Website Guide: Should you recommend beer to buy diapers?

Our American peers have long discovered that beer on the edge of a supermarket diaper can boost sales. Because in the United States, the father who is responsible for buying diapers, when they are sent to the supermarket shopping, will conveniently bring back a few bottles of beer.

So what should you recommend when a customer buys a diaper on an electric dealer's website?

This depends on personalized recommendation technology. One of its basic ideas is to use "swarm intelligence" to speculate about what products customers like, through historical data from all customers on the site.

Now there are two main ideas:

1. Based on user recommendation

This is based on the similarity of the customer, is a customer and which group of people more similar? Give him the recommendation to buy this group of people. NET Bo Station guide Rui Purchase: The Tube should buy urine in the slice of the line recommended beer?

One of the most popular technologies currently used in the industry is collaborative filtering based on users. The name is very advanced, but the truth behind it is very simple: flock together. People who buy the same goods often have similar preferences.

For example, customer a bought "Bill Gates biography", B bought "Jobs biography". Through the data, A and B have bought a lot of the same book, judge their preference is similar, said a and b these two customers are more "similar", you can buy the "Steve Jobs biography" recommended to A.

A famous retail website based on user collaborative filtering is cdnow. This is a music album of the retail website, the website has a personalized recommendation module: Mycdnow. Mycdnow is actually a personalized store, and everyone's Mycdnow shows their favorite albums.

This personalized system is a reflection of Amazon founder Bezos: "If I have 1 million to customers, I should have 1 million stores." "CDNow is recommended based on data from consumer ratings, and the site knows what albums the user a bought and how they rated the album." Based on this data, CDNow uses collaborative filtering to find the neighbor of user A. Then, the album that the neighbor likes but a has not bought is displayed on the Mycdnow page of a.

Jack uses collaborative filtering techniques to help the site solve problems. Jack developed a "swarm intelligence platform" to help the site increase consumer clicks and purchase conversion rates. Its solution is to connect customers with people who are similar to them, and let members of a group do the shopping.

2. Product-based recommendations

The main idea is to determine which products are more relevant. If many users have iphones and accessories at the same time on the site's record, the two items are relatively relevant.

Another example, many customers purchase records are "Bill Gates biography" and "Jobs biography" These two books, you can speculate that the two books are relatively relevant, there is a relatively high "relevance." If a new customer is found to have bought one, we can recommend another one to C.

If a person who buys a product a never buys a product B, the person who buys the product B will not buy the product A, then the two goods are far away, the correlation is low.

The advantage of this association is that for large retail sites, their number of users is far greater than the number of products, counting users who are more similar may be counted 10 million times, but the calculation of the product between 100,000 times.

Interestingly, many retail websites now mix applications based on product recommendations and other technologies to achieve good results. The most popular hybrid application is the combination of referrals and social networks. Mobile application services company Goodrec recently put social networks into the personalized recommendation system, mainly through the use of customer friends, family rating information for product recommendations. It allows customers ' friends and family to do "shopping", recommending the products they buy to their customers. For example, your friend recently bought a book, he has a good evaluation of the book, Goodrec will recommend this book to you. Goodrec can also help customers buy gifts, you often see some MP3 recently, your friends will receive a recommendation: "If you want to give him a gift, send a MP3."

Personalized recommendation technology can achieve customer loyalty. Retail sites provide customers with a valuable shopping experience, and for customers, they are more efficient in their shopping processes and do not have to worry about looking at unrelated items. As customers visit the site more frequently, our preferences are predicted more accurately, thus giving customers a higher value.

In the future, personalized E-commerce will become more accurate because of social media and mobile Internet. On SNS and Weibo, for example, you can purchase items without leaving the page of social networking sites such as Facebook. More importantly, user networks, access histories and shopping histories on social networking sites provide richer data for personalized product referrals.

Future retail sites may start to talk to you when you log in: "Today you are not feeling well, you have just read a ' How to reduce the pressure of work ' article, we have some related books, or to see?" ”