Machine learning is often mentioned but also often misunderstood, leading to a lot of confusion about what it actually involves and whether it’s truly as useful as it’s said to be (blockchain is a useful point of comparison, being presented in vague terms and thought to have limitless potential). The obvious question, then, is this: how valuable is machine learning really?

Before we consider that, let’s lay out what machine learning means. I would describe it as using iterative computer systems to achieve steadily-improving results. You provide data, choose your success goals, define some viable actions, and set the results to feed back into the data. The longer you use machine learning on a particular task, the better the results should get.

Having established that, we can return to the question of what it offers — and the truthful answer is that it’s already proving transformative throughout the business world, with online retail being one field that’s ahead of the curve. In this post, we’re going to look at 5 smart ways it’s being used in the ecommerce industry. Let’s begin:

Boosting service with chatbots

Chatbots are incredibly useful for enhancing customer service. A human support assistant can only multitask to a limited extent, needs to take regular breaks, and can’t work all hours of the day and night — but just one chatbot with enough processing power behind it can scale indefinitely and run on a 24/7 basis. But where does machine learning enter into it?

It comes down to a chatbot’s actions. How does it process natural language? How does it greet someone? How does it react to a particular query? Suppose that a visitor asks something that a chatbot can’t parse, and it gets escalated to a support assistant for resolution: that resolution can be automatically passed back to the chatbot’s library data, leaving the chatbot more capable than it was beforehand.

Automating website A/B testing

The design of a website is extremely important (just take a look at some of the top site designs to see how much appearance impacts first impression), and even a minor tweak in the layout can make a big difference in conversion rate. This is why ecommerce merchants tend to invest in A/B testing (creating two versions of a page and running them in competition to see which one performs better) — but it\s an arduous process making all those changes.

Machine learning is a perfect fit for this process. You identify the elements that can be changed as well as the targeted performance metrics and just let it run. It won’t work for everything, of course — writing creative copy, for instance — but it can massively overhaul your design and achieve an impressive layout.

Providing dynamic recommendations

Personalized recommendations tend to sell very well. Amazon pioneered the system years ago, and other industries have been playing catch-up since then. It’s only with the entry of machine learning into the mainstream that it’s become possible for average businesses to implement compelling recommendation systems.

Such a system can recommend certain items, see how they perform, them keep them or replace them as needed until the best conversion rates are achieved. Given how much revenue can be earned through upselling and cross-selling, this isn’t something to overlook.

Achieving optimized pricing

Traditionally, pricing has been updated manually on an occasional basis, but this isn’t viable in a hotly-competitive online marketplace. Since shoppers can compare prices in moments, and easily buy from different stores if their usual sellers are charging too much, any retailer putting too much of a premium on its products will see its sales plummet.

The problem, of course, is that prices can change unexpectedly. Machine learning is the solution. Monitoring the prices across the web for each product, it can adjust a particular store’s pricing in real-time and see how the conversion rate is affected. Over time, it can discern which products can be given healthy profit margins and which can’t, leading to greater profitability.

Maintaining stock levels

Managing stock for a virtual store is easier than doing the same for a physical store, because you don’t need to have items at a specific location ahead of shipping them — but it’s still a complex matter that can cause major problems if mishandled. Overstocking something that subsequently loses its appeal might demand damaging reductions, for instance.

Understocking is often worse, of course: with Amazon having massively raised the bar for service, shoppers can be extremely demanding, and not having a popular product in stock will lead to lost sales and reputation damage. By reading into buying trends over time, a machine learning process can optimize restocking, ordering the right amounts at the right times.

The basic formula of machine learning can be applied to so many things throughout the tech world, so this list is far from comprehensive — but it does give a solid indication of how this process is reshaping the online retail world.

–EOF (The Ultimate Computing & Technology Blog) —

推荐阅读：
地图管理条例（国务院令第664号） 　 中华人民共和国宪法　 全国社会保障基金条例（国务院令第667号） 　 2016年国务院关于修改部分行政法规的决定　 居住证暂行条例（国务院令第663号） 　 国务院关于修改《建设工程勘察设计管理条例》的决定（国务院令第662号） 　 国务院关于修改《中国公民往来台湾地区管理办法》的决定（国务院令第661号） 　 存款保险条例（国务院令第660号） 　 博物馆条例（国务院令第659号） 　 侵害消费者权益行为处罚办法（工商总局令第73号）