E-commerce activity review: How to efficiently achieve data-intensive operations?

In an era where everyone is talking about operations and big data , everyone will say to operations: "We must speak with data and achieve refined operations", but how should we do it?

1. Four prerequisites for refined data operations

1. Obtain the data needed for operation in a timely manner:

• Clarify what data should be obtained, such as order volume, registration volume, reading volume, page views, visit duration, etc.
• Data can be obtained, but not all data can be retrieved. Only data that was collected in the early stage can be obtained.
• Obtain data in a timely manner. Many companies cannot directly obtain data through operations. Generally, they need to communicate with technology first to clarify needs and schedules. And a lot of data is time-sensitive. For example, during an event, the potential purchasing user ID is not obtained in time, resulting in a delay in sending discount information, and the user purchases the product through other channels .

2. Reasonably define the dimensions and indicators of data analysis :

The closer the "defined dimensions and indicators" are to business needs, the more the true value of the data can be brought into play. However, many companies have a vague division of data. Even if they can reasonably define it during analysis, they cannot perform analysis because they have not collected data on these dimensions in the early stages.

3. Select and use efficient data analysis tools :

Choosing the right data analysis tool can do more with less effort. A good data analysis tool should not only meet the data analysis needs of the current business, but also meet the data analysis needs as the amount of data increases and the business changes during the development of the enterprise.

Therefore, tools such as Excel, SPSS, SAS, SQL, Clementine, R, Rapid-miner, etc. may be used. The requirements for operators to master these tools are too high. The financial resources and energy spent on training an operator according to such standards are equivalent to those of training a data analyst.

4. Have strong data analysis ability and be able to combine it with practical work:

In short, data analysis ability means being able to discover problems, summarize patterns, and provide optimization solutions from complex data. To combine it with actual work, operators are not only required to have a deep understanding of the business, but also to have strong logical analysis capabilities in order to integrate data with actual business.

So, is data-intensive operation a false proposition? Of course not. Below, we take the review of the 618 event by Xiao Zhang, an e-commerce operator , as an example to explain how to efficiently achieve data-driven operations.

II. Background

During the 618 period, Xiao Zhang, an operator of a small beauty e-commerce company, tried a brand new 15-day promotion called "Mid-Year Promotion, Compensate You to the End", which was quite effective. Therefore, the leader asked Xiao Zhang to review the activity for everyone and emphasized the need to use data to speak.

During the 618 event (June 6-20), Xiao Zhang made two main attempts as follows:

1. Personalized SMS reminders. During the event from June 16 to June 20, in order to focus on guiding users to purchase, personalized SMS reminders for different users were added. For example, send the message "The product you want to buy has been reduced in price!" to users who add products to the shopping cart before the event; send "Facial mask coupons" to users who browse facial mask products during the event, etc.

2. Refine the recommendation model. The main focus is on refining the personalized recommendation model of the detail page, for example, adding recommendations for products purchased during promotions for users with the same consumption habits.

After completing the necessary work for the day, Xiao Zhang began to prepare for the event review. Xiao Zhang first sorted out the key points of the review:

• The overall effect of the activity. For example, total sales, total order volume, total number of purchasing users, etc.;
• Analysis of user behavior during the event. For example, access time, geographical distribution, coupon usage distribution, etc.;
• Analysis of active products. For example, sales analysis of different categories of goods, sales analysis of similar goods, etc.
• Activity resource allocation analysis. For example, banner position, pop-up window, related recommendations, etc.
• Activity flow analysis. For example, the number of traffic and ROI from different channels.

3. Overall effect of the activity

The overall effect of an event is a comprehensive evaluation of an event. The total sales, total order volume, total number of purchasing users, average customer unit price, etc. of the e-commerce event need to be analyzed, and the month-on-month and year-on-year figures must be analyzed to see the specific growth.

Xiao Zhang found out from the data that the sales during the 618 promotion period were about 20 million higher than the same period last year, a month-on-month increase of 40.9% (as shown in Figure 1) and a year-on-year increase of 55%.

Note: The pictures in the article are all from Sensors Data products, and the data involved are all virtual.

