Really nice to seem directed edge is 'language agnostic' and has provided a lot of pre-cooked interfaces for their API. PHP, Ruby, Python and Java ready to go.
It wouldn't be a big deal for people competent enough to use the API in the first place to cook something up but it will definitely help speed up adoption and cut down on debugging time.
You should ask wheels about how much of a pain it was to write all the language bindings ;-)
Agree that it is a great idea to release language bindings for an API.
The big question in my mind is what else can you do to increase API adoption? One of the biggest advantages that twitter had was their massive developer and API support and I haven't seen anything that talks to how would you go about increasing developer adoption? Make something people want for sure, but make something developers want is perhaps a smaller subset of that axiom?
It is definitely a big deal to write scalable, fast and a good recommendation engine. Item level collaborative filtering is easy to write - but anything beyond that will require significant effort.
> It is definitely a big deal to write scalable, fast and a good recommendation engine.
I know, I've invested $30K US in to a dutch company that does this for one special interest field. It took them quite a bit of time to get it working perfectly. One very big reason why I'm so impressed with Directed Edge. And why I think they'll make it big.
Yup, providing easy to use API for developers to do recommendations can be a breakthrough. But I believe DirectedEdge has a ton of sales job to do in terms of convincing developers to take it up. They would need case studies and whitepapers where they demonstrate how plugging in their engine increased revenues by X% and engagement by Y and what ROI can they expect.
What I'd like to see is sites like HN using tools like this. So that rather than see some kind of popular consensus about what makes a "good" posting, I see (or have the option of seeing) the kind of thing that I want to see.
About "wouldn't be that difficult": no, not given a good and appropriate recommendation engine. However, is there one? For example, in context, we're talking about DirectedEdge. It seems to be aimed primarily at product recommendations, not link recommendations. And I'm not sure these are really even the same problem.
According to my reading of the documentation I think the recommendation engine operates on the data set that you specify regardless of the content. It operates on link graphs and group recommendations not the data itself. Basically, the data that is represented is invariant. The operations it does is like a group based recommendation service.
Really awesome stuff and I can't wait to start playing with it.
Minor tiny bug report for that second link. Maybe it's just my Safari but where it says:
"Price
$9.00 USD per month*
Website
Directed Edge ExpressRex"
-- the link is broken, it starts with http:// 3x before the domain name.
Grml, yeah, that's a Shopify bug that I already mentioned to them. Every time you edit the template it adds another http:// in front of the URL, so they pile up if you forget to remove them with every edit. Just edited it again to remove that.
Nope -- you don't need a separate Directed Edge account. It's a totally frictionless thing where all of the billing for the DE + Shopify stuff is handled on the Shopify side. It's our first experiment on a reseller model where the app has access to our admin API which can create accounts automatically.
ok - I clicked around again, and see you have a different pricing page on Directed Edge.
The Venture Beat article mentions increases of 20-30%, but this page talks of recommendations only, without personalization. What sales lift can we expect out of related products only?
How much you'll see just from related products varies significantly based on the type of products you've got and how much potential there is for selling additional products based on related products. i.e. if you've got a bunch of disjoint products, you're probably not going to see huge lift, but if you've got products that tend to compliment each other (and, of course the data we can infer that from), you'll see a decent bump.
That said, we'd like to have personalization, from the line in our FAQ:
We've been talking with the Shopify team about making personalized, e.g. Welcome back Bob, here are some things we think you'll like, possible. The Directed Edge recommendations engine supports that type of recommendations, though currently we don't have the ability to get the necessary data from Shopify stores.
I think you guys should have links to some of your customers on your sales page, so that new customers can see how this actually works on other sites(without having to visualize how they'd implement it on their own site)
Something like "________ is using DirectedEdge to recommend _____, "after installing DirectedEdge our sales went up 15%!!!omgWTFBBQ!" - CEO of ______
I have an idea that I've been planning. It's got a lot of little pieces, and one by one, I've discovered that someone else is already making them. It's pretty great living in the future.
I'd like to sign up for a trial, but I have to provide my credit card information first. That's awfully annoying. Is there any chance of removing that requirement?
On second thought, this may be a non-issue, since a huge dataset that is just product0-29389283923 and customer0-29389238923 is probably not that useful without product names.
