In what now seems inevitable, an online fashion retailer in India owned by an e-commerce startup that’s backed by Walmart is doing research with Deep Neural Networks to predict which items a buyer will return before they buy the item.
With this knowledge, they’ll be better able to predict their returns costs, but more interestingly, they’ll be able to incentivize shoppers to NOT return as much, using both loss and gain offers related to items in one’s cart.
The nuts and bolts of it is: the AI will assign a score to you based on what it determines your risk of returning a specific item to be.This data could be from your returns history, as well as less obvious data points, such as your search/shopping patterns elsewhere online, your credit score, and predictions about your size and fit based on aggregated data on other people.
Then it will treat you differently based on that assessment. If you’re put in a high risk category, you may pay more for shipping, or you may be offered a discount in order to accept a no-returns policy tailored just for you. It’s like car insurance for those under 25, but on hyper-drive. If you fit a certain demo, you may start paying more for everything.
Preliminary tests have shown promise in reducing return rates.
So many questions:
Is this a good idea from a brand perspective? If this becomes a trend, will retailers with cheap capital that can afford high-returns volume smear this practice as a way to gain market share?
Will this drive more people to better protect their data and “hide” themselves online? We might be OK with being fed targeted ads based on our data, but what happens when your data footprint and demo makes that jacket you wanted cost more?
Will this encourage more people to shop at brick and mortar stores to sidestep retail’s big brother? Or will brick and mortar stores find a way to follow suit?
How much might this information flow back up the supply chain, to product design, even?
Why it’s hot
Returns are expensive for retailers. They’re also bad for the environment, as many returns are just sent to the landfill, not to mention the carbon emissions from sending it back.
So, many retailers are scrambling to find the balance between reducing friction in the buying process by offering easy returns, on the one hand, and reducing the amount of actual returns, on the other.
There’s been talk of Amazon using predictive models to ship you stuff without you ever “buying” it. You return what you don’t want and it eventually learns what you want to the point where you just receive a box of stuff at intervals, and money is extracted from your bank account. This also might reduce fossil fuels.
How precise can these predictive models get? And how might people be able to thwart them? Is there a non-dystopian way to reduce returns?