Why chainsaws aren't books: product recommendations in eCommerce

@lizrice#DS15

Hello! This is a talk I gave

at Digital Shoreditch

2015. Let me share just a

few things I’ve learnt

about recommendations

and personalization

@betterrecs

Chainsaws are not books

flickr: pasukaru76

… but apparently there are

folks in ecommerce &

advertising who don’t know

that!

This is my friend Piers. He

bought a chainsaw online…

…and then he couldn’t

escape from ads about

chainsaws.

They followed him

everywhere!

flickr: kalexanderson

The thing is, he only wanted

one chainsaw.

He wasn’t building a

chainsaw collection! He just

wanted to cut up some wood

with the chainsaw he had

bought.

Books are different. You are

quite likely to buy more than

one of them.

You might well buy several

similar books.

Dorling Kindersley

Different recommendation

types:

<- Complementary items

<- Alternative choices

flickr: michaelpardo

Recommendation engines

often use an algorithm that

finds people who interact with

items in a similar way.

Doesn’t really matter what

the products are - it’s the

user behavior that counts.

flickr: epriyanto

Other algorithms use product

features to make

recommendations.

Or you might combine

behavior and features,

maybe other factors too like

quality or time of day.

eurobricks: scorpio

If you’re selling a chainsaw, it

makes sense to help the user

find alternatives. Once they

put it in the cart or buy the

item, they want to see

complementary items not

alternatives.

When Piers bought that

chainsaw, he got tagged with a

browser cookie as someone

interested in chainsaws. That

cookie didn’t get cleared when

he made his purchase, and

doesn’t say he’s someone who

now owns a chainsaw

stickeryou.com

lego.com

Product recommendations can lift

store revenue by as much as 30%according to Forrester

Just make sure you’re using the

right kinds of recommendation in

the right places on your store

betterrecommendations.com @betterrecs