ViSenze Blog

Machine learning for e-commerce: Awesome or hype?

Posted by ViSenze 10-February-2016

It’s hard to think that machine learning (the current buzzword in science and tech) can be applied to the world of e-commerce. For those unfamiliar with this term: Machine learning explores the study and construction of algorithms that can learn from and make predictions based on data. These algorithms are able to model complex relationships between inputs and outputs, and to find patterns in data. Yet, this “techy” subfield of computer science does have profound impacts for the industry.

What can machine learning do for e-commerce?

To start with, e-commerce success involves taking in a massive amount of data and trying to make the best decisions based on those data sets. The ultimate challenge is in making sense of all of that data, and machine learning algorithms can help companies to achieve that, providing actionable insights to improve the shopping experience for consumers. Let’s have a look at some of the major applications of machine learning in the e-commerce field:

1. Product recommendations

Currently, most e-commerce recommendation systems are built upon the principle of popularity - based on user click-through rates or product sell-through rate - which has nothing to do with relevancy for consumers. Machine learning makes this challenge easier by using predictive analysis to understand product attributes, user purchase patterns, the performance of different products on the site, to determine relevant recommendations that have a higher probability of generating a sale.

enhance product recommendations

In addition, through large-scale data analysis of query logs (both keyword and image queries) and user behavioral data, we can create patterns and intelligence between queries and products, and even get insights into understanding query intent to improve search results and product rankings. E-commerce sites can then offer these highly-relevant recommendations in real-time, helping consumers discover items in your catalog and improving conversion in this era of personalization.


2. Fraud detection and prevention

fraud detection and preventionThis is a pervasive problem faced by all e-commerce companies. In 2014, the annual fraud costs reached $32 billion, a 38 percent increase from the previous year. Fraud detection involves the constant monitoring of online activities, and automatic triggering of internal alarms. Data mining uses statistical analysis and machine learning for the technique of “anomaly detection” - detecting abnormal patterns in a data sequence, such as buyer profiles and activities. With machine learning, we can achieve accurate and fast fraud detection in real time.


3. Trend forecasting and analytics

E-commerce players, especially fashion retailers, often experience frustration due to the lack of information available to help them understand and respond to the latest trends in this fluid and fast-moving industry. Most of them have internal data on the performance of previous seasons’ products and access to inspirational trend sites, but no way to understand opportunities they have missed or concrete data on how they can improve their product assortment.

Big data and machine learning allows for the aggregation of trends and sales information from a wide variety of sources around the globe - from retail sites, social media, designer runway reports, and blogs - and then makes it understandable and accessible in real time. Analytics and insights can be created in chosen market categories for all parts of the supply chain - merchandising, buying, trading, and strategy - to track competition, anticipate demand, and refine product planning for upcoming seasons.


4. Product tagging automation

By deploying image recognition and computer vision, automatic product tagging can be conducted at scale. New items added to your inventory can be automatically tagged with the relevant visual attributes and product categories with machine learning. Product titles and descriptions can even be created automatically, For example, this is a  “sleeveless, midi floral dress with sweetheart neckline”.

This saves so much manual effort on your backend by streamlining and accelerating new product uploads. It also unifies and standardizes your product categories to make search, browsing and filtering much easier for end users. 

Machine learning for e-commerce: Definitely worth the hype

We are just beginning to witness the growing need of machine learning and its huge potential for the e-commerce industry. The use cases listed above (far from being exhaustive) are just the beginning of this technological wave that would float all boats - from suppliers to retailers to end consumers, from the frontend to the backend.

With this technology and the changes it brings to the industry, it unlocks the ability for customers to have exactly what they want and not necessarily what’s been decided for them. It also allows for the market to be more efficient in satisfying different tastes, and address concerns on fraud, product visibility, and relevance. With the deployment of machine learning and its applications, e-commerce retailers can finally deliver the right products at the right place and the right time, providing intelligence-powered shopping experiences.

 

New Call-to-action

 Image credits: 1 | 2

  E-commerce,   Deep learning

Comment