next »

[arms]

best thing i could come up with is natural language search in mysql (MATCH ... AGAINST) for the product name, take n best results. could further filter that by same category or a complimentary one (ie. current product category is "tops", do search in "bottoms"). removing common words from the search could also improve relevency.

this is actually working pretty well with some stuff i tried, but i could see it failing on other things.

also i am just working with affiliate stuff so i don't have buying habit/pattern data from previous or logged in customers. would be very curious though how anyone here is doing this.

anyone care to share what solution they use for this or just throw around ideas?

[KaptainKrayola]

fancier than what i have going on for my aff sites

i have all of my feeds bundled up into categories and for related products i'll just grab some products from the same category.  i also have groups of categories bundled together so that helps spread the results out some on categories that are light on product.

seems to work out ok as long as i keep a good eye on the category relations

[arms]

i was doing something similar, just grab next n products by id in category.

doesn't hurt for indexing and adding some extra related content on the page but in broader categories it can be off.
for example someone looking at "little pink dildo" probably doesn't want the "giant black destroyer dildo".

for blacker sites it doesn't matter, but for whiter sites i'd be curious if it increases conversions or adds legitimacy to the site.

[nutballs]

my "may also like" algo was really simple.

same 5 products that made me the most profit. LOL

[perkiset]

Bang. We had a change of heart a couple years ago when we decided to stop selling what we thought the consumer wanted, and started selling what we wanted to sell. So through a simple keyword/color match, but then prioritized by price and inventory levels, that's how I show what people might also like. And you know what? Works like a charm. Whatever color we want to sell, that's what I put up and it sells - by a wide margin - more than the other available colors on a detail page. Whatever we put up as "You might also..." sells better than anything else to a returning or already-carted customer.

[arms]

Bang. We had a change of heart a couple years ago when we decided to stop selling what we thought the consumer wanted, and started selling what we wanted to sell. So through a simple keyword/color match, but then prioritized by price and inventory levels, that's how I show what people might also like. And you know what? Works like a charm. Whatever color we want to sell, that's what I put up and it sells - by a wide margin - more than the other available colors on a detail page. Whatever we put up as "You might also..." sells better than anything else to a returning or already-carted customer.

thats pretty interesting, but not all that supprising.

i guess it's similar to an in store salesperson telling a woman how such and such a hand bag would go great with that dress she's trying on.

[KaptainKrayola]

very tasty ideas - if i ever get around to working on that store project again i'll have to implement some of these


you guys are full of win

[NYDAz]

[DangerMouse]

A simple version would randomly display products from the same cat / tag á la the Kaptains method, but record which items generate the most clicks, or basket adds. Once statistical significance has been achieved you can standardise on displaying those products.

Throw in a bit of logic á la Nuts to establish that although one product may generate the most purchases the product with the highest profit per click should be prioritised and the system improves further.

How about taking it to another level again and segmenting by traffic source and / or inbound keyword and displaying related items based on inferred intention?

I guess other considerations include whether you end up cannibilising your own sales where visitors go off at a tangent viewing related items rather than converting, or whether they end up purchasing lower value products.

Just some rambling thoughts.

DM

[nutballs]

ok I slightly over simplified. I basically have always done a sort by profitmargin, but within the same category only. So no one got car tires if they just bought a backpack.