When doing data analysis, keep this underlying logic in mind!

"The key to revealing the underlying logic of data analysis is to determine the standards." Why is it so difficult to determine the standards in data analysis? And how can we determine the appropriate standards for different business types?

A basic common sense of data analysis: data itself is meaningless, only data + standards have meaning. However, it is precisely the word "standard" that has killed countless data analysts. Common problems, such as:

No standards: The business person said, "My activities have improved performance. I don't know how much it will improve. Can you analyze it?" Then no matter what the data is, the business person said, "It's too little. You didn't think it through."

Uncertain standards: Business wants to improve "customer health" and "channel quality", but no matter what indicators the data analyst uses, the business will ask, "Can these indicators represent health? Then another indicator is not healthy?"

The standard jumps back and forth repeatedly: when the indicator drops by 0.1%, the business gets extremely nervous and insists on in-depth analysis; when the indicator drops by 30%, the business says “this is normal, your analysis is pointless!”

Finding standards is the core problem of data analysis. If the standards are not determined, how can we analyze the data if we cannot determine whether it is a problem or not? How big is the problem? Whose problem is it? How can we analyze the data in depth if we cannot determine it? So how do we determine the standards? Let’s talk about it systematically today.

01Why is it so difficult to set standards?

Essentially, it is difficult to set standards, mainly because the difficulty of assessing the business itself varies:

n Some businesses are like moving bricks, doing one thing at a time. This is the easiest to set standards, and piece-rate wages can be used.

Some businesses are buff-style, which improves the efficiency of doing things. This kind of standard is not easy to set, because it is difficult to separate "how much is there without buff", and countless disputes come from this.

Some businesses are icing on the cake, just like the vendors in the vegetable market shouting "My tomatoes are big and red." If you don't do this, you won't die. If you do it, it may seem better, but it is very difficult to verify the effect.

What’s worse is that some data analysts are not clear about the above differences and think they can all be mixed together.

What’s even worse is that some businesses are very clear about the above distinctions, so when they don’t do well, they deliberately try to fish in troubled waters in an attempt to blur the standards and cover up their mistakes.

When a data analyst who doesn't know how to distinguish encounters a business that attempts to get away with it, it's like a blind man riding a blind horse. This is why the complaints at the beginning appear. What's even funnier is that at this time, the ignorant data analysts haven't discovered the problem yet, and they're still asking everywhere on the Internet: "Where is the unified standard definition of Chinese Internet data analysis?"

Therefore, the solution is not to expect the business department to have a change of heart, but for the data analysts themselves to develop a keen eye, distinguish right from wrong, and see how the standards are set according to the types.

02The first category: making money by moving bricks

For example, the promotion and investment of the Internet industry, and the sales and stores of traditional enterprises. This kind of work can be evaluated on individuals, and it is clear at a glance how much revenue and new users each person contributes to the company. This kind of work is related to the company's revenue and business growth, so there are usually hard evaluation indicators, and they are usually imposed by the boss.

In this case, remember the three principles of not talking

n Don’t talk about “natural growth rate”. Just do what you need to do, and don’t argue with your boss if you’re not satisfied.

n Don’t argue about “reasonable or not”. Just follow the boss’s order. If you don’t agree, go argue with the boss*2

n Don’t talk about “other far-reaching effects”. If the boss doesn’t make a decision, ignore it. If you are not convinced, go and argue with the boss*3

What data analysts need to do is to break down the boss's goals by business line/time. Break down the annual goals into each time period and then track their completion (as shown below).

03 The second category: supply of brick moving

For example, the preparation and supply of goods. This type of work must be prepared according to sales conditions, but it cannot be completely copied from sales targets, because sales targets may not be achieved or be achieved overdue. Once the target cannot be achieved, too much stock will be accumulated, resulting in losses. If it is achieved overdue and insufficient stock is not available, some sales opportunities will be lost. Therefore, when setting goals, there is often a double assessment: supply adequacy rate/inventory loss rate.

04The third category: general buff

Popular BUFFs, such as various big promotions, new member gift packages, 100 off for purchases over 500, buy three get one free, become a platinum card member after spending 10,000, etc., are usually organized by operations, marketing, and growth departments. The rules and participants of this kind of activity are open and transparent, and users who meet the requirements can receive the prize.

Popular BUFFs all have clear goals, such as product operations, and the goals/methods are different at different stages (as shown in the figure below).

For example, for user operations, the goals/methods are different at different stages (as shown in the figure below).

Attention! The departments that add BUFFs like to talk about "natural growth", "far-reaching impact", and "extra income". Because these BUFFs are superimposed on the work of others, the departments that organize activities are afraid that they cannot show their own merits, and they want to make the natural growth rate negative and attribute all the growth to themselves. This also causes a lot of quarrels.

The way to resolve the dispute is to separate "deep-level effect analysis" from "target assessment". Achieve your own goals first, and then review the results after achieving the goals. If you haven't even achieved your goals, your overall performance is still declining, your products are still unsalable, and you can't attract users, what "other far-reaching effects" are there to talk about? It's all self-deception.

05Fourth category: precision buff

The most common form of precision buff is precision marketing. Different marketing plans are given to different users through phone calls, text messages, and push notifications within the APP, and other users are unaware and cannot participate.

Note: Precision marketing can measure the natural growth rate relatively accurately by setting activity groups and reference groups through closed information. Therefore, the precision BUFF can directly set the goal as: compared with natural growth, additionally improve the conversion/consumption effect XXX.

I am afraid that the activities I am doing are precision activities, but I don’t do grouping, reference group, or AB test beforehand. I am not prepared at all, and then someone asks me afterwards: How can I measure accurately? I definitely can’t answer.

06Category 5: The icing on the cake

All tasks that cannot be recorded by internal data are icing on the cake tasks. Common examples include brand, communication, and service, and the corresponding indicators are user awareness, user satisfaction, and user loyalty.

These data sources either rely on market research or data records from external platforms, such as the number of article readings on public accounts, the number of video views, etc.

The biggest problem with this kind of standard is that the data is easy to manipulate. The quality of market survey questionnaires is too poor, and the samples are too small; the data from external platforms can be easily made by manual inflating. So there is a simple way to set this kind of standard: let the responsible department set the number in advance, and then see if they can achieve it.

As for whether this icing on the cake can bring actual results such as sales and users, data analysis only recognizes links. If there are conversion links and internal data records, it will be converted into a general buff-style work to assess conversion efficiency. If there are no links, no conversion paths, and no internal data, they will not be recognized at all, which will cut off their thoughts of fishing in troubled waters.

07 Summary

Essentially, the standard issue involves performance appraisal, so the business side has the motivation to fish in troubled waters. This is an office politics issue disguised as a data issue. Therefore, if you want to solve the problem, you must not lead the solution to "there is a perfect mathematical algorithm". No mathematical algorithm can solve the problem of human greed.

Therefore, what data can do is to implement the measurable parts, help everyone return to their original intentions, reduce nonsense, and better achieve the overall goal.