Presented by AWS Machine Learning
Amazon’s 2019 Climate Pledge calls for a commitment to net zero carbon across their businesses by 2040. Since then, the company has reduced the weight of their outbound packaging by 33%, eliminating 915,000 tons of packaging material worldwide, or the equivalent of over 1.5 billion shipping boxes. With less packaging used throughout the supply chain, volume per shipment is reduced and transportation becomes more efficient. The cumulative impact across Amazon’s enormous network is a dramatic reduction in carbon emissions.
To make this happen, the customer packaging experience team partnered with AWS to build a machine learning solution powered by Amazon SageMaker. The primary goal was to make more sustainable packaging decisions, while keeping the customer experience bar high.
“When we make packaging decisions, we think about the end-to-end supply chain, working backward from the customer in terms of the waste they get on their doorstep, but we are also really cognizant of how our decisions in packaging impacts speed to fulfillment,” says Justine Mahler, Senior Manager, Packaging at Amazon.
Whether it’s sending off a water bottle or a grill, her team’s objective is to use ML to deliver packaging that delights customers, arrives undamaged, and contributes to a reduction in Amazon’s carbon footprint.
“We try to minimize the amount of packaging customers have to dispose of, and drive toward recyclability in our packaging as well,” Mahler says. “Carbon is the primary metric that we hold ourselves accountable to when we think about sustainability for the customer – and our corporate responsibility to be a leader in that space.”
The sustainable packaging challenge
Amazon sells hundreds of millions of different products, and sends billions of shipments a year. To ship all with minimal packaging, maximum speed, and customer satisfaction, the team must innovate on a large scale.
“This is a challenge that machine learning is uniquely able to solve,” says Matthew Bales, a manager of research science at Amazon. “Instead of having someone inspect these products individually for things like fragility or how they would eventually ship, we use machine learning.”
The goal was to scale decision making across the hundreds of millions of products that are shipped – to not automatically default to boxes, but instead identify items that can be packed in a mailer, polybag or even paper bag instead. Both mailers (padded paper envelopes) and polybags (the familiar plastic padded bags) are more sustainable choices. They’re 75% lighter than a similarly sized box, and will conform around a product, taking up 40% less space than a box during shipping – which means a lot fewer trucks on the road.
The machine learning difference
In practice, this meant creating a machine learning algorithm built on terabytes of product data from product descriptions to customer feedback. Working closely with AWS professional services, these terabytes of data are cleaned, catalogued, and ready for mining. The ML algorithm then ingests that data to identify the best packaging with the least waste.
Some of the most impactful ML models identify products that don’t need any packaging at all – like diapers. Others can look at a category like toys and differentiate between collectible items where protecting the original packaging is key vs. the rest of the category where fragility is much less important.
By 2020, ML tools changed the packaging mix significantly, reducing the use of boxes from 69% to 42%.
“It turns out that we know a lot of things about the items in our catalog, but for many items we don’t have detailed fragility information that’s relevant for Amazon’s complex shipment process,” Bales says. “Before we built this model, everything had to be caught by very generalistic rules – but it turns out there are a lot of exceptions to those rules.”
The model allows them to dig into all these exceptions – like collectible action figures that required more packaging than first assumed. It ensures every item is packaged in the correct size mailer or box, or no box at all, and all at scale.
Using Amazon SageMaker, they can analyze hundreds of millions of products, billions of customer shipments, and multiple channels of customer feedback, providing actionable insights in real time.
SageMaker ended up being key for them, Bales says, in part because it offers full customizability. As the models got more complex they were able to move from built-in models to custom models. Amazon SageMaker helped facilitate the launch of new models in just weeks, allowing them to continually invent new ways to eliminate waste. From ML models that predict what products might leak, to identifying products that can be shipped in a paper bag to finding products that can be folded into smaller packages – the possibilities are endless.
Looking to the future of sustainable packaging
As the packaging experience team monitors social media, they’ve seen that customers are noticing the change and offering positive feedback. And today, thanks to Amazon’s efforts, thousands of vendors are working alongside the company to adjust their own packaging to make more sustainable choices.
The team’s customer obsession is driving them to see how far they can reduce wasteful packaging, to evaluate new items quicker, to design better packaging, and to meet their larger carbon pledge.
“We’re focused now even more increasingly on elimination to reach those goals,” says Mahler. “That’s going to require more machine learning, infrastructure investments, and breakthroughs in materials science. This work has certainly given us a head start.”
Dig deeper: See more ways machine learning is being used to tackle today’s biggest social, humanitarian, and environmental challenges.
Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. Content produced by our editorial team is never influenced by advertisers or sponsors in any way. For more information, contact [email protected]
Source: Read Full Article