Data Operations: How can operations train their data thinking?

What does “data sensitive” mean? You may have been operating for three years and have never looked at the key data. A company may have been developing a product for three years but still not have a data display platform. In this case, is it impossible to train yourself to be sensitive to data?

Let’s use a case to describe what it means to be data-sensitive and how to train your data thinking.

Please listen to the question: Suppose you work in the 3C category department of JD.com, and you find that mobile phone sales have been steadily declining in the past three months, what should you do?

Key point: To solve these problems, do not start directly from the solution, but must first study the problem itself. You have to get to the bottom of the problem, diagnose its cause, and then solve it. Cultivate data thinking and data sensitivity, starting with identifying the problem.

Be aware that in some cases, you may end up at a fork in the road where you are faced with an irreconcilable conflict.

For example, let’s say you’re given a question like this: “You’re launching a product for sharing premium content, and you’ve been experimenting with a new interface for posting content. This new interface has increased the amount of time people spend on the site, but has decreased the amount of content shared. What do you do?”

If you have to choose between the two, decide based on the company's goals.

Not only does every company have different goals, but even a product may have different goals at different times.

For example: A startup company may initially consider acquiring users as its primary goal during its A and B rounds, but after its C and D rounds, it may prioritize revenue.

The focus is different at different stages. A company that wants everything will be neither dead nor alive even if it does not die.

The same is true for a career. At different stages, the focus is different. If you want everything, you will get nothing.

When faced with this kind of problem, breaking it down can help you find the real cause of the problem. The following are some common data problems. Let’s take a look at how to solve these data problems.

1. Common data problems and causes

Here, we provide some common data problems and their causes.

At the same time, using formulas to solve problems can simplify a complex problem. When you encounter a problem, try to break it down into a formula.

Profit decline: Profit = Revenue - Cost. So, either revenues are down or costs are up.

Revenue decline: Revenue = sales volume * price. So, either sales are down or prices are down.

The reason for this data problem: If there is a decline in profits or revenues, this may mean a shift in user purchasing behavior at different product levels.

Sales decline: sales = users * purchase conversion rate, users = new users + old users. So, either the number of new users decreases, or old users are lost, or the purchase conversion rate decreases.

The number of new users decreases: New users = traffic * acquisition conversion rate, so either the traffic decreases or the acquisition conversion rate decreases.

The reason for this data problem: if sales volume declines or the number of new users decreases. These two may be due to changes in the number of old users and new users of the product respectively.

Cost increase: Cost = fixed cost + additional cost, so either the fixed cost increases, such as labor, raw materials, purchasing, supply chain and other costs, or the additional cost increases, such as discount promotion expenses, postage to remote areas, etc. are all additional costs.

The reasons for this data problem: Fixed costs or additional costs, both of which may be due to supplier price increases, dealer changes in profit structure, increased returns, or a variety of other changes.

Decrease in users: Possible reasons include a decrease in the number of new users, a decrease in the number of old users, or a decrease in the length of time these two types of users stay. The number of new users decreases: the number of new users = new users from search + new users from recommendation + new users from direct visit.

So, the reason for this data problem: it may be a decrease in search traffic, a decrease in referral traffic, or a decrease in direct visits.

2. Using first principles to identify the real problem

For example, how do you figure out why profits are falling?

First of all, let’s be clear, what does profit equal?

Profit = Revenue - Cost = Sales Volume Price - (Fixed Cost + Additional Cost) = (Number of New Users + Number of Old Users) Purchase Conversion Rate - (Fixed Cost + Additional Cost)

Of course, you can further break down this formula. For example, the purchase conversion rates of new and old users may be different. You can calculate the conversion rate of new users and the repurchase rate of old users.

For problems like this, you can use the formula to break down [profit] into revenue, sales volume, number of new customers, user conversion rate, etc.

The key here is to determine the root cause of the problem. Once we figure out which variable is changing, we can analyze why the problem occurred. But this type of problem can be caused by any reason.

However, for any problem, there is a most basic problem, and this problem cannot be violated or deleted. This is what Aristotle proposed - the first principle.

Here are some suggestions. If you encounter this kind of problem, you can try this way of thinking - first principles to identify the basic problem you encounter.

For example: You need to use the first principles to first confirm where the problem lies.

1. Have all regions experienced the same changes?
2. How many business lines do we have? Has this change occurred on all products?
3. As far as we know, are competing products in the same industry experiencing the same problem?
4. Have other related products suffered the same impact?
5. Is it related to seasonality?
6. Have we made any changes to the production line?
7. Are there new competitors entering the market?
8. What differences would we see if we separated users into new and returning users?
9. What about the data on user stickiness?
10. What have our customers been saying lately?
11. Have we received more complaints than usual recently?
12. Have you noticed any changes in referral traffic?

The above is a basic idea of ​​decomposition, constantly using questions to locate the most basic propositions.

3. Clarify your goals

Before you start solving a problem, be clear about your goal in solving it.

If you start thinking about solutions right away, there is a high possibility that what you produce will fall into the same pit again, so you must break down the problem.

Once you are able to determine where the problem occurs step by step, you are not far from solving it.

Next, let’s try to break down this question and try this answer.

IV. Case Analysis

Therefore, we still need to use a case to illustrate how to solve practical problems.

Suppose you work in the 3C category department of JD.com, and you find that mobile phone sales have been steadily declining in the past three months. How do you figure out the reasons behind it?

In fact, there are two key words in this question: [mobile phone] and [steadily]. To break down the problem, we need to follow these key words.

If you just find that sales have declined, we need to determine what has declined? Still all are declining. At the same time, regarding this [decline], we also need to confirm whether it is a steady decline or a sudden cliff-like decline, as well as the extent of the decline. We must pay attention to this and make sure what this keyword means.

Mobile phone sales are down, assuming that means a significant decline. We need to know more to find out why.

Firstly, why the past three months?

This is actually very strange. There must have been certain changes in certain aspects, which may be mobile phones, 3C products, JD.com, e-commerce, or they may just be related to this specific period of time.

We can rule out e-commerce or JD.COM because if either of them were hit hard, we would definitely know about it, and at that time, we would not only be worried about the decline in sales of mobile phones, but there would definitely be global changes.

Likewise, we also exclude the decline in JD.com’s 3C sales because, if it were a problem with the entire 3C category, we wouldn’t be particularly worried about mobile phone sales.

The timing issue is also interesting. Comparing sales to three months ago is not a good option. Because, within the current year, mobile phone sales should not be fixed every month, but will vary with some factors.

Therefore, even if a comparison is to be made, it should be compared with the same period of the previous year to eliminate seasonal reasons.

At the same time, we assume that there are actually some other problems behind this issue. As mentioned at the beginning, sales volume has been steadily declining. This means that the problem is not caused by a product feature or other external negative news changes. Because in that case, sales would fall suddenly, off a cliff, rather than falling steadily.

First principles thinking is important, and it is also a kind of structural thinking.

At this time, we want to understand the sales volume within the mobile phone category. Are sales of all mobile phones declining? Or only certain brands? Are mobile phones on Taobao experiencing the same situation?

We can use crawlers to obtain the other party's public data, especially data such as sales volume and user reviews, which can be crawled and analyzed. A big company like ours must be constantly crawling competitor data. If Taobao is also affected, then we can actually assume that all mobile phones on the entire e-commerce are affected. This is largely an external factor rather than a specific internal reason.

Therefore, we assume that the mobile phones on JD.com are affected, while Taobao is not affected. In this way, we can attribute the problem to the formulas that we have previously broken down, and these specific reasons are actually the reasons that we can control.

As mentioned earlier, sales are a function of users and conversion rates. If sales volume declines, it means that either the users have changed or the conversion rate has changed.

We can divide users and conversion rates by user type:

1. JD.COM search traffic (users searching on JD.COM)
2. Browsing traffic (users who browse and find the target product category on JD.COM)
3. External search traffic (search engine traffic, such as Baidu, 360, Soso, etc.)
4. Direct traffic (users who directly open a specific product page).

Our goal is to target those visitor types that are experiencing a decline in user volume or user conversion rate.

For example, if external search traffic is down, we can investigate changes that may have caused the Page Rank (Page Rank) to drop. If you don’t see anomalies in these types of user volumes, you’ll have to look elsewhere for something that might be affecting sales.

Another example: This may be because the price of mobile phones has risen sharply. It is also possible that there is an error in the system's search function or purchasing process, affecting sales. (If this happens, can the department bonus be cancelled or can employees be laid off?)

The above is the basic process for us to deal with this problem.

OK, the logic for this kind of question ends here. Looking back at the example just now, we have firmly grasped the key words of this problem in this process: this is a "steady" decline in sales of "a certain type of product". To find the possible causes of the problem, you must start with these two keywords.

At the same time, a balance is maintained between "logical deduction" and "intuitive answer".

The logic is that you need to constantly break down this problem: sales = users * conversion rate. Then how are the users and how is the conversion rate. For example, intuition is just a guess. For example, mobile phones generally don’t sell well suddenly.

Logic is important, but intuition is sometimes needed as well.

Author: Dalin's Little White

Source: Dalin's Xiaobai