Pune: As various state governments set up containment zones to curb the spread of Covid-19, e-commerce marketplace Amazon faced a challenge in managing deliveries to these locations.
Soon, using Machine Learning tools, the US-based online retailer could identify buildings and send emails to impacted customers on how to order once the ban on e-commerce to that area was lifted. Amazon helped provide specific services, including digital-only payment options, to customers in containment zones.
“We created virtual pick-up points by extracting building information from the addresses to identify customers in quarantined buildings,” said Rajeev Rastogi, VP-ML at Amazon India.
The company also used its Artificial Intelligence engine to ensure the safety of its employees, using Bluetooth to ensure that social distancing norms were followed.
Apart from helping with its response to the pandemic, the machine learning team at Amazon has been instrumental in several new initiatives that have been introduced in India and then exported to other geographies.
Rastogi’s team is playing an important role in helping the online retailer strengthen its presence in the country. “We’ve been working on demand forecasting for different products to give guidance and insights to sellers,” Rastogi said.
This includes analysing details like what features are more popular for specific products—like customers’ preference for coconut oil in glass bottles instead of plastic or opting for luggage that comes with locks approved by the US Transportation Security Administration.
Rastogi said the e-tailer also filters search results to show products in line with regional preferences. For instance, a search for sarees in Chennai would throw up a different set of results compared to a similar search in Ahmedabad, due to local preferences.
Rastogi’s team is also currently focusing on the grocery business—Amazon’s big bet in the country. Using machine learning tools to improve customer experience and help people make better purchase decisions by creating a better question and answer engine through analyses of customer reviews are some of the things the company is working on.
It has also been using machine learning to validate customer addresses and fix mismatches in pin codes and identify whether they are commercial or residential buildings to ensure more effective and faster delivery.
It has also been able to weed out catalogue defects like incorrect product descriptions or information – like analysing an image to determine the colour of a product if it has been left blank by the seller, ensuring that it shows up in relevant searches.