What does it mean to be data sensitive? How does APP do data analysis?

Today I will talk about my understanding of data.

1. Split from the data dimension to make the goal more realistic

I have been engaged in e-commerce operations for nearly two years, and one point that impressed me most was the splitting of goals from the dimension of data.

Tmall ’s Double 11 just passed, and Jack Ma created a new achievement of 91.2 billion. From 57.1 billion last year to 91.2 billion this year, how dare Jack Ma say that he can make 90 billion this year? Before setting this goal, it is necessary to break down the goal.

The 90 billion transaction volume is first divided into each category according to the proportion of categories in the past, and each category is responsible for a certain sales target. The categories are then divided into each seller according to the proportion of sellers' past transaction volume, and each seller is responsible for a certain sales target. Sellers can then reversely calculate how much traffic is needed based on their daily store conversion rates . By combining each category with the traffic that the platform can provide, they can get the traffic gap. Next, by calculating the cost of acquiring traffic through each channel, we can determine the amount of marketing funds that the Double 11 platform needs to invest. Through such a split, the entire 90 billion target becomes clear and feasible.

No matter what you do, if you want to succeed, you can't do without breaking down the goal. Any abstract thing can be solved through mathematical methods. Datafication will make things simpler and more executable, and it will also be easier to evaluate the results.

2. Many businesses are actually a formula

When I first started e-commerce and received business training, the first class only taught one formula.

Transaction amount = number of buyers x average order value

If you want to increase transaction volume, you can either increase the number of buyers or increase the average order value. We can take stock and see that among all the promotional methods we have seen, there is none that is not aimed at increasing these two values. Discounts for purchases over a certain amount, free gifts for purchases over a certain amount, and buy two get one free are means to increase the average order value; flash sales and group buying are means to increase the number of buyers (the core of flash sales is to gather a large amount of traffic for related sales).

Not only that, this formula can also be broken down into multiple forms based on different business scenarios.

Number of buyers = business details UV x order rate x payment rate

Business details UV = Ad impression x Ad conversion rate = Search impression x Search conversion rate = Activity impression x Activity click rate

Therefore, the factors that determine the transaction amount become the conversion rate of each channel, the click-through rate of pictures, the order rate of products, and the payment rate. These many details jointly determine the final transaction amount. Next, we optimize these details separately. This process is called refined operations based on data.

If you think about it, isn’t your own business also a formula? Try to find your own formula and break it down, and you may find many ways to improve it.

3. Operations is ultimately a funnel

Under the Internet model, no matter what product is made, the fundamental purpose is to realize it, and as long as it is monetization, it involves conversion. And conversion is actually a funnel model.

The funnel model is the most frequently mentioned word in operational data. In the business chain, the number of users in each link is constantly decreasing. What operations have to do is to try every possible way to improve the conversion rate of each link in the funnel.

For example, the funnel model of an e-commerce activity page should be like this:

With such a funnel, I can analyze what each link represents and how I can improve it:

1. pv/uv: Page visit depth directly reflects whether the page is attractive and whether users are interested in clicking on the page.
2. Activity page -> Details page UV: Whether the content on the page is attractive and whether the products are liked by users, products with poor click effects need to be replaced in time according to the page click situation.
3. Details page uv—>number of orders: what is the conversion rate of the product, whether it is a hot-selling product, if the conversion rate here is too low, it needs to be replaced with a high-conversion product.
4. Number of people placing orders -> Number of people paying: If the payment rate of a product is lower than the normal value, the seller needs to urge the order.

It should be noted that the funnel model needs to be compared. If there is only one funnel model, it is just a display of data. If you want to do analysis, you must compare it, such as comparing it with previous funnels, or comparing it with the platform's average. Problems can only be discovered during the comparison process.

As product operators , we must be familiar with every key data in our product, such as the average daily UV, conversion rate, and download volume. This way, we can discover any abnormalities in the data in the first place. Being familiar with product data is a prerequisite for being sensitive to data.

4. What should a complete data analysis report include?

I have talked about some theoretical aspects before, and finally I will give you a data analysis template for your reference.

1. First, you need to determine your target achievement rate, completion percentage, and improvement percentage based on the activity goals.

This is the result of this activity, written at the beginning. like:

This event UV 24w (20w, ↑ 20%), UV value 3.6 (3, ↑ 20%)

2. If you are publishing weekly or monthly reports, the next step should be the core data trend chart

In this graph, we need to analyze the turning points of each data. For example, the UV value on November 7 and 8 in the graph has increased significantly. The reason for this needs to be found and written in the report.

3. The next step is traffic analysis, which mainly focuses on the distribution of traffic sources and the analysis of traffic conversion rates of each channel.

When traffic increases, we need to find out which channel brought the traffic increase, why it increased, and analyze the reasons. What is the quality of the traffic and which channel has a high traffic conversion rate. Two pie charts are needed here, one is the proportion of traffic channels, and the other is the proportion of conversions brought by the channels.

From the two pie charts above, we can see that the conversion rate of in-site traffic is obviously higher, while the conversion rate of traffic brought by GuangDianTong is relatively low. In addition, by comparing the proportion of channel sources with previous periods, we can see the changes in the current traffic structure.

3. Conversion rate analysis, also known as funnel model analysis.

As mentioned earlier, the funnel model requires comparative data, so in the analysis here, we need to list two funnel models.

The analysis of conversions in each link of the funnel model mainly compares with past data, combines multiple factors such as activity pages, traffic, product functions, etc., and tries to analyze the reasons for the increase or decrease in conversion rate of each link here.

4. Module click analysis

When we design a product page or activity page, we need to know whether the structure of the page is reasonable and the user click distribution, which will help us improve it. When we try a new page style, we should analyze the module clicks here to verify whether our structure has brought improvements to the data.

Module click analysis is mainly analyzed from the perspective of click pie chart and conversion rate of each module. Click pie chart can show user needs, and module conversion rate reflects whether the content of each module meets user needs. If the module conversion rate is low, you need to consider whether the content of this module is high-quality, or even whether this module needs to change its style.

5. Improvement and optimization

Every activity always has its good and bad aspects. The purpose of our data analysis is to accumulate experience and consolidate methodology. At the end of each data report, we need to make a summary of this activity. For example, we tried a new gameplay and saw how effective it was, tried a new page style and saw whether the click-through rate increased, etc. Apply the experience to future event planning.

5. Data is not everything

Finally, I want to say that data is not omnipotent.

The data analysis we often do is based on massive amounts of data, but in start-up companies, the data system is often not perfect and the amount of data is insufficient. The data can only be used as a reference, and over-reliance on data often leads to wrong judgments.

There are many indicators in the data and many statistical dimensions. If you dig deeper, it will take a lot of energy, but it may not be effective. Therefore, it is very important to find the most critical data indicators and analyze them most reasonably.

That’s all for today. When doing data analysis, the focus is not on the data but on the analysis. Being sensitive to data means being able to clearly understand the reasons behind data anomalies. This requires experience as well as your thinking and execution capabilities. I hope you can become an Internet person who is sensitive to data.

End.

APP Top Promotion (www.opp2.com) is the top mobile APP promotion platform in China, focusing on mobile APP promotion operation methods, experience and skills, channel ASO optimization ranking, and sharing APP marketing information. Welcome to follow the official WeChat public account: appganhuo

[Scan the top APP promotion WeChat QR code to get more dry goods and explosive materials]

This article was compiled and published by (APP Top Promotion) by @酥酥. Reprinting this article requires the consent of Top Promotion , and please attach the link to this article!