Ecommerce logging and recommendation system


#1

Howdy,
since the introduction of collecting ecommerce data (logging of purchased products) it would be great, to build something like product recommendations via the API.

I’ve setup a simple test case where I pull the Top-Seller from yesterday and display it on the shop pages.
I feel this is easier and faster than programm this by my own, because it’s already there in statistic data.

Now I read a lot of product recommendation in style of “Customers Who Bought This Item Also Bought…” and I think it’s not an easy task to build something like that for a webshop with 60.000 products and 1000 orders a day. (it’s possible with precalculation, but it’s heavy and I am not able to implement this by myself to be honest :S )

So if PIWIK is about statistical data, this would be a nice addition. What do you think?
Sure, the precalculation has to be made too, but I think PIWIK could handle this smart, because it can handle all the tracking data, so the logic is there?!.

This way, PIWIK would be much more attractiv not only for page tracking, also for other statistics.

It’s just an idea…

cheers m.


(Matthieu Aubry) #2

Thanks, this is a great suggestion. We are definitely interested in this kind of features and will let you know if anything is planned in the future. I added this in the ecommerce analytics wishlist: Ecommerce analytics improvements · Issue #2448 · matomo-org/matomo · GitHub


#3

oh man, thanks for the link to Liftsuggest
I searched for something like this since days. Just found google prediction API (for maschine learning, which costs money) and Mahout Apache extention, which I am not able to install B)-
Ended up in writing my own simple system with item2item recommendation.

OK, I don’t even use Google Analytics :slight_smile: But it shows new ways and it would be really really cool, if PIWIK add such a feature in future. So we can keep our data and have not send it to some other servers…

Thanks for positive response.