Collection is operated from scratch (Advanced-Chapter 6)

Chapter VI

At the end of this article, let's leave a little space to talk about data.

data problem has always been a headache for many operators. In fact, we said something in the introductory article, but we didn't expand on it. I don't know how far the introductory article can go, but let's talk about it first.

1

Definition of data

Data is actually a bunch of numerical values.

but these values are calculated from the user's behavior. Basic materials used to facilitate students who need to use data for research and analysis.

2

What data are there

At the end of the introductory article, we listed some core data, which I briefly summarized with a brain map, and entered the content of our section:

This brain map simply shows some operational data that may be common, but if we look carefully, we will find three data types that all operations need to have:

If you want me to simplify the above brain map, I will tell you that the data you need to obtain for operation are these three types of data:

channel data, cost data and revenue data.

channel data is used to measure channel quality and channel function, which is determined by the positioning customer group of the product itself and the characteristics of the product. In fact, we can easily push it down. If a wealth management product is put into the channel of game community, its operation effect may not be too good, but if it is replaced by lottery or gambling, it may be very effective; In the same way, if the publicity and activities of games such as Legend are put on the women's community platform, the effect can almost be ignored, and if they are replaced by a Q version of the game, the effect may be very good.

cost data and revenue data will reflect the effect of operation from different levels.

By the way, don't believe all kinds of chicken soup articles circulating on the Internet, such as "XX is highly disciplined and you don't spend money to operate". There must be costs in operation. If you think that an expert operator can do things without spending money, it is better to believe that men can get pregnant and have children. The efficiency of operation can be improved by various means such as experience, proficiency and creativity, but the cost of operation is inevitable and is generally proportional to the operation effect. A very simple truth:

two activities, one activity to send 1 iphone6s, and one activity to send 1 iphone6s, which effect will be better?

Students who are operating, please be sure to seriously evaluate the cost behind each operation.

The so-called "revenue" is not equivalent to "revenue". Getting money is revenue, getting users is revenue, and getting word of mouth is revenue.

if we know the channel, cost and revenue, which are the core data to guide our operation, we can set what data we need to obtain according to our product characteristics.

Let's take Footnote, an App—— that is very popular recently, as an example.

Footnote is popular because of a non-core function, but as such an application, what data will it pay attention to?

from the product level, it will pay attention to:

1) the daily opening number of the app

2) the use times and frequency of various functions

3) the click times of various tabs and the opening frequency of corresponding pages

from the operational level, It may pay attention to:

1) the daily number of active users of the app

2) the daily number of UGC (to distinguish between new and old users)

3) the daily number of UGC shared to social media (taking into account the number of content generated by unit users)

4) the number of newly installed machines and newly activated users brought by UGC shared

and so on.

What we need to pay attention to is that these concerned data points are not static, and they will be adjusted according to different stages of the product. If we assume that the future footnote has a profit model, then its core data will shift from content to income, and at this time, the conversion rate-related data will become important.

Similarly, we gave an example in the second chapter of this article:

A travel website initiated an old user to invite new users to join, and both old users and new users can get vouchers from 1 yuan. If the new user completes a travel order during the activity, the old user as the inviter can also get vouchers from 1 yuan, regardless of the amount.

At that time, we analyzed the activity process and sorted out the key points of the activity process. These key points are the data that need to be obtained.

The data we need is designed according to the actual needs, and there is no completely universal standard. Of course, if you do more, you will find that your data feels unconsciously improved, which is very important.

3

How to get data

There are many channels to get data, and the basic ways are to do it yourself and use external tools.

if you do it yourself, the App can choose "embedding point", log and other ways, while the Web can make records through log, log and button embedding.

external tools, many third parties will provide services.

in fact, there are various ways to obtain data, but the key point is, as an operator, what kind of data is important, and what is the context of these data? This is a linkage process, not a single behavior.

4

How to analyze the data

Everyone has different ways to interpret the data. If we want to make a simple summary, the methods of data analysis are nothing more than:

1) Determining the accuracy of data

This includes the rationality of selecting data dimensions and the accuracy of data statistics. If the data dimension selection is unreasonable and the statistical results are inaccurate, we may not be able to get the correct analysis results. This is the foundation.

2) Identify the factors that affect the data

A data will be affected by many factors, both internal and external. Operators should know as much as possible about the influencing factors at all levels, so as to help us interpret the data in a relatively correct range.

3) Pay attention to long-term data monitoring

In the analysis of operational data, ring-on-ring and year-on-year methods are often used to compare data. Simply put, the ring ratio is the comparison between today and the previous day, this month and last month, and this quarter and last quarter; Year-on-year is the comparison between this year's day and last year's day, this month and last year's month, and this quarter and last year's quarter. The ring comparison helps us to see the short-term data fluctuation, while the year-on-year comparison helps us to understand the data fluctuation in the big environment.

4) keep an objective perspective

in the process of data analysis, objectivity is very important. You should not be fond of things but sad about yourself. You should admit that you have done wrong operations and brought adverse effects, and you should be calm when you have achieved unexpected results. Don't choose conclusions that are beneficial to you. This is a problem of professional ethics, and it is also a very common problem in professional development.

5) Pay attention to eliminating interference items

In actual work, we will encounter many problems, which are interference items. For example, in a relatively stable curve, there is a sudden strong fluctuation at a point. At this time, we need to fully understand the cause of this fluctuation. If we can't confirm the reason, we will eliminate this fluctuation, otherwise it will be difficult for us to get a correct conclusion.

personally, I can't say anything more valuable about the content of the data. I write it here with a responsible attitude, hoping that students who are interested in it will study hard through professional channels and improve their level in practice.