when Amazon bought CDNow ages ago. CDNow let me easily find specific CDs by artist and title AND they even had a great recommendation algorithm. I discovered some great new music that way. In 2002 Amazon shut down CDNow with the idea that you'd just use Amazon.com. Unfortunately, search has been one of Amazon's weakest features. Sixteen years later and I still miss CDNow. Amazon rarely shows exactly what I…

What you're describing is called "faceted search". For that to work properly, two things are needed: The product data needs to be structured properly, so that you can actually find the attributes on each product variant. This has to be consistent for every vendor, of which Amazon has many, of course. So they would have to define a common data format for every vendor selling M8 bolts, or convert their formats into a c…

For, say, computer monitors Amazon has a size, resolution, 'tecnology' (LED, OLED or LCD) and the 'brand' parameters for filtering. That's it.