Research the topic of e-commerce returns and you’ll find that various sources quoting return rates of 20% to 30%, with some items like luxury goods and swimwear even higher. That’s as much as three times the rate for purchases from brick-and-mortar stores, which average about 9%.
With global e-commerce orders exceeding $16 trillion in 2022 (according to IMARC Group), there’s a significant cost in the processing of returns, but with artificial intelligence (AI) now in the mix, there’s a lot of opportunity to get it right.
Discover how AI has created a new ecosystem of partnerships with a fresh spirit of customer-centric cocreation and a renewed focus on reimagining what is possible. The Acceleration Economy AI Ecosystem Course is available on demand.
The machine learning (ML) component of AI detects patterns. ML depends on data. So, in industries where there is a lot of data and a high level of transactions, AI is able to make recommendations to the buyer. Accurate recommendations increase the likelihood that customers are happy with the items they’ve ordered and are less likely to make returns.
The familiar music streaming services Pandora and Spotify utilize AI to understand users’ music preferences. They employ ML to create personalized playlists, discover new music, and recommend songs based on a user’s listening history. This personalized approach helps reduce dissatisfaction related to music choices.
Stitch Fix, an online personal styling service, sends curated boxes of clothing to subscribers based on an AI algorithm that considers their users’ preferences in fit and taste, as well as their ordering patterns. Since it was implemented, AI has helped the company reduce returns by 25% and improve recommendation accuracy by 30%.
If this technology were installed in every site that expects returns, the incidence of returns would plummet.
In the case above, returns were reduced by improving the probability that people will be happy with
their purchase. In another example of reducing returns by stopping them from happening in the first place, ML analyzes returns and identifies commonalities in defects that it can then direct to certain follow-up actions. For example:
- Products arriving broken — the action is a product or package redesign
- Late delivery — the action is carrier changes or adjustments to committed delivery dates
- “Not what I expected” — the action is to place more accurate descriptions on the website
- “Buyer beware” — the action is to flag items that have a high rate of return with a “frequently returned” badge, which Amazon has begun to do, giving potential buyers more information before purchase
I’ve written several times over the last year about AWS Supply Chain software and how it uses ML to direct items to the right warehouse so rebalancing can occur quickly. Software like this can factor in distance, sustainability, and demand to get non-defective items returned back into the system quickly.
According to the research firm CBRE, the average cost of an e-commerce return ranges from $20.75 to $45.25. That means you need margins at least that high to break even on a return. And that dollar amount is greater than the typical e-commerce margin.
So much so that the math suggests the most prudent thing to do in some situations is to refund the purchase price and not try to manage items back into the system. But that becomes a delicate balance between sustainability goals and the cost of disposal.
With AI-driven warehouse management, there are opportunities to make dynamic decisions on return disposition. But the biggest bang for the buck comes with analytics and action that stops companies from getting returns in the first place.