Data drives business growth, real cases are here!

In the retail industry, how to increase the proportion of users placing orders on multiple platforms at the same time is a typical data-driven business scenario. This article will use specific cases to explore the working methods and thinking patterns of data analysts when facing such problems.

The previous article "Data-driven business: Are you the driver or the donkey pulling the cart" shared the basic logic of data-driven. This article will explain it with specific cases.

Problem scenario:

A retail company has both offline stores and online self-operated micro-mall. Now the boss requires the operation department to "increase the proportion of users who place orders on both platforms at the same time." The operations director said: "Data drives the business. Please ask the data team to give clear guidance." You are the data analyst of this company. Ask: What should you do at this time?

1. Difficulty Analysis

Would you do this:

• Calculate the number, percentage, and growth rate of users placing orders at the same time
• Calculate the average payment amount and payment frequency of users who place orders at the same time
• Compare the top 10 products purchased by users who placed orders at the same time vs. users who placed orders through a single channel
• Compare the gender, age, and region of users who placed orders simultaneously vs. users who placed orders through a single channel
• …

After doing the above things, do you feel a deep sense of powerlessness?

Yes, this is the biggest difficulty of data-driven business: data cannot directly lead to a business action. 60% of users who place orders at the same time are women, and 55% of women who use a single channel account for women, so there are 5% more women who place orders at the same time.

So what? So we should promote women? This idea is too stupid! It's just 5 more points, what does it mean? The comparative analysis of other dimensions is similar. Perhaps the closest way to directly derive a conclusion is to look at the ranking of purchased products, and promote the one that sells more. However, in reality, it is easy to be criticized by the business to nothing (as shown below).

Many students start to feel confused after achieving this, so they search the Internet for "The great god from Touteng'a Company, you can get it for a fee!", search for the PDF version of "21 Days to Master Online and Offline Order Analysis", and so on.

2. Key points to solving the problem

The key to solving the problem is to link business actions with data. Use data to prove judgments and collect data in action, so that data and business can be combined. For example, for the task at hand, pay attention to the question. The boss only gave a direction: increase the number of people placing orders on both platforms at the same time. As for how many people are there now, how much to increase, and why to increase? No information was given at all.

So at this time, the first task is to clarify the direction. It is not the time to think of a specific plan. The first judgment related to the plan is: should we do it or not? Note that the boss did not mention specific numbers. It is very likely that he just vaguely felt that he should say so. Therefore, both the positive and negative conclusions of "do it" and "don't do it" can be verified with data (as shown in the figure below).

Similarly, all judgments in the entire decision-making process can be listed in the form of hypotheses to be verified, so that the pure data such as "currently users account for 20%, the monthly growth is 50,000, and 75% will continue to place orders at the same time" can be converted into a conclusion that is meaningful to the business: "There is room for improvement, the growth rate is not high, and there are rules to follow." In this way, the data can be continuously used to deduce. The same problem can also be demonstrated from different angles and directions. Students can think about the specific ideas themselves.

Assume that the current situation is: "Orders are mainly placed in offline stores, with few online orders. There is room for improvement, but the growth rate is not high and there are regularities to follow." What can be done next?

Since there is a pattern, we need to see what the pattern is. Here we need to use comparison. Note that making a comparison and making a random comparison are two different things. It does not mean listing a bunch of columns and putting whichever one is higher. Instead, we need to base our analysis on our own assumptions. For example, there may be four patterns behind the fact that users do not go to retail stores but instead go to WeChat malls to place orders. Therefore, we can find the corresponding dimension data to verify them (as shown in the figure below).

Because we are still in the process of exploring the direction, we should try to sort out the subdivided issues that affect the judgment of the general direction in advance, such as:

• What are the specific criteria for “placing orders at the same time”?
• How many people currently meet the criteria?
• Are the number of people who meet the criteria increasing, decreasing, or remaining the same?
• What are the qualitative and behavioral differences between those who meet the standards and those who do not?
• Are people who meet the standards and those who don't meet the standards related by evolution or are they two different groups of people?
• Why do you want to increase the number of people who place orders at the same time?

After doing these homework, communicating with the business/boss is much more meaningful than asking "why?" Many idiots directly ask a bunch of "whys", but end up being scolded: "What's the point of you!" "Don't you have a brain!" "What analysis!".

With homework, you can:

• You can actively guess the purpose of the action. If you guess correctly, you will immediately make people think that you understand the business well.
• Can actively prompt verification results, whether it is true or not, it shows that you think carefully
• Can actively guide the direction, tell everyone which rules to follow, and make the analysis feasible

In short, bosses like to come up with a rough idea and let their subordinates do a lot of very detailed work. Many successful data analysts have risen to the top by this method, so remember this.

3. Implementation Methods of Data-Driven Business

However, having a direction alone is not enough. Many businesses like to say: "implementation". So how to implement it? Note that in implementation, data analysis methods cannot directly generate creativity, but are more about summarizing past results and testing the effectiveness of creativity. However, the actions of the business side can directly generate creativity, and do not necessarily require data support.

For example, if the business side launches a free delivery service within 2 kilometers of the store for online orders, it is very likely that online orders will increase significantly. If it has never been done before, or it has been done but without data collection, then even a good cook cannot cook without rice, and you will not be able to analyze how it is done.

For a new solution, it is necessary to carry out detailed implementation and testing to truly confirm the effect. For example, if a free delivery service within 2 kilometers is a new business, the business party needs to provide a detailed plan before testing.

Therefore, at the implementation stage, the most critical issues are:

1. Have you done any related actions before?
2. If you have done it, what were the results and what was the input-output ratio?
3. If you haven't done it, is it because you haven't thought about it, or if you thought about it but felt there were obstacles? What are the obstacles?
4. What do your peers do and what results can you find out?
5. Are there any preliminary plans, and if so, what are they?

Similar to the direction setting stage, these five questions are first asked by data analysts themselves. To be sensitive to the business, you should collect more activity announcements and version update information, and then the idea of ​​implementation will be clear. Simply put, it is: when there are a lot of actions, do optimization and select the best action route; when there are not a lot of actions, do testing and explore feasible paths.

Of course, it is very likely that even if you do this, your business will still say "not specific!", and ask for a solution from the data, specific to how many pages there are, how many buttons there are on a page, how to write the page code, how to draw an event poster, and whether it should be drawn with 3 strokes or 5 strokes...

If you really want to dump all these business tasks to data, data analysis can directly suggest: it is recommended to replace the business with those who are capable of making posters and H5. You can even further analyze it based on the recruitment information: the monthly salary of business with these capabilities is only 8-12K, and it is more cost-effective to fire the current XX, thank you.

IV. Summary

Of course, during the execution phase, data can also be used for monitoring and problem diagnosis. During the review phase, experience can also be summarized. These are all driving actions that can be taken.

The reason why we emphasize the planning and design stages is that data-driven business makes the most mistakes in these two stages. They are often:

During the planning stage, data analysts work in isolation, without integrating business aspects or clarifying goals, and blindly hope that "super awesome models" and "national unified templates" can sort out the problems.

During the design phase, the business department blindly passes the buck and relies on data for everything. They have no opinions or ideas and wish the data could do all their work, otherwise they would say: It’s not specific enough.