This article was originally published in Forbes by Tracey Wiedmeyer.
Frictionless checkout is getting a lot of attention with retailers these days — and for good reason. The idea of zipping in and out of a store to grab a snack or some essentials without ever having to stand in line or open a wallet is an enticing shopping experience to present to your customers. Amazon just opened its first AmazonGo location outside of Seattle, and it happens to sit right around the corner from my company’s headquarters in downtown Chicago. So far it’s a hit with our employees who have tried it out.
For retailers, cashierless checkout trades cashier labor for labor that can restock shelves more frequently, assist customers or be deployed to other, more productive areas. The idea is that the more memorable and convenient the experience, the more visits you’ll have with possibly higher basket sizes. But despite the hype and promise, it’s actually a very hard technical problem to solve.
Obstacles Of Seamless Checkout
Frictionless checkout uses a combination of sensors and cameras that monitor consumers as they shop to determine who the person is and what products they are selecting. This generates a lot of data — much of it just noise that needs to be discarded — and some of it relevant, determining who and how much to charge. Camera feeds must be outfitted with sophisticated machine learning algorithms so they can sift through the noise of each video frame and determine the relevant actions to track. This requires a lot of training data to recognize real-world products and the relevant contextual actions they are involved in, like being put into a basket or being put back on the shelf.
Continue reading at Forbes.