Technology has the potential to throw the old adage “it’s the thought that counts” into disarray, if an algorithm does the thinking for you. Enter Spanish startup Giftri, another attempt to tap Facebook’s data hive to provide personalized gift recommendations for friends.
Specifically, it analyses your ‘Social Graph’ (your friend-list) and the corresponding ‘Interest Graph’ (the products, brands, and other things your friends have ‘Liked’) and then plugs this data via its own algorithms and machine learning into Amazon’s huge product database to help you pick out the perfect gift from hundreds of thousands of options. That’s of course also how the company generates revenue. Giftri takes a small affiliate commission from Amazon for sending the buyer their way, and in the future it plans to add more online retailers, particularly smaller merchants, and offer ‘featured products’ as an additional revenue stream.
To begin the process of finding a gift, you log in to Giftri with your Facebook credentials and select a friend from your friend-list. The service’s recommendation engine then goes to work and a selection of gift ideas are displayed. The suggested gifts can be rated up or down — helping Giftri to get smarter — and you can filter by price range. You can also manually add products to a friend’s gift list.
“We look at the general profile, such as ‘Single woman in her 30s living in California’, and parse it though a huge data map for Amazon that we call ‘The Matrix’,” explains Giftri founder Søren Beck Jensen. “It uses weighing algorithms to find Amazon categories, such as ‘Ladies purses’ or ‘Art supplies’, for that profile. From the category we then look for ‘Most gifted’ and ‘Most wished for’ items.”
In addition, Jensen says Giftri analyses any Facebook ‘Likes’ the person may have, such as ‘Windsurfing’ or ‘Guns ‘n Roses’, and, depending on the ‘Like’ category, parses it though a filtered search in specific Amazon categories to return gift recommendations specifically for that ‘Like’.
“We also allow users to rate our suggestions and use the ratings to generate gift recommendations based on ‘collaborative filtering’,” he adds. “This is currently only a small part of our algorithm as we do not have that many users yet, but will eventually be the main factor for generating gift suggestions, as we will be able to accurately calculate that there is a big chance you will like a ‘Battlestar Galactica t-shirt’ if you also like ‘Star Wars’ and ‘Star Trek’.”
When I wrote about a similar service — the UK’s ThePresent.co — I noted my uneasiness at asking a third-party app to mine my friends’, albeit public, Facebook data without their explicit permission. More broadly, these types of offerings open up a whole can of worms around who not only owns your data, but how that data can be used by the friends you’ve connected with on various services, even if that data is publicly scrape-able anyway.
On the issue of privacy, Jensen had this to say: “In regards to privacy, we do store your profile data and your ‘Likes’ and we do require that you connect with Facebook to use our service. We do however make it easy to disconnect and delete your data and we take privacy and security very serious. We do not allow you to see the gift suggestions for someone unless they are your Facebook friend, although we do plan to make it possible to make your [own] gift-list public in the future.”