Figure 1 Comparison of total sales during this year’s 618 promotion and last year’s 618 promotion

Xiao Zhang further analyzed and found that the sales from June 16 to June 20 were very good (as shown in Figure 2). Analysis shows that this may be related to the personalized SMS reminder attempt of this event. The conversion rate of reminders to promote orders is as high as 30%, which is twice the original conversion rate.

Figure 2 Single-day sales during the event

Analyzing the overall effect of the activity is only the first step in the review. Whether the effect is good or bad, further segmentation analysis is needed to analyze the reasons and make subsequent optimization suggestions. For example, we learned above that personalized SMS reminders can effectively improve conversion rates. In subsequent activities, we can set more personalized reminders or extend the cycle.

4. Data analysis based on user behavior

In today's increasingly fierce competition, whoever can capture users will increase their chances of winning. User behavior analysis is an essential part of retaining users, and it is also a point that must be emphasized in every review.

Xiao Zhang conducted a series of analyses on consumer users’ access time, geographical distribution, coupon usage, etc. The following is an example of Xiao Zhang’s analysis of users’ coupon usage (as shown in Figure 3).

Figure 3 Distribution of coupon amount usage per day during user activity

Figure 4 Distribution of overall coupon amount usage during user activity

Setting a reasonable discount level for a big promotion will achieve twice the result with half the effort. This discount level should be attractive to most people and maximize the profit after the discount under controllable costs.

As shown in Figure 3, the largest proportion of users using coupons during the activity period is 50-100 yuan per day. As shown in Figure 4, the largest proportion of users using coupons during the entire activity period is 100-200 yuan, indicating that users are likely to use coupons repeatedly, and subsequent activities can be optimized based on this situation. For example, increase the usage rate of high-value coupons, because generally high-value coupons often correspond to high order amounts; set up single product coupons, etc.

There are many points that can be analyzed in user behavior. Regarding the distribution of coupon usage mentioned above, we can further analyze the use of coupons by users in different regions. For example, the amount of coupons used in first-tier cities is large, while the amount of coupons used in third- and fourth-tier cities is relatively small, so coupons of different amounts can be sent to people in different regions. By analyzing data , you can unlock many operational codes instead of walking through a maze where you don’t know which path is right.

5. Analysis of Product Activities

One of the core competitive advantages of e-commerce platforms is product competitiveness. The value of products directly affects whether users are willing to buy, and product recommendation management will directly affect product sales and thus total sales.

Xiao Zhang Company is a small beauty e-commerce platform, and its product categories include cleansers, lotions, essences, facial masks, sunscreens, etc. Xiao Zhang first analyzed different categories of goods and found that the sunscreen category accounted for the largest share of total sales in this event. This is as expected, as the weather gets warmer, sunscreen has become almost a must-have and most used item for women every day.

Xiao Zhang also analyzed different products in the same category. Taking the sunscreen category as an example (Figure 5), it can be seen that the top-ranked sunscreen is ZIL AI X sunscreen spray (the same sunscreen as the Douyin influencer), the second is RECIPE sunscreen spray, and the third is Anessa. Therefore, when expanding product lines in the future, you can try popular products on Douyin, because Douyin has already increased product awareness and there is no need to educate users.

Figure 5 Distribution of sales amount of the top five sunscreen brands

In e-commerce activities, high-value goods are generally divided into two categories: one is goods with low unit prices and large sales volumes, and the other is goods that may not have the largest sales volumes but have high unit prices and high total sales volumes. Xiao Zhang obtained the product IDs of these two types of goods through analysis, and created product groups for them to facilitate subsequent activity calls. However, during the analysis, Xiao Zhang discovered an anomaly: the sales volume of a sun-protection shirt that was usually popular was abnormally low. In order to find out the reason, Xiao Zhang analyzed the conversion funnel (Figure 6).

Figure 6: Purchase conversion funnel of an abnormal sunscreen shirt

As can be seen from the above picture, the sun protection shirt had no sales after June 9. After further analysis, Xiao Zhang found that the sun protection shirt was sold out on June 9, but because he was too busy during the promotion period and did not check the inventory in time, he did not replenish the stock in time, resulting in a huge loss.

