The Farfetch Marketplace Flywheel & Product Development Supercharged
By Pedro Cerqueira, Product Owner
As explained in this article, the Farfetch marketplace flywheel is powered through a virtuous cycle that starts with a great experience that drives traffic to the platform and third-party sellers, consequently improving the selection of goods and cost structure contributing to the flywheel spin.
The Farfetch Marketplace Flywheel on Steroids is, in essence, the same virtuous cycle, but leveraging on top of another extremely powerful virtuous cycle driven by Big Data, Machine Learning and Data Science.
As the flywheel spins, there’s significant growth in the user and customer base, as well as more visits and purchases (demand) led by improvements in experience, increased user acquisition and an ever-increasing item selection and range (supply).
This type of growth results in a significant increase in the data available (Big Data) about our users/customers’ behaviour and buying preferences at an aggregated anonymous level (the same applies for sellers and partners).
By having machine learning algorithms (called learners) absorbed in learning models on how to learn from data on our customers and their tastes, Farfetch creates a hard to copy core competitive advantage. This is our ability to optimize our customers’ experience to discover the items they want from the ever-increasing long tail of fashion items we showcase and sell.
These learners will guide the discovery and inspire each individual customer (without knowing who they are, just because of the way they interact with the site/app) to find and discover the items they are looking for or didn’t even know they wanted.
This can all be accomplished through the right recommendations and ranked items shown at the right time to the right customers. Effectively generating and moving demand, selling more, generating more data, and starting yet a new cycle revolution for learners to act upon.
Darwin’s Theory of Evolution by Natural Selection: adapted for Product Development
Besides being able to address the long tail challenge presented earlier, learners can improve Farfetch technology products by applying the scientific method of experimentation. This means learners will rapidly and constantly change our products by learning what experience works best for each customer, leveraging build-measure-learn cycles unconstrained by human capacity.
Each of our customers operate within their own complex environments — e.g., political, economical, social, technological. Their interactions with our products reflect conscious and unconscious behaviours and micro/macro decisions.
Product evolution occurs when this sort of “natural selection” happens on top of these interactions and decisions favouring certain characteristics to the detriment of others, becoming more prevalent or rare within a product. It is this process of Product evolution that gives rise to mass Product personalization and market fit. Over successive iterations, our machine learning algorithms learn from its customer interactions and adapt our product characteristics accordingly. This results in a kind of “survival of the fittest” product, whatever that might be or look like in the end. Our learners are therefore able to achieve product/market fit for each customer without being able to clearly explain all the reasons why! It just works.
As Farfetch products adopt more learning-enabled features, this creates a new massive competitive advantage: each customer will feel like a luxury private client with a designated concierge. Each learner will become a Farfetch virtual shopkeeper working as a private client assistant alongside its human counterparts.
These learners will expand Farfetch’s talent community, and whoever has the best data, data scientists and learners in the industry will have a competitive edge in the arms race to be the global luxury fashion leader.
Conclusion: Learners as Farfetch’s First-Mover Advantage
New users and more engaged repeat customers will generate more data available for learners to learn, which will help learning models to create products our users love. This will draw in even more new users and customers, which will generate even more data to learn, thus triggering yet another virtuous cycle.
The first player to harness this power will have the first-mover advantage, having gone through more learning cycles than the competition. The higher the number of flywheel revolutions, the better the learning ability and speed. This transforms the first-mover advantage into an almost unsurpassable barrier to new entrants who won’t have the necessary data volume, connections nor knowledge to remain competitive.
References:
The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World, book by Pedro Domingos
Originally published at https://www.farfetchtechblog.com on April 1, 2020.
