Lead: If your impression of the data model is still limited to the boss's reading, and the professionals' forcible weapon, read it when you have time, this is definitely an article you can't miss, because it will subvert your original understanding of the data model and teach you how to run through the data operation in your work!

First, let's talk about why we need to build a data model:

? 1. Quantitative display of all the status quo: Usually, we are most concerned about the final sales of the merchants, but what we want to say here is that the lost orders are far more important than the finalized orders. Why do you say that? The main reason is that the users who clinched the deal have subjectively recognized our ideas, prices, products, services and other attributes, while the lost users who have not clinched the deal may indeed be our potential customers. If a business model only knows,

? 2. Theoretical basis of all objectives: After a period of business development, all aspects of the business can basically run smoothly, and historical data is the only criterion for us to set goals. For example, the current user-transaction conversion rate is 10%, and we want to increase it by 20%. All improved strategies and methods can only be put forward by optimizing the current model, and we can't spread our thinking to other irrelevant methods;

? 3. The discussion basis of everything: This function of the data model can basically solve more than 80% of the reasons that are consistent with the argument. With or without specific data support, it is quite necessary to use a reasonable model to constrain the scope and boundaries of conference speeches. If your meeting is full of "I think", "I think" and "no one will do that", a model with data can slap the face well and then make the meeting return to normal.

Next, let's talk about commonly used data models:

? 1. Funnel model: This may be one of the most widely used models in our daily work. Its main function is to show the business status and conversion rate of each link, and to show the development of a business from the whole picture in a month-on-month or year-on-year manner. All loss-related businesses can be built with a funnel model. Common situations are AAARR (acquisition, activation, retention, income and dissemination) mode and retention mode.

? 2. Distribution model: the main function is to show the distribution of data in a certain range and integrate the data in the form of classification. Commonly used situations are: consumption distribution, commodity sales distribution, user portrait, 28 law and so on.

? 3. Path analysis: the main function is to indicate the situation of the service at each stage, and finally choose the optimal path by displaying the status quo of all nodes;

Here are some principles for building a data model:

? 1. All links or nodes need to be directly related to the result. Whether it is funnel, publishing or path analysis, all links in the process of data model construction are directly responsible for the results. For example, we are a sales company, so whether employees are willing to work overtime is not directly related to the results, so extending working hours will not be regarded as the assessment point of a data model here;

? 2. The finer the model, the better. Paying too much attention to details can easily lead to insufficient overall control or wasting a lot of manpower and material resources on links that have little impact on the results. At the beginning, the model is often only focused on large modules, and only when it is really necessary to improve the conversion rate or cardinality of a specific module does it need in-depth analysis;

? 3. The data model is not recommended to be used in small and medium-sized enterprises without system support, because the data model needs to collect data from key nodes. Without system automation, manual work is basically impossible, especially when multi-departments cooperate, but if there are only a few people, some people can be familiar with the whole business line, and they can also use Excel as a support to build it.

Finally, let's talk about how to build a business data model:

Let's take an O2O door-to-door platform as an example. After the user leaves his phone number on the platform, the telemarketer will follow up the matching requirements and finally choose the right staff to provide on-site service:

1. If sales are the goal, then we need to build a funnel model, and finally improve sales by analyzing customer acquisition channels, telephone conversations, business matching and other links. This is the operation of the four words "open source and reduce expenditure". Open source means expanding customer acquisition channels and increasing user base, and throttling means improving the conversion rate of all links and ultimately improving the development of the whole business;

2. If recommendation is the goal, then I need to establish a path analysis model to improve the recommendation rate of users by analyzing their motivations (service itself, quality, price, etc.). ), the quality of sharing channels and recommending users. As can be seen from the following figure, interest-based recommendations account for about 20% of the total, and the most conversion channel is WeChat, with more than 65,438+recommenders.

3. If the goal is to increase the unit price of customers (continue to consume other high-value services after cleaning), then it is necessary to build a distribution model, and introduce different preferential policies by analyzing the secondary consumption characteristics of different users to push users and promote their consumption. As shown below, if you want to increase the unit price, the best way is to start with the whole age group of 4 1-50. Because their spending power is obviously stronger than other age groups, this consumer group obviously can't take inducements as a reference (mainly from the perspective of marketing cost), so the construction of data model should not only consider the current model, but also relate other options to design a model that conforms to the business form;

? To sum up, all models are based on objectives and may be completed before the business is promoted. Otherwise, the data collected later will not only fail to guarantee timeliness, but also cause a lot of troubles, such as incomplete data and mixed data that cannot be cleaned up.

Summary:

This is a pit filling job. I think the office software that Xiaobai also needs to know will be completed soon. However, the first series of business models that Xiaobai still needs to know still has 1 business model introduction and two case sets unfinished, so we have to fill in this pit recently, and the next issue of our content is "Those products that died because of the wrong business model choice"!