Skip to content Skip to sidebar Skip to footer

How Does Machine Learning Help Peer to Peer Online Marketplaces Succeed?

Machine learning technology brings a real ROI to peer to peer marketplace owners. Airbnb, a company known for their personalized approach to each customer has seen a significant growth in the number of bookings after they started using machine learning. Other popular P2P marketplaces such as Etsy, Uber, Lending Club also apply intelligent algorithms to increase their conversion rates and acquire a competitive advantage.

 

Machine learning helps leading marketplaces and companies that provide AI consulting services develop features that bring tangible value to business and consumers. So how exactly do P2P marketplace companies use this cutting-edge technology? Let’s find out.

 

Product recommendations

 

Predicting which products a user will like is a widely adopted use case for machine learning. A variety of digital products today have recommender systems powered by this technology. In the past product recommendation systems were based on hard-coded rules. Those rules determined what item to show to a user based on some predefined scenarios. For example, if a user buys a fancy red dress, she’s likely to also buy shoes that match this dress.

 

New marketplaces that don’t yet have a lot of users and items to sell can still use rule-based systems. But once the amount of data increases, a simple rule-based algorithm will no longer keep the customers satisfied with the provided recommendations. Because the more users and items you have the more exceptions to the rules there will be.

 

Etsy has tons of data to deal with. There are  21 million buyers, 1.4 million sellers, and 35 million items on Etsy. While this enormous volume makes Etsy an excellent place to find anything you want, having millions of items listed represents a huge challenge.

 

To make this huge market smaller and more discoverable Etsy’s data science team is engaged in building machine learning algorithms that provide item, user, and shop recommendations to buyers that perfectly match their tastes.

 

They use two approaches to develop their recommender systems: Matrix Factorization and Latent Dirichlet Allocation (LDA). The data that these algorithms work with include user historical preferences, preferences of users with similar tastes, and shops that sell items of similar styles to the ones a user shopped at.

 

The best price predictions

 

Sellers on peer to peer marketplaces often find it hard to price their offerings. Price is determined by supply and demand. For example, a large festival in town can increase the demand for hotel rooms and accomodations, so the costs for bookings will go up. On the other hand, the price for hotel rooms near ski resorts in summer won’t be high because the supply will be plentiful. It is quite difficult for sellers to keep up with the fluctuating forces that go into market pricing.

 

In fact, Airbnb had a problem with hosts dropping off at the stage when they had to define the price for their listings. To solve this problem Airbnb used machine learning.

 

Their new Price Tips feature helps to predict prices based on all of the historical data on travel patterns. The Airbnb’s hosts can take those price suggestions by default. Every day the algorithm recalculates the price predictions for all listings around the world, up to 12 months in the future.

 

Fraud prevention

 

Because peer to peer marketplaces are so popular they attract a lot of fraudsters. The most widespread fraudulent activities include Fake Profile Fraud when a seller sells items that the buyer will never receive, and Fake Buyer and Seller Closed Loop Account Fraud when buyer and seller carry out a transaction for goods that don’t exist using a stolen credit card.

 

To combat different types of fraud marketplace developers use risk-based systems that validate sellers and buyers based on various information including social accounts they use to login, transaction amount, geography, and products.

 

The data science team at Uber, for example, built a machine learning model called Locality Sensitive Hashing (LSH) algorithm. This algorithm detects a large number of similar trips around a city which helps Uber capture fraudulent drivers.

 

They also wrote a special rule engine called Mastermind which allows for detecting fraud in a fraction of a second. To build this engine Uber’s engineers didn’t use any machine learning methods.

 

User verification

 

If you’re going to get into a car of a complete stranger or rent a room from people you’ve never met you need to trust them. In a peer-to-peer marketplace, verifying user identity increases trust.

 

To verify users on the platform marketplace owners use different methods: they do manual verification through their admin systems, make people snap a webcam shot of themselves, holding their ID in their hand next to their head, integrate third-party services for background checks, and more.

 

While all these methods may work well enough, machine learning makes the process of user verification less time-consuming and much more efficient. For example, machine learning and AI can aggregate the data from social networks and draw conclusions. This method is widely used by P2P lending companies like Lending Club to qualify borrowers for loans.

 

Machine learning makes it possible to accurately process, verify, and authenticate identities at scale. And it’s going to change sharing economy for the better!

This Pop-up Is Included in the Theme
Best Choice for Creatives
Purchase Now