Better Ecommerce Personalization: Conquer the Cold Start Problem
Today, ecommerce personalization makes it easy for companies to cater to each customer’s specific interests…when they know the customer.
However, when first-time visitors arrive on site, it’s really more of a guessing game, one that even smart algorithms can’t always get right.
After all, it’s not so easy to offer personalized product recommendations when you know nothing about the visitor.
This is known as the “cold start problem”, and it is one that all ecommerce companies face.
Without historical data on hand about browsing history or previous behaviors, the default strategy is to offer general recommendations such as the best-selling items.
In the fast-paced digital age where machine learning enhances the customer experience, the issue of the cold start problem persists.
If not addressed, it can cause significant damage to your business.
The Pain of Cold Starts in eCommerce for First-Time Visitors
A site visitor will be classed as a cold start if the site doesn’t have information on any of these aspects:
- Browsing history
- Transaction history
- Interaction with a store item
- Ratings or some form of feedback on any items
- Known name, email, or IP address
Studies from Retention Science indicate that ecommerce companies class almost 50% of their users as cold starts.
First-time visitors are rarely willing to hand out their email address or even considering buying anything yet, which makes them notoriously difficult to convert right away.
Typically, just over 5% of first-time visitors take any desired action on an ecommerce site, such as filling out an email form.
Many ecommerce sites will employ a simple email capture form, optimized to convert by using the minimum number of fields.
This then contributes to the cold start problem, as you may only end up with a first name and email address and know nothing else about their interests for their next visit.
As a result, bounce rates can skyrocket. Industry standard is 50%, though it can vary dramatically depending on the industry and services on offer.
Personalized product recommendations can drive sales, however if they aren’t of interest to visitors, they can be one of the major factors impacting bounce rate.
This issue can make it harder to secure sales, even if it comes from paid advertising. Because of this, ecommerce companies will be faced with higher acquisition costs and a lower return on investment (ROI).
Remarkably, it costs five times as much to attract new customers, compared to the cost of keeping an existing customer. If an ecommerce business is keen to attract new customers, there is a big risk of a low ROI, unless they learn how to deal with the cold start problem.
How Machine Learning Can Eliminate the Cold Start Problem
Machine learning is advancing at an astonishing rate. Forbes predict global revenues for AI will soar from $1.62B in 2018, to over $31B by 2025.
Today, smart algorithms can gather enough information to generate accurate personalized product recommendations for almost every visitor – even first-time visitors.
Collecting data on visitors, such as their gender, devices, location, and referral site, will help the software create accurate micro-segments for first-time visitors.
Considering that 97% of visitors to ecommerce sites leave when the first product they see is isn’t of interest, improving ecommerce personalization is of critical importance.
Boosting conversion rates of new visitors will lower acquisition costs and increase ROI.
Here’s how to do it.
- Content-Based Filtering
The cold start problem can relate to new visitors and also to new products, which have yet to receive sufficient feedback or reviews from users. This can make it difficult to know when and where it should be displayed.
Through content-based filtering, the system initially relies on the metadata of new products to create recommendations, putting visitor actions second.
Furthermore, it can separate users by their click behaviors. Normally, this starts to track on the first visit, after the first few clicks.
On a new site, the first few clicks tend to be sporadic, as first-time visitors are finding their way around the site and store. By starting to track after a few clicks, the system can discover what the user’s true interest is.
- Use Dummy Data
The issue with new visitors is that the machine has not encountered them before, and so it is trying to figure out preferences and offer recommendations with no data to work from.
By feeding the system with scenarios of dummy data, businesses can effectively train their ecommerce AI ahead of time, preparing it for live customers. Using a virtual assistant app is a good way to do this, enabling you to kickstart the growth of your machine learning.
- Form Strategic Partnerships
Some businesses have access to huge amounts of valuable data, but they lack the means of generating insights. Pharmaceutical companies are an example of this, but it is the drug development companies who possess the resources and skills to actually draw valuable insights from the data.
By partnering with companies who hold proprietary data, ecommerce businesses can gain a lot from this data access by applying their AI software and analytics skills.
- Scrape Publicly Available Data
Through public forums and public databases, companies can gather relevant data to develop smart algorithms upon. One caveat of this method is the possibility of breaching the EU General Data Protection Regulations (GDPR).
Smart Algorithms make Smarter Offers
First-time visitors don’t want to have their time wasted. While they may be less likely to convert straightaway, companies can use data insights to boost the chances of conversion.
Through accessing as much data as possible on products and visitor behaviors, machine learning can segment visitors, and develop a smart algorithm capable of delivering more accurate personalized product recommendations.
Over time, this allows companies to map their customer journeys from the beginning, and ultimately find a solution to the cold start problem.
The end result will be a level of ecommerce personalization that enhances the user experience, fostering customer loyalty to take your company above the competition.