From the above examples, we can see that through data analysis, we can not only find the problem, but also find optimization solutions. For example, if it is found that losses have been caused by not discovering the out-of-stock situation in time, an inventory alarm mechanism can be set up in the future. When the inventory is less than 20 pieces, an email alarm will be automatically sent to the product operator to allow the operator to replenish the stock in time. In addition, the product inventory must be checked before each promotion to prevent out-of-stock situations during the promotion.

6. Resource allocation analysis

Resource allocation in an activity is an extremely important matter. A good allocation method will bring exponential growth to the activity effect. Similarly, an unreasonable allocation method will also bring great negative impact.

In the past, Xiao Zhang could only see the PV of each resource position. He could not judge the value of the resource position through a single data. Many times the PV was shown to be high, but in fact no actual conversion was formed. Xiao Zhang has always thought that since the banner has a lot of clicks, the conversion rate must be very good, so he spent a lot of time thinking about it. But this analysis found that the conversion rate of the banner position is actually not good.

Therefore, it is necessary to optimize the placement of banner positions. For example, when clicking on the banner, the user will not be redirected directly to the product details page, but will first be redirected to the coupon page for different products and then to the product purchase page.

Figure 7 Conversion of each resource position

Xiao Zhang also found that the purchase rate of recommended products at the bottom of the details page increased significantly (as shown in Figure 7), which is related to the new recommendation mechanism adopted in this event. This time, a new personalized recommendation mechanism is adopted, that is, relevant segmentation models are set according to the user's original browsing behavior. When the user's browsing behavior meets the characteristics of a certain segmentation model, recommendations will be made according to this model.

7. Analysis of traffic flow from activities

For e-commerce platforms, total sales = number of users * average customer unit price. The average customer unit price is often relatively fixed. To increase total sales, it is mainly through increasing the number of users, that is, traffic . At this time, it is necessary to divert traffic, especially during big promotions.

In today's business environment where fake traffic is rampant, it becomes more important to identify good channels, good delivery methods, keywords , etc. to obtain high ROI. Therefore, Xiao Zhang also conducted some analysis on the delivery channels (Figure 8):

Figure 8 Traffic diversion from different channels

As can be seen from the above picture, Baidu has the largest number of traffic, and Didui has the least traffic. But is Baidu traffic better than all other channels?

Of course not. To judge which channel is better, we need to further analyze the quality of the users attracted, that is, whether these users have registered and purchased. At this time, the registration conversion rate is a good indicator, that is, the number of registered users divided by the number of users who launched the APP. The number of users here refers to the number of users attracted by each channel. This indicator can more accurately show the channel effect. Similarly, the purchase conversion rate can also be calculated.

In addition to analyzing the channels, Xiao Zhang also classified and analyzed the keywords used in the channels. Xiao Zhang divided these keywords into four different quadrants: keywords with high traffic and high quality, keywords with high traffic and low quality, keywords with low traffic and high quality, and keywords with low traffic and low quality, and summarized the optimization suggestions:

In subsequent promotional activities, for keywords with high quality and high traffic, you can increase the delivery; for keywords with low quality and low traffic, you can save costs and deliver higher quality keywords; for keywords with low traffic and high quality, you can deliver them in channels in a targeted manner; for keywords with low traffic and low quality, you can stop delivering them or optimize them.

In the end, Xiao Zhang made a very detailed analysis report in just 3 hours (the above example is only part of Xiao Zhang’s analysis). The next day, Xiao Zhang’s operational review received unanimous praise, many suggestions were adopted and included in the scope of the second in-depth discussion.

Conclusion

From the example of e-commerce entrepreneur Xiao Zhang’s 618 event review, we can know that although it is very difficult for a company to meet the four prerequisites for refined data operations, there is a solution, which is to use a better data analysis tool.

Good operations need to be combined with one's own business, "squeeze out" the data, extract valuable and nutritious information, and form one's own data/analysis report. This not only allows you to give your leader a perfect answer, but also allows you to summarize and refine your own operations and gain growth. I hope this article helps you!

Author: Yanu, authorized to publish by Qinggua Media .

Source: Everyone is a Product Manager