Today I met Emma. Emma is an e-commerce chat bot. Yap, it’s a software that is designed to interact with people and give recommendations for what they should wear.
Created by ChatShopper, She is integrated in the Facebook Messenger and interacts with people based on automated conversational patterns.
Emma is very kind, polite and even a bit funny. However we couldn’t hit it off at first glance since Emma doesn’t speak English, not a good start.
Ok, I will make the first move, I will try to communicate with Emma using my mediocre German. With all the might of my German iPhone keyboard, I said:
“I am searching for a red dress” (in German).
I spelled it wrong so she showed me a blue one:
Ok, fair enough. Second try, this time I spelled it correctly.
Emma came up with something that looked like a Christmas gift wrapper dress:
“That’s not quite my style”.
Emma came up with a worse dress, it looked liked cotton candy.
I kept giving Emma simple and short directions; shorter, for summer, modern, classical style, more trendy.
This is how it looked like:
No, no, no!
At some point she gave up and said she doesn’t have more recommendations for me, and I got mad for her incompetence.
I had to face the simple fact that Emma has absolutely no Style!
Talking to the creators of Emma at a Zalando Tech Open Air Satellite Event, they explained to us that she still has a lot to learn.
Just like a child, she needs to learn more languages, with all of their accompanying nuances.
But more importantly, Emma has to learn style and trends. This would be the hardest challenge for E-Commerce bots, to be able to recommend and display items that are compatible with current fashion. She also needs to know what is the difference between a classical red dress and a modern red dress. She needs to cross reference her own inventory and other fashion content providers and influencers to understand what is trendy right now.
This needs to happen on a daily basis since fashion is an ever changing current.
Emma also needs to learn how to see.
She will have photo recognition capabilities which can identify what items are displayed in a certain photo and try to match those with the ones she has in her own inventory.
For example, I found a really nice coat on an Instagram picture, I would be able to upload that onto the chat and ask Emma to find me one that is either the same or very similar to it.
That technology is already available on the Zalando app, with questionable performance:
On the Zalando app, you enter the search tab. In the search tab you choose the little camera icon at the top right:
Then you will be requested to take a photo of the desired item or to upload one from your library. Next step is to focus on the desired item with a grid.
Here is an example:
Once you have chosen an item, zalando searches it’s inventory for most resembling items:
One alternative in the market of curated shopping, Zalon app, enables users to chat with human stylists.
Based on their experience and the photos you upload to the chat, they will create a curated shopping box matching to your taste, needs and measurements, with no service costs! (I will talk about Zalon more in detail in my next post).
You can upload a picture of a bag you like on the chat and a stylist will do the web searching for you. Downside is, it’s a human at the other end, who are known to be disappointing sometimes, and my stylist never answered my inquiry. Meh.
Correspondence in Zalon chat:
Over a human who doesn’t answer, and a bot without style, I prefer Emma the bot.
Taking me back to Emma. While ChatShopper claims to have a conversion rate of 9% from a recommendation to purchase, I have yet to see these outstanding recommendation capabilities.
Once Emma learns how to speak English (and other languages!), is caught up with current trends and becomes a real Fashionista, and starts to recognise pictures, then Emma and I can become best BFFs.