To interpret the data, I found a very useful seven-step standard method

In our daily lives, we always see various data tables or charts, but how do we know the information behind the data? Let’s read this article to find out.

"Take a look at the recent sales data. Do you find anything?" This kind of problem without a clear goal often occurs at work. The annoying thing is that most of the time, the daily data only fluctuates a little. If you directly report a conclusion like "3% month-on-month growth", you will be accused of "I know this too, and I need to analyze it in depth!" So what should you do? Let me explain it systematically today.

There is a standard sequence for interpreting data, which is divided into 7 steps: look at the numbers, find patterns, set standards, look at the structure, make assumptions, verify the authenticity, and draw conclusions. We are not in a hurry, and we will explain step by step:

1. Look at the numbers

This is the most basic, year-on-year, month-on-month, absolute value, up, down... Daily reports are all these things. But these things are not popular. First, anyone can understand the situation at a glance, so there is no need to write anything; second, these things have no business meaning, and it is the same as not saying anything, so we must go deeper.

2. Find the pattern

If you want to go deeper, you can extend the data time to see if there is any natural pattern. This step does not require any technical content, just connect the daily reports together, but it is very useful! Because many regular data fluctuations have periodic patterns. Mastering the pattern can avoid making a fuss and misreporting. It can also keenly observe the real problem (as shown in the figure below).

3. Setting Standards

If you want to go deeper, you have to find a judgment standard. Data + standard = judgment. Only after you have a good or bad judgment can you continue to think about why it is good/why it is bad. The best standard is to have a KPI value on your head, so that you can directly compare the KPI completion rate to get a conclusion.

However, for some non-core indicators, there are no KPI requirements, so other criteria must be found. For example, by using the scenario decomposition method, the relationship between non-core indicators and KPI indicators and the numerical range of non-core indicators when KPIs are met can be found, which can also form judgment criteria and make judgments.

4. Look at the structure

After you have made a good or bad judgment, you can further think about the reasons. But before thinking about the reasons, it is best to look at the internal structure of the indicator and find the major factors that affect the indicator. This will make the focus clear and make it easier to see the problem.

For example, when looking at sales, sales focus on people, goods, and places. First, look at the internal structure from the three dimensions of users, goods, and channels to see which type has a high proportion and which type is performing well/poorly at the moment. This way, it is easy to form ideas by distinguishing the key points.

For example, when looking at the cost situation, we can distinguish between variable costs and fixed costs. Variable costs can be distinguished between product costs and marketing costs. Fixed costs can be distinguished between front-end costs and back-end costs. This makes it easier to see which part is the source of fluctuations.

With this step, it will be much easier to find the cause later, and you can get straight to the point.

5. List of Hypotheses

Some lazy students directly draw conclusions in the previous step. For example, the recent poor sales are because product A is not selling well. The high cost is because too much money is spent on promotion...

But this reason is often too superficial. First, it is possible that product A did not sell well because of other, deeper factors (there are deeper factors); second, it is possible that product A did not sell well because certain types of users are leaving (other factors are involved); third, even if product A did not sell well because it was not good, it is not necessarily correctable in the short term, and other solutions must be considered (the feasibility of problem analysis)

Therefore, if you want to go deeper, you must make clear assumptions and figure out the logic behind the problem. Many students will be dumbfounded at this step, thinking that there are so many reasons, how can I list them reasonably?

Here are two simple solutions:

1. Start with recent events.
2. Start with possible actions the business can take.

Starting with recent events can help you quickly find hypotheses that explain the source of the problem. We can first collect recent positive/negative events and then look at them one by one:

In theory: Which indicators does this event affect?

In fact: the extent of this event, the corresponding data changes

In this way, check one by one to find out the source of the problem.

Starting with the possible actions that the business may take, we can quickly find hypotheses about the business response measures. For example, when faced with declining performance, the business has three options in the short term:

1. Go on sale and send out a bunch of coupons
2. Conduct training and select a few typical examples
3. Change the copy and change the promotion link

Then, we can make the assumptions:

1. According to the past input-output ratio, promotion can boost performance
2. People are uneven, there are benchmarks to refer to
3. Promotion is uneven, there are benchmarks to refer to

Then just check them one by one.

6. Step 6: Verify authenticity

Now we have a hypothesis that can be verified. Note that we do not have the resources to conduct AB tests one by one to verify many daily data fluctuations. Therefore, the verification mentioned here is more about finding evidence. Find enough, obvious, and statistical evidence to confirm the point of view.

For example, if I receive information about the recent price adjustment of a product, theoretically, if it is a best-selling product, adjusting the price quickly due to insufficient supply will increase revenue, while adjusting the price of ordinary products out of necessity will only hurt sales. So the idea of ​​verification is:

1. What are the past sales and inventory data of the price-adjusted products? (determine the type)
2. When did the price adjustment start? What changes have occurred in sales since the adjustment?
3. How big is the impact of the price adjustment? Excluding this product, are there any other problems?

By making comprehensive use of the data, we can make a judgment.

For example, let's assume that promotions can boost sales. Then we can take out the previous promotion effect data for reference.

1. How much did you invest and how many days did you work on it?
2. How much was increased at that time?
3. At present, can we fill the gap with this quantity?

This way we can also judge whether promotions now can save the situation and what other measures are needed.

7. Conclusion

At this point we have done enough homework, and when we hand in our assignments, we can make a very detailed report:

1. The current situation is good/bad, as shown by... (Conclusion of step 123)
2. The current situation is good because... (Conclusion of step 4)
3. The deeper reason is... (Conclusion of step 5)
4. This good forecast is sustainable/unsustainable because… (Conclusion in step 6)
5. Therefore, it is recommended that... (continue to observe/take measures/collectively discuss further plans)

In the attachment, the detailed data process is included, which makes it appear comprehensive and in-depth.

8. The order of the 7 steps

Note that you don’t have to wait until someone asks you a question before you start these 7 steps. Because:

Steps 1, 2, and 3 are all basic data interpretations, which can be done at ordinary times.

Step 4: Collect recent business actions and industry events.

Step 4: Review past business actions, which are recorded in history

Do your homework well in normal times, so that when the time comes, all you need to do are actually the two things in step 5: use historical data to calculate and verify the impact.

Therefore, we often say that if data analysts want to enhance their data insight capabilities, they must accumulate more analytical experience, collect business actions for specific business problems, and review them more often, so that they can gain a deeper understanding. Every time a specific problem comes, they will have a rich arsenal of ammunition available.

Author: Down-to-earth Teacher Chen

WeChat public account: Down-to-earth Teacher Chen