Eric Bowman, VP of Engineering at Zalando, writes at Minutehack about the importance of machine learning for the future fashion industry:
For example, to correctly recognise a blouse, thousands or even millions of images are used to train computers to identify items that are visually similar. This knowledge is then linked together to reveal symmetries and fashion trends that enable computers to make recommendations based on buying or browsing behaviour.
Fashion data, such as article representations and trend popularity, is critical to advancing ML capabilities, and companies like Zalando and Google are aggregating a lot of this data. […]
In a nutshell, this is how we see the next few years of software in fashion progressing: a cooperation between man and machine where each complements the strengths of the other.
Machine learning (ML) certainly makes a lot of sense in the fashion industry, as it does in many others. (ML will change almost all industries)
And given the aggregated data, Zalando (and, say, Amazon) is in a good position to take advantage of this. (More so than Google, which is probably just named there because it is another company with huge sets of data but is not a competing online retailer.)
But color me sceptical on the immediate impact of ML. Zalando itself is for its mobile marketplace app Fleek an approach that, indirectly, shows the limits of ML in fashion: “Influencers”, idols, stars, and brands, define an important subset of fashion. This social context can not be replaced by automatic curation and recommendation (aka ML products).
It can be enhanced though. A combination of all this will certainly be very powerful.
Still, as Bowman writes himself, accurately extracting the proper intent and context of buying behavior online remains tricky:
Was the item bought as a gift? Were multiple sizes of the same item purchased? Without insight into the reasons behind a decision, purchasing behaviour can create a false narrative. […]
Machines can analyse this information, weaving it together to create a tapestry of consumer tastes and preferences. Over time, machines will begin to understand what a consumer is searching for or what they want to achieve. And then it will be able to offer recommendations that are on par with a highly-skilled stylist.
The main point: For ML, one needs huge sets of relevant data. Zalando has those for fashion. And they are going to use them.
This is important for the fashion industry: Machine learning’s inherent scale effects (the more, the merrier) mean that the bigger companies will get better and thus biggger and so on.
(But again: The social, emotional aspect of fashion will temper this to some extent.)