Why Most Ecommerce Sites Make Bad Product Recommendations

Product recommendations have become a staple on ecommerce sites and can be found on all of the major ecommerce platforms and over 90% of all online shops. However, less than 20% of shoppers say their online product recommendation experience was relevant to them. (Bazaarvoice, Ecommerce Consumer Survey 2018)

Product recommendations work but the majority of sites are doing it wrong. This is especially true for the fashion industry, where sites often find it difficult to satisfy each shopper’s unique taste and style preferences.

We will take a quick look at:

As the number of items on an ecommerce site grows it becomes increasingly more difficult for shoppers to find what they are looking for.

The recommendation engine, when done right contributes to a positive shopping experience. A good recommendation experience is one where shoppers find the products they were originally going to buy in less time than browsing the directory or searching for the product. The easier it is for a shopper to find what they are looking for the more likely they are to buy.

A 2019 study from McKinsey observed that 35% of purchases made on the ecommerce giant Amazon’s platform came from product recommendations. This means if your ecommerce site does not have a recommendation engine you are guaranteed to underperform by around 35%.

That same year Amazon got into a lot of trouble when their engine was caught recommending knives to students looking to buy backpacks in the UK.

While Amazon was big enough to deal with the blowback and was able to quickly fix the problem, a negative recommendation on a smaller ecommerce site can be devastating. The Bazaarvoice survey also found that 38% of shoppers won’t return to the online shop that recommends things that don’t make sense for them.

Shoppers hate getting annoying recommendations for products they are not interested in.

80% of shoppers have bad product recommendation experiences because most sites get recommendations wrong. The recommendations on most sites are not personalized enough, not based on good product information, and not being optimized for the shopper’s experience.

1. Not personalized

Most ecommerce recommendation engines recommend products based on the site visitor’s demographics, such as age, geography, and income. While this may work some of the time, the recommendations made are not accurate enough to truly deliver the best product to each potential shopper. For certain types of ecommerce sites such as those selling fashion products such as apparel, this type of recommendation rarely works.

Prince Charles and Mick Jagger are both wealthy British men in their 70s, therefore most recommendation engines would group the two men together but if you are a site selling menswear you would not want to show the same recommendations to these two demographically similar but otherwise very different men.

2. Bad product descriptions

Another big issue with ecommerce recommendations is having bad, inaccurate, or incomplete product descriptions. By not having the most relevant information about a product, you will be able to make good recommendations as recommendation engines rely deeply on having relevant descriptions and product information.

3. Not optimized for the shopper’s experience.

Recommendation engines like all software tools need to be optimized for a metric. And the most frequently used metric for ecommerce product recommendation engines is increasing order value or AOV.

At first glance, this seems like a great metric to optimize, because isn’t the goal of having a recommendation engine to increase profits? Yes, but being overly-optimized for short-term profits such as recommending impulse buys that the shopper will regret after purchase will create a negative shopping experience. Loyal returning customers with a high customer lifetime value (CLV) are much more valuable than a slightly higher one-time transaction.

Based on the problems that the majority of ecommerce sites’ recommendations face, a better way to recommend would be to have recommendations that are truly 1-to-1, based on accurate product descriptions, and optimized for the shopping experience (customer lifetime value, CLV).

There are two approaches to fixing this problem: one is to deploy a team to build the recommendation engine for your ecommerce site from the ground up with this in mind.

For product recommendations, especially fashion recommendation engines an individual’s taste profile is especially important. Recommendation engines based on machine learning algorithms that can learn what a shopper wants within a couple of clicks are a lot more effective than recommendations based on demographic information. This type of recommendation also respects a site visitor’s privacy but not needing additional information from Google or Facebook to work effectively.

Korean fashion retailer Codibook has over 120,000 items in their catalog and they understood early on that recommendations would be a key part of their shopping experience.

To help make the recommendations as accurate as possible, every visitor to Codibook’s site has a unique preference profile created and is only shown items that match their preference profile. Their profile is updated in real-time based on on-site behaviors and their engine can even make accurate individualized recommendations for first-time visitors.

Codibook’s recommendation engine also takes the visual attributes such as style, fabric, and fit directly from the product image which enhances the accuracy of the recommendations by adding another layer of depth to the products.

All of this created an improved experience for shoppers and contributed to higher revenue and customer lifetime value (CLV) for Codibook.

Another example of a successfully implemented recommendation engine is from fashion retailer Ahua, where they increased their AOV by 42% in just 14 days.

Related posts:

Getting up at 5am. Taking cold showers. Meditating. Keeping a diary. Reading tons of books — at least 52 a year, so one per week. Planning your days minute after minute. Do these sound familiar…

bibit pisang merah Bibit yang kita kirim Tinggi bibit 30–50 cm Tergantung stok yang ada Bibit hasil semi/cacah bonggol Bibit yg kita kirim kami sertakan bonggol sedikit agar bibit lebih kuat Pisang…

Deluxe houseboats in Alleppey are finances like minded houseboats encompass widespread centers like dwelling area, A/C bedrooms, connected lavatories and kitchen. A/C operating from 9:00pm to 6:00am…