Request new feature

[turbodonkey]

Ok want to help, cause this is amazing!

Firstly, i've installed this on a dev store, works great. BUT I really want to pre-load every product with the maximum slots. So for example, I've set that I want 5 products shown per product detail page. Currently without having a 'history' nothing is shown.

What I'd like to propose is a simple "Day 1" init which loads the positions. It would in order of priority look like this:

1. Load current manual "Related Products" (if that is more than the 5 i need, stop)
   2a. Load similar named products (say my product name is ACME | Widget - A, I would text search for other products very similar to this, specifically if there exists ACME | Widget - B and ACME | Widget - C I want to give them high priority in the prepositioning) (if that is now more than required, stop)
   2b. For special marks trawling through entire order history and understand what was bought together and apply a weight for these associations
2. If there is still nothing found above 1 and 2, then pick from the category the product is in (this can be a set a categories of course, and potentially give higher priority to brand / manufacturer)

Once everything is loaded the current code will make it dynamic and hopefully positions fine tune over time. But its great to be able to pre-load at the start without having a history. 1 and 2a would be a great start!!

I want to help, because this is a great module which bucks the trend against all these SaaS players.

[vovayatsyuk]

Hi!

I understand your request, but I'm not sure about the ways of implementing this great feature.

1. Will work nicely, if product page does not have the related block. Otherwise you'll receive
   2 blocks with the same content.
2. 1. This is really interesting way, but it's possible by using the full text search only.
   2. Maybe
3. It will not provide relevant results

So, for now, I think that 2b option is a golden mean among others.

[turbodonkey]

Ok so I ran a little test on the weekend, using gensim - https://radimrehurek.com/gensim/index.html. I created a corpus of all my products and their descriptions then using the above python library created an LSI index of all products. I then ran a similarity search based on the product title, and was easily given 30-50 related products and suggestions making this algorithm really easy to fullfill 2a. Only issue with this code is that it runs python, and is completely outside the coding domain of this module, which makes it really hard to produce an all-in-one package here.

Now that I have this database, I was going to go through the entire product portfolio and pre-load a weighting for these suggested / related products so they appear in the mostviewed block. BUT instead of just using the existing 1 variable weightings, I'm keen to introduce another weighting column in the table which is a weighting given from this LSI similarity score, and it this variable will not be effected by views over time, its essentially static.

My aim for 1. would be that instead of using the related product field in your xml layout, you would replace it with your mostviewed block.

[qx54]

+1

Also, it would help if the extension would inject into the native upsell, cross-sell and related products blocks.

Regarding 3): I think this would still provide somewhat relevant results, still better than having no related products.

[steig]

The extension could ship with a sql script that one has the option of running that would query the report viewed index table joining on customer ids and adding to the process log table.

insert into yavva_alsoviewed_log(product_id,related_product_id)
select 
    cpe.entity_id 'viewed_product',
    rvpi2.`product_id` 'link_product'
from catalog_product_entity cpe
inner join `report_viewed_product_index` rvpi on rvpi.`product_id` = cpe.entity_id
inner join `report_viewed_product_index` rvpi2 on rvpi2.`visitor_id` = rvpi.`visitor_id`
where rvpi.`visitor_id` is not null and cpe.entity_id <> rvpi2.`product_id`

[vovayatsyuk]

👍 Great idea.

I will add it as "Import relations from magento reports" feature.

[steig]

Only problem is depending on the size of those reports you might need to process the relations in batches.