I'm using a Viewed This Viewed That criteria and finding that with no attribute weighting, a certain type of product is dominant and shows up about 65% of the time. I'm trying to use attribute weighting to "boost" the other product types. When I set those product types to 50% weights (and no weight for the dominant type), it does not seem to change the proportion of product types that I see in the resulting recommendations data. When I set the alternative product types to a 75% weight, they become too dominant, and my previously dominant type goes down too much (to about 3%). There's no option to set a weight in between 50% and 75%, so I'm wondering if there is something else I can do to achieve an intermediate result.
More broadly, I don't fully understand how to interpret the 0/25/50/100 attribute weights. If something is set to a 50% weight, does that mean that products with that attribute will be recommended 50% of the time? This doesn't seem to be the case since you can set the rules to sum to more than 100% (e.g. 3 product types with 75% weight each). Does 50% mean 2x as likely to recommend as another type?
Any help is much appreciated!
The attribute weightings only allow the 25% increment options. In stead of boosting up that single attribute to 75% you could keep it to 50% and then boost another attribute (that addresses a similar set of products/entities) up (maybe 25%) to nudge the results a little more. That way you might be able to achieve the mix you are looking for.
To your later question, boosting 25% doesn't mean it will achieve 25% of recommendations. It's just an increase or decrease of weighting on that set of products. A "nudge" or "boost" are pretty accurately descriptive terms for it's impact. Here's a link to the attribute weighting help docs which describe it and may be helpful.