Source: Down-to-earth teacher Chen
Many students asked, "Is there a common and general way to sort out the indicator system?" Most of the common indicator systems shared on the Internet are AARRR on the Internet, but in reality, the situation is very complicated.
Of course, there is a common method, which is to sort out the indicator system based on business logic. Essentially, there are four types of data metrics for four different business logics. Let's introduce the system today.
Evaluative
"What is the quality of this product?"
"Is this event working well?"
"Is this customer base worth it?
This kind of index system that uses data indicators to evaluate the good/bad of a thing is an evaluative index system. Note that the evaluation does not necessarily use the data indicator system, such as the quality of the product, you can simply and rudely look at the number of sales. But looking at only one indicator will be problematic, such as good sales but low profits, good sales but poor reputation.
The evaluation index system is mainly to solve: the problem of deviation in the evaluation of a single standard. Therefore, when constructing such an indicator system, it is important to consider:
1. Reflect the quality from multiple indicators
2. Try not to overlap between the indicators
3. Distinguish the importance of each indicator
4. Comparability of various indicators
In the evaluative indicator system, the relationship between indicators is often juxtaposed (as shