There are many ultimate problems in the field of data analysis. If you are dealing with marketing, operations and other departments, one of the most common problems must be called: natural growth rate!
★ How to calculate the natural growth rate
★ Why do they say it is unreasonable when I calculated it
★ Why the natural growth rate they give is so low
Many difficult and miscellaneous diseases, let’s explain them systematically today
1. What is the natural growth rate
The natural growth rate is relative to the artificial growth rate. Strictly speaking, there is no strict natural growth rate in business. All performance is made by people. However, some departments can directly produce performance, and some can only play auxiliary and stack buffs. Typical direct output of performance, such as sales, intuitively, all performance is sold by sales. Internet advertising is similar to this. Once an advertisement is cast, the user clicks it and it directly brings in revenue.
A typical overlapping buff department is branding, marketing, promotion, user operations, and event operations. They stack the buff on the basis of sales and promotion.
such as:
★ The original product sold for 30 yuan, now I will send a coupon to save 5 yuan
★ The original product is food, now add a "eat a life extension" promotion
★ The original product has no brand, now add an "international brand" and "famous trademark"
★ Doesn’t it look a bit more awesome, maybe more people buy it?
It is possible and impossible! In short, it is difficult to say clearly.
Therefore, these departments especially want to export one: the concept of natural growth rate. Stripped out "Which are the natural sales, which are my coupons/points/small gifts/gifts/advertisements/slogans/levels/honor badges". This is the origin of the natural growth rate problem.
What’s interesting is that since these people invented the natural growth rate, sales people have also started to use this concept, but the usage is: calculate the natural growth rate, and then prove that the big environment/weather/operational planning stupid activities have a negative impact. As a result, sales were not done well.
2. Natural growth in theory, like this
It seems that the BUFF that wants to spin off the brand, operation, and marketing is a good solution. Just divide the sales into natural and artificial parts!
In theory, there are three algorithms for natural growth rates.
Method 1: Distinguish by time. The natural growth rate is between no activities, and artificial growth is during the activities (as shown below)
Method 2: Distinguish by group. People who do not do activities grow naturally, those who do activities grow artificially (as shown in the figure below)
Method three: distinguish by product. Products without activities are natural growth, and activities are artificial growth (as shown below)
Fuck! How easy it is.
Yes, it is so easy in theory, but in practice, troubles come.
3. Challenges from reality
Challenge 1: Inactive time data is not neat.
In essence, customers will not even go to the store to buy things all the time. If you want to go shopping, you have to wait for get off work, weekends, and holidays; if you want to go online, you have to wait for free time, fish, and baby sleep. Therefore, the data itself during inactive periods is ups and downs. Choose a 3-month average? Choose the last month? Choose the last week? Because of this, there are often quarrels.
Challenge 2: Do activities almost every day, no inactivity period.
This is very common in the retail, e-commerce, gaming and other industries. Activities are done almost every day, and inactive periods cannot be selected. Or the inactive period is only a few weeks between two major events, and it is in a period of recovery after the end of the major promotion, which is not justified. In this way, the time law is basically abolished.
Challenge three: The attributes and life cycles of commodities are different, and there is no comparison.
First of all, it is difficult to select exactly the same products for comparison, and there are more or less differences between the two products. Second, the sales trend of the product itself is also artificial. If the goods are sold well, an order needs to be added, and the goods are sold badly and need to be cleared. Therefore, the current sales can hardly be identified as "natural".
Challenge 4: Not all activities are suitable for dividing the reference group.
For example, in 618, the Double Eleven promotion, there are not enough people who are still participating, and it is impossible to exclude some people from participating. For example, for non-e-commerce channels and non-instant consumer products, if pricing is based on different groups of people, it is easy to cause stray goods, or be reported to the Market Supervision Bureau by consumers, and be convicted of "big data".
Challenge 5: Refer to the crowd's drawing method, it is difficult to be smooth.
It is difficult to explain the problem even if it is divided into the reference group in real time. Because the final test is buying behavior, and there are many variables that affect buying behavior. Gender, age, past purchase frequency, brand loyalty, promotion sensitivity and other factors all have an impact. Therefore, through analysis and sampling, a reference group with a very low purchase rate can be easily made, thus making ABtest invalid.
Challenge 6: External influences are not considered.
Yes, even if all of the above factors are considered, some people still jump out and say: You did not consider the impact of the macro environment/weather/policies/community ethnic group, etc. In short, if it should have fallen by more than 30%, look at the XX peers. Many, so it’s normal for us to fall by 20%. Well, it’s our result.
Looking deeper, the reason why there are so many messy arguments is essentially two words: throwing the pot
It's not to make your work flourish. Who would push it so hard, with a brilliant tongue. What you said makes people's performance look bad. People will find all kinds of reasons to spray you, it's that simple.
Fourth, the theoretical break
Is there a reasonable solution? In theory!
The prerequisite for the solution is to stand on the second floor and look at the problem and move the butt away from the small department. Thinking: To what extent is really helpful to performance, how to complete your tasks and improve overall efficiency.
The first thing to eliminate is to throw the pot to external factors.
Whether it is a change in the external environment is actually very easy to identify from the data, as long as the four major conditions are met, it can be said: This is mainly affected by external influences (as shown below). However, if the four major conditions are not met, just seeing a news report and hearing a complaint from a colleague will not be the case.
Secondly, don't talk about natural growth at all for departments that directly generate performance. Reaching the standard means meeting the standard, and not meeting the standard means not meeting the standard. It just depends on whether you don't find a way or find a way from outside.
Again, for the stacked buff department, it can be regarded as natural growth. But there are three types:
Simply put, it is:
★ If you don’t memorize hard indicators, compare yourself with yourself.
★ If you memorize hard indicators but have a clear task, complete the task first.
★ If you recite hard indicators and remember the overall indicators, pay attention to the whole and don’t entangle the details. The overall standard is not up to the standard, you just beep, I did a good job, but no one believes it
★ If you have hard indicators and group marketing, you can go directly to ABtest.
Of course, there is the easiest way, which is to buy away. Everyone's consensus algorithm in advance, whether it is the last X weeks or the same period last year, in short, the consensus is good before the project starts. After that, I bought it and left the hand, but the effect was not good afterwards and I reflected on the reasons. The reference group is uncertain at the beginning, and naturally it will be sophistry afterwards.
V. Helplessness in reality
However, the above is only theory. In reality, no matter what, it is:
★ Responsible for stacking buff operations, always want to prove that they are overwhelming.
★ The sales department always likes to complain about the lack of back-end support.
★ The boss in charge of supervision always has his own small nine and nine standards.
Therefore, the quarrel about the natural growth rate will continue endlessly. In particular, sometimes the data analysis position is set under the operation, and the boss of the operation needs data analysis to help his platform. At this time, the science is unscientific, reasonable and unreasonable, so they can't manage that much. To eat the monarch’s profit, divide the emperor’s worries, just think of a way to get over. However, as a data analyst, you have to know how the game is played, so that you can move forward and backward freely when you use various methods to round out the story.
In fact, if you look closely, you will find that the so-called ultimate problem of data analysis has never been the calculation itself, but the difficulty of using data as a gun in all departments. The ass determines the brain. When you want to collect a favorable data evidence, you can always find one. The most classic example is ABtest. If you are interested, please pay attention to the public account Grounding Qi School. We will share the next article, so stay tuned.
The down-to-earth teacher Chen, master of mathematics science, business management from 985 universities, data science management expert, has 11 years of rich experience as data director and senior consultant. Served large enterprises such as Ping An Bank, China Guangfa Bank, Tencent, and created online and offline integrated digital transformation solutions for traditional enterprises such as Vinda Paper, Overseas Chinese Town, Guangqi Honda, and World Union Real Estate.
The creator of the public account [Grounding Qi Academy] has independently launched a series of data analysis courses, which has more than 20,000 students and is well received.