Recommendation - criteria - Media with Similar Attributes - Content Similarity - expected lead time and keywords comparison logic

[Jerome1978]

Hi,

I have few questions around the Media with Similar Attributes Algorithm

1) once I have updated a criteria it takes between 30mn (best case) to 1h (average) up to 1h30 (worst case) for the recommendation to be ready - it really make the tweaking of the weighting of the attribute very time consuming

2) The scale goes from Baseline to 20x can you please

• Confirm that baseline means that the attribute must exist
• does 20x mean that the attribute must be an exact match? or does it mean that the same keyword is used and put a weight of 20 (points?)

     I have tested a very simple attribute that only contains one word and I can't understand the logic behind it -  the recommendation completely ignored

     Try with another simple attribute still one word and it seems to work. 

3)    Do I need to add the entity field that need to be ignored? why is that?

Many thanks,

Jerome

[MihneaD]

@Jerome1978,

You've raised some good questions. I like to provide documentation where I can so here you go:

1.) For more on "Expected criteria processing time" see: https://docs.adobe.com/help/en/target/using/recommendations/criteria/create-new-algorithm.html#proce...

2.) For more on "Difference betweeen Attribute Weighting and Content Similarity" see: https://docs.adobe.com/help/en/target/using/recommendations/recommendations-faq/recommendations-faq....

3.) Ignored means it has no bearing on the results. If you dont' want to use that enity.variable you can remove it by cliking the little: "X" to the right of it.

If this satisfies your query please like, mark as helpful and mark as answer. Otherwise lets keep the conversation going. Hope you have a wonderful day.

Mihnea Docea | Technical Support Consultant | Customer Experience | Adobe | (:: 1 (800) 497-0335

[Jerome1978]

Hi Mike,

As usual many thanks for your quick reply.

1) With "Media with Similar Attributes"  there is no behavioral source, so this only based on content tagging - but our catalog is really not that big and it seems a quite a long time.

2) Yes I understand the difference one is based on Static value the other one is based on the recommendation key (current item) so as I mention I only a attribute with one keyword and in none of the recommended content there is this keyword (and I know that this keyword is broadly used - fixed term - in our catalog.

What kind of text comparison does it do?

3) If I need an attribute I add them to the list, if I don't need one I would just not add it -> in which case do I need to put ignored? do I need to add all the field I don't need as ignored?

The Inclusion Rules with Entity Attribute Matching works perfectly so but without being able to score / rank the content it's pretty much random.

Kind regards,

Jerome

[Jerome1978]

it will seem very odd - but I am just asking: Do I must use "Content Similarity" and "Attribute Weighting" together for example:

[MihneaD]

@Jerome1978,

Others feel free to chime in if you have further clarification on any of these. I'll do my best to address below:

1.) Per the KB article previously linked it can take up to 12hrs. (Search the article for the word hours)

2.) Per the KB article previously linked if you based it off of something like the word: "Adidas" it would make it more likely that all things containing "Adidas" are recommended such as: "Adidas Sneakers" "Adidas T-Shirt" etc...

3.) Why would I use ignored vs just remove it. I'm trying to think of a good reason and the only one I can think of is if you temporarily want to ignore it but are planning on using it again in the future.

4.) Do I have to use content similarity and attribute weighting together. I'm not certain on that one it would depend on what the UI is allow you to do.

Mihnea Docea | Technical Support Consultant | Customer Experience | Adobe | (:: 1 (800) 497-0335