Blockchain circle

One stop hot information platform

About us:

Blockchain circle provides the latest information about blockchain, digital currency, digital wallet, exchange, metauniverse, bitcoin, Ethereum, contract, financial management and so on, and always pays attention to the latest market...

To e-commerce practitioners: you don't need machine learning and artificial intelligence. What you need is SQL

Time : 18/08/2021 Author : feanlp Click : + -
        Editor's note: with the passage of time, some interesting new technologies and concepts such as machine learning, blockchain, artificial intelligence, virtual reality and augmented reality continue to emerge, while at the same time, some existing technologies can only retreat to the second tier. In this situation, many people like to get involved with these new technologies when launching products or financing. In this paper, celestineomin, a software engineer, takes the e-commerce field as an example to demonstrate in detail why the value of traditional structured query language technology has remained the same since it appeared more than 40 years ago. Not long ago, I posted a topic on twitter, proposing the need to use traditional and existing tools to solve daily business problems, rather than blindly believing in the latest popular skills and some fancy and complex technologies.
        This topic has received a good response and triggered some interesting discussions. Some of them agree with me, while others disagree completely. They think I'm stupid and vain. I am not writing this article to convince you to accept my idea, but to further explain what I originally posted in my twitter post. We can all find that over time, some interesting technologies and concepts such as machine learning, blockchain, artificial intelligence, virtual reality and augmented reality continue to emerge, while at the same time, some existing technologies can only retreat to the second tier. Now I often hear people build excellent products supported by blockchain technology. I also know that there are many new things such as e-commerce services and social networks supported by blockchain technology.
        These days, I often hear such words: if you want to complete the financing as soon as possible and quickly, you must invest in "blockchain" related technology, even if it has nothing to do with your macro plan. Not long ago, machine learning and artificial intelligence were in the same position. Whenever someone launches a login page, there may be words of machine learning or artificial intelligence on the page, as if it is impossible to create an initial page without mentioning these words. Seriously, are you really in business? To be honest, you don't have to. Structured query language (SQL) is a technology that I have been optimistic about until now. This technology, which first appeared in 1974 and has a history of more than 40 years, is still of great importance today.
        Although this technology has also experienced some improvements over the years, it is still as powerful as before. I have been dealing with technology throughout my career, and I have also spent a long time working in the field of e-commerce, so I have witnessed the value of this technology for our growth and expansion of business. This technology is very beneficial for us, because it allows us to explore some interesting information from the collected data. These data include but are not limited to consumer behavior, shopping patterns and habits. It even allows us to make predictions about the minimum inventory units we should hold and the minimum inventory units we should not hold.
        In addition, it can also help us please our customers and re interact with those customers who left halfway. Next, let me tell you how we do it, and how you can do it. When I talk to founders and potential founders, they always tell me quickly how much they want to use AI and machine learning technology to improve customer retention and improve customer lifetime value. But in fact, they don't need machine learning or any fancy technology at all. What they need is to write structured query language correctly. Previously, I have written SQL queries to extract valuable information and insights from the data we generate.
        Once I need to know who the customers are this week in order to 1) recognize them and 2) reward them. The simple but unexpected measures taken by the company towards customers will always make them feel ecstatic and turn them into communicators. On social media platforms, like "Wow, I became a customer of Konga (Nigerian e-commerce) this week, and they even rewarded me with a voucher of 2000 naira. I didn't expect that, thank you, you are the best!". Facts have proved that this is more effective than spending money on advertising. Don't get me wrong. Traditional advertising still has its value, but no marketing method is more effective than personal recommendation from your trusted friends.
        And the key is that it is not so difficult to obtain this information. Besides excellent traditional SQL, you don't need any fancy technology. In order to lock in customers this week, we have compiled an SQL to select the largest order of this week from the order list. After getting the customer information, we will send a thank-you letter to the customer by email, with a coupon or voucher attached. Guess the effect? 99% of them have become repeat customers. We don't need machine learning. We just write a simple SQL and get this information. Once, we need to get in touch with customers who haven't shopped on our website for a while.
        As the person in charge, I wrote an SQL query to collect customer information that has not been shopping for three months or more. This is very simple to write. For example, you can select an order from the order list that the last shopping time was 3 months or more ago. When we get these information, we will send an email to the corresponding customers, "we miss you, welcome to come again, with coupons". After taking such an approach, we found that the customer conversion rate gained from this has always remained above 50%. In my opinion, these two strategies are much more effective than advertising investment on Google and Facebook, which was the case in the past and is still the case today.
        We also apply the same idea to the briefing sent to customers. I mean, if you can personalize this, why send a generic briefing to everyone? What is the solution? I wrote SQL queries to view shopping cart items and extract individual items. Based on these items, we can create a briefing and cover relevant content. For example, suppose a customer buys a pair of shoes, sunglasses and a book, then in the briefing given to him, we can show items including shoes, sunglasses and books. Obviously, this is more meaningful than randomly recommending items. Why send a promotional email recommending breast pumps to a man who just bought a pair of sneakers? This obviously makes no sense.
        The click through rate of most marketing emails is in the range of 7% to 10%, but when we realize personalized matching, we find that the click through rate has reached 25% to 30%. This is almost three times the industry standard. In addition, in these emails, we will start with the name of the recipient, rather than the general name of dear customers. This makes these emails more human and shows our concern. All this is achieved through excellent SQL, rather than fancy machine learning. For those customers who cannot complete their orders for one reason or another, we will not give up easily. As long as customers add products to the shopping cart, it indicates that they are interested in buying.
        In order to let such customers complete the checkout process, I wrote a good SQL script and cooperated with a cron planning task. This combination will lead to the instruction of sending an email to those customers whose shopping cart was last updated for 48 hours or more. Guess what? This is an effective approach. Because we can track these emails, we also know whether they will come back to complete the order payment process because of these emails. Of course, this SQL is also very simple to write. It will select those orders whose shopping cart status is not empty and the last update time is greater than or equal to 48 hours.
        We set cron to start running at 2 a.m. every day, because there is less activity and traffic at this stage. When customers wake up and see our email, they will think of the previously shelved shopping cart. There are also no fancy technologies in this process, only SQL, Bash and cron. In addition, cash on delivery is now very popular, for which SQL can also come in handy. For those customers who cancel orders for three consecutive times, we will put them into the high alert category. Next time these customers place an order, we will call them to confirm whether they really need the ordered items. In this way, we save time and unnecessary pressure.
        In the field of e-commerce, logistics costs are expensive, so we need to focus on those who really want to customers. We don't need machine learning or fancy AI technology to solve this problem. Similarly, we just need good SQL. We will also use SQL to manage customer expectations for those customers who meet the service level agreement standards but whose orders are not delivered in time. From the order list, we will select those orders whose order date =7 days (standard delivery period), but still in the undelivered status. We will send emails and text messages to customers in combination with cron planning tasks. Although they will not immediately give feedback and praise, at least this action can let them know that we care about these problems and are trying to solve them.
        After all, nothing is more annoying than late delivery. This special solution also has a great impact on our net recommendation. In that sentence, excellent sql+bash saved us. In addition, siftscience (a company that uses artificial intelligence / machine learning to develop network security applications) has done a very good job in preventing fraud, but in fact, SQL can also be used for this. If a person tries to pay with three different cards at the same time, these cards will rebound. The first and most obvious thing to do here is to temporarily block their accounts. In this way, you will save potential bank card owners a lot of trouble.
        You don't need to store the details of your bank card, just the checkout attempt of this specific order number. These are very simple things. Machine learning is also not required, just write an appropriate SQL. Personally, I also like machine learning and artificial intelligence, but if you operate a small online store with 1000 to 10000 customers, you can still rely on SQL. After all, talents in machine learning and artificial intelligence are not everywhere.
Previous:Defi accelerates the new development of blockchain smart contracts
Next:No more

Related articles:

© 2005-2032 | Blockchain Circle & & All Rights Reserved    Sitemap1 Sitemap2 If there is infringement, please contact us at: