Ramgopal Prajapat:

Learnings and Views

AI/ML for Reducing Returns in Ecommerce

By: Ram on Aug 03, 2022

Summary

In this blog, I will share a point of view to address higher product return challenges using AI/ML.

  • Product Returns in Ecommerce
  • Reasons of Product Returns
  • Ideas to reduce returns (cancellations post return)

 

Higher level of order cancellations and returns has huge impact on profitability for the ecommerce businesses.  In the earlier blog, we discussed about the order cancellations and returns in detail.

Diagram

Description automatically generated

Orders are cancelled across stages 1. Cancellation Before Delivery (Cancel an Order), 2. Could not be delivery so cancelled by delivery partner (RTO – Return to Origin) and customer initiated return post-delivery. Typically, 20-30% of the orders are cancelled across any of these stages and of these 60-70% are returned post-delivery (Return Product or Customer Initiated Return).  

 Chart, bar chart

Description automatically generated

In this blog, we will discuss about Product Returns and what can be done to reduce return percentages.

Some of the returns are part of business model and must be considered as an expected cost for the business model. But a higher level of return can be a killer for the sustainable business model and action must be taken to address these returns.

We can look at the return reasons and design AI/ML based strategies to address these.

Chart, bar chart

Description automatically generated

We can link these returns to further factors and identify ways to address requirement of genuine customers and not all the customers are genuinely interested in the delivery of the products.  These returns can be linked to Product, Quality/Review Process and Delivery.

  • Product
    • Provide accurate and improved product descriptions and the details so that customers get what they intend to
    • Enable size fit checking on the website. We know different brand have different sizes for the apparel category and allowing them to do the comparison will help
    • Virtual Try On – a lot of ecommerce website enable virtual try on for some of the categories to return returns
  • Quality
    • Rigorous quality review process before handing over the parcel to delivery partner can help
    • Deep Learning/Computer Vision based product quality review process can enable systematic quality check in a faster way
    • Additionally, the computer vision-based product matching between ordered product image and selected for packaging can address the gap of ‘Incorrect product received’
  • Product Delivery
    • For the potential suspect products and customers, the customers and probably delivery partner can be advised to capture product images at the time of delivery so ensure whether product was defective or not at the time of delivery.

Summary Action Strategies

Some of the reasons can be tackled by AI/ML actions directly. And for the other reasons, we first need to set up the processes and then improve the intelligence of the processes using AI/ML.

Table

Description automatically generated

Above recommendations can help in reducing process gaps for an ecommerce business to manage the returns.

Additionally, there can be various proactive measures to reduce returns and operational costs especially for the repeat offenders and fraudulent customers.

Managing Return Frauds for Ecommerce

Around 30% of customers order online and return later as they know about getting full refund. There is huge cost in terms of inventory loss, logistics costs and other expenses to the company due to actions of these fraud customers.   

In the next blog, we will focus on identifying the orders linked to fraudulent behaviour using ML techniques and actions for the team to take.

Concluding Thoughts

AI/ML can help in reducing inefficiency in Ecommerce. And one of the key use-case is reducing customer initiated returns by proactively deploying AI/ML frameworks.  Hope you found some of these ideas useful and relevant. 

What can be additional reasons of product return?

What additional actions can be taken up to address these?

Looking forward to hear from you.

Leave a comment