Seven steps, in-depth interpretation of data meaning

“ Look at the latest sales data , What did you find ？” At work , This kind of problem with no clear goal often occurs . What I hate is ： Most of the time , Daily data is just a little fluctuating . If you just take “ Chain ratio 3% growth ” Report this conclusion , Will be detained again “ I know that, too , In depth analysis ！” The hat of . Then what should I do ？ Today, the system will explain .

There is a standard order to interpret data , It is divided into 7 Walking ： Look at the numbers 、 Looking for a regular 、 Set standards 、 Look at the structure 、 Assume clearly 、 Verify the truth 、 Draw a conclusion . We are not in a hurry , Step by step ：

The first 1 Step ： Look at the numbers

That's the bottom line , Year on year , Chain ratio , The absolute value , Up 、 fell …… Daily reports are all these things . But these things are unpopular . in the first place , As long as a person looks at it, he will know the situation , There is no need to write ; both , These things have no business meaning , Talking is the same as not talking , So we must go further .

The first 2 Step ： Looking for a regular

Want to go further , You can prolong the data time , See if there are natural laws . There is no technical content in this step , Just connect the daily reports directly , But it works very well ！ Because many conventional data fluctuate , It's periodic . Master the law , It can avoid making a fuss , False positives, false positives . You can also keenly observe the real problem （ Here's the picture ）.

The first 3 Step ： Set standards

I want to go deeper , You have to find the criteria . data + standard = Judge , With good or bad judgment , To continue thinking ： Why / Why not . The best standard , There's one KPI The value is on the head , This is a direct comparison KPI The completion rate has a conclusion .

But some non core indicators , No, KPI requirement , At this time, we have to find other standards . For example, use scene disassembly , Compare non core indicators with KPI Relationship between indicators ,KPI Find out the value range of non core indicators when reaching the standard , This can also form a judgment standard , use sb.'s judgement .

The first 4 Step ： Look at the structure

With good or bad judgment , You can think further about the reason . But before thinking about the reason , It's best to look at the internal structure of the index first , Find the big end of the impact index . So the focus is clear , It's easier to see the problem .

For example, look at the sales , Sales stress people 、 cargo 、 site , Start with the user 、 goods 、 Three dimensions of channels , Look at the internal structure alone , See which type accounts for a high proportion , Which type is performing well at present / Bad . This distinguishes the key points , Easy to form ideas .

For example, look at the cost , Differentiate between variable costs 、 fixed cost , Variable costs differentiate commodity costs 、 Marketing costs . Distinguish the front desk in the fixed cost 、 Back office costs . This makes it easier to see which is the source of fluctuations .

With this step , It will be much easier to find the reason later , You can go straight to the point .

The first 5 Step ： Column assumptions

Some lazy classmates , Draw a conclusion directly in the previous step . For example, the recent poor sales are due to A The goods didn't sell well . The cost is high because the promotion costs too much …… But this reason is often too superficial . in the first place , There may be A The reason why the goods are not bought well is due to other hidden factors （ There are deeper factors ）; both , There may be A The reason why the goods are not bought well is that some types of users are losing （ Other factors affect ）; Third come , Even if A I didn't buy it well because A no way , That doesn't necessarily mean that it can be corrected in a short time , Still have to think of other ways （ The feasibility of problem analysis ）

therefore , Further down , Be sure to make clear assumptions , Roll out the logic behind the problem . Many students will be silly at this step , I think there are a lot of reasons , How can I make a reasonable list ？ Here are two simple ways ：

• Start with recent events .
• Start with the possible actions of the business .

Start with what happened recently , Can quickly find assumptions that explain the source of the problem . We can collect the recent positive / Negative events , Then look one by one ：

Theoretically ： What indicators have this event affected

actually ： The extent of this event , Corresponding data changes

Check one by one like this , Find out the source of the problem .

Start with the possible actions of the business , Can quickly find the assumption of business response . For example, in the face of declining performance , In the short term, the business will be threefold ：

• On sale , Send a bunch of coupons
• Training , Grasp several typical demonstrations
• Copy changing , Change the promotion links

that , We can list assumptions ：

• According to the past input-output ratio , Promotion can improve performance
• The staff is uneven , There are benchmarks to refer to
• Promotion is uneven , There are benchmarks to refer to

Then check one by one .

The first 6 Step ： Verify the truth

With hypotheses, we can verify . Be careful , Many daily data fluctuations , There are no resources for us to do it one by one ABtest To verify . So the verification here , More is to find evidence . Find enough 、 The obvious 、 Data evidence , To prove the point .

For example, I received the latest price adjustment information , So theoretically , If it is a best seller , Fast supply exceeds demand and price adjustment , It will increase income , The forced price adjustment of ordinary goods will only hurt sales . Then the idea of verification is ：

• Past sales of price adjusted goods 、 How about inventory data （ Judgment type ）
• When did the price adjustment begin , What's the change in sales after the adjustment
• How much influence does the price adjustment have , Remove this product , There are other problems

This comprehensive use of data , Can judge .

For example, let's suppose ： Promotion can improve performance . Then you can take the previous promotion effect data for reference

• How much was invested at that time , For a few days
• How much did it improve at that time
• At present, according to this quantity , Can you fill the hole

This can also make a judgment ： If there is a promotion now , It can save the situation , What other measures are needed .

The first 7 Step ： Come to a conclusion

At this stage, we have done enough homework , When you hand in your homework , You can make a very detailed report ：

• The current situation is very good / Is very poor , Manifested as ……（ The first 123 Conclusion of step ）
• Good current situation , Because ……（ The first 4 Conclusion of step ）
• The deeper reason is ……（ The first 5 Conclusion of step ）
• This good prediction can be sustained / Unsustainable , because ……（ The first 6 Step conclusion ）

therefore , Suggest ……（ Continue to observe / Take measures to / Brainstorm further solutions ）

In the attachment , Attach the detailed data process , It is comprehensive , There is depth again .

7 Sequence of steps

Be careful , this 7 step , You don't have to wait until the moment someone asks questions . because ：

The first 1、2、3 Step , It's all about basic data interpretation , I can do it at ordinary times

The first 4 Step , Collect recent business actions , Industry events , I can do it at ordinary times

The first 4 Step , We should review the past actions of the business , There are records in history

Do your homework at ordinary times , It's time to do , In fact, only the 5 In this step, we use historical data to calculate and verify the impact .

So we often say , Data analysts want to enhance data insight , You have to accumulate more analytical experience , For specific business problems , Collect business actions , Multiple duplicate plates , Only in this way can we know more and more deeply . Every time a specific problem comes , Only then can there be rich ammunition depots available .