转载声明：转载时请以超链接形式标明文章原始出处:
http://www.marsopinion.com/2007/12/15/make-personalized-suggestions-to-enhance-conversion-rate-for-ecommerce-website/

 

We all know, for our ecommerce website, making the right (personalized) suggestion to customers will enhace our conversion rate significantly.

But, if we do not have 1000 MIS team support (like amazon) or genenius formula, how can we make personalized suggestoin in a simpler way?

　　

OK, at least the answer is clear:

1. Find out the products this customer would be interested
2. Suggest this customer to take a look at those products

　　

Therefore, our question becomes to be: how to know what our customers want or need?

　　

There are 2 different theories (applied in different condition, I think)：

1. The customer would like the products which are similar with the products he used to be attracted by
2. The customer would like the products which usually are attractive to those people who are similar with him/her

　　

OK, now our question is:

1. How to calculate the similarity among products?
2. How to calculate the similartiy among customers?

　　

The magic formula is: ………..

I do not know.

But I have an idea.  Perhaps we do not need a formula.

What we should do is to treat every product, every customer like a cute doggy with crazy genes.

　　

What is the gene for a product? Tags.

First, the system can label the product with basic tags automatically. The basic tags inlcude: brand name, manufacturer name, category, condition, price range, mail-in-rebate, promotion, the quantity range of reviews, customer rating…everything related.

Then we should allow our Product Manager to label the products with whatever tags he/she likes. For example, we can give tags like “fasion”, “expensive”, “UI”, “Apple”, “cute” to iTouch.

And finally, we should give the right to our customers, allow them to give tags to whatever he/she likes.

　　

OK, we get a lot of “tag”s, but where is the “gene”s?

Don’t worry.

1. Let’s adjust the weight  for different kind of tags. For example, the influence of tags labeled by system or Product Manager is five times as tags labeled by customers.
2. After calculation, an iTouch may have tags like: “fasion(2)”, “expensive(7)”, “UI(1)”, “Apple(23)”, “cute(8)”, “Chrismas(1)”, “5 star(12)”,”Video(5)”, “sexy（12)”，”defectivebydesign(2)”,”mail-in-rebate(9)”, “mp3(22)”, “beautiful(9)”…..(the number indicates the weight of tag).

The list of combinations of tag and weight is the gene of the product.

　　
What about customers?

At the begining, their genes are their basic profile (gender, age range, professional, spent range on our website, how long has he/she registered…), the weight is 500 (for example), because for new customer, the browse and purchase history is insufficient for us to make suggestions. At this time, we should allow basic profile information
play a more important role.

But once they view a product page or purchase the product, they “eat” the genes of the product.

For example, the gene for a new customer is “male(500)”, “34(500)”, “IT(500)”, “2000 ~3000(500)”, “1 year ~ 2 years(500)”. After he browsed the page of iTouch, he gets the gene of the iTouch, and now his gene is “male(500)”, “34(500)”, “IT(500)”, “2000 ~3000(500)”, “1 year ~ 2 years(500)”,  “fasion(2)”, “expensive(7)”, “UI(1)”, “Apple(23)”, “cute(8)”, “Chrismas(1)”, “5 star(12)”,”Video(5)”, “sexy(12)” “defectivebydesign(2)”,”mail-in-rebate(9)”, “mp3(22)”. 

And, if the customer really purchased the product, he/she will “eat” the gene 10 times! ( Q1: Why 10 times more? A1: Because the purchase is more serious action which could illustrate his preference better. Q2: Why only 10 times? A2: Because under most circumstances, he would browse many similar products before this perchase, he has already eaten enough genes.)

　　

Finally, we have genes for produts, and genes for customers.

Find a friend good at math, ask for a formula to caculate similarity between two genes.

Find a friend good at arithmetics design (eg. someone used to participate in NOI or ACM : P), ask him to simplify the fomula (otherwise you may have to get more servers to support your website, I guess).

And I assume you already have those friends : P, and I assum we already have the arithmetic and we can continue now : P

 　　

With the magic arithmetic, we can look back to the theories:

1. The customer would like the products which share the silimar characteristics like the products he used to be attracted with.
2. The customer would like the products which usually are attractive to other people who share the silimar characteristics like him/her.

I believe you already know what you should do now:) haha

　　

Emm…In case you still have no idea what I’m talking about….for example, If we use theory 1, when the customer comes to our website, the system will suggest products which have most similar genes with him/her; If we use theory 2, the system will find out the top 5% customers who is most similar with this customer, calculate which products they are intrested in common, and make suggestions.  (certainly we can apply the theories more creatively! here are just examples! Do NOT limit your imagination~: )

 　　
The “suggestion system” not only can help you arrange promotions, personalize the advertisement, and personalized newsletter, but also can give you more insight of your products and customer behaviors . Enjoy your way, and enjoy your success!

　　

Emm…It’s the first time for me to write blog  in English on www.marsopinion.com. Your comments will be much appreciated!

Popularity: 3% [?]

相关文章

• 网络营销数据解读（六）——自顶向下，逐步求精 (23)
• 网络营销数据解读（五）——完善指标 (19)
• 网络营销数据解读（四）——目标和指标 (24)
• 电子商务网站社会化营销的几个例子 (19)
• 促销也是可以有趣的 (8)

• Previous Entry: 一周碎碎念
• Next Entry: Huntington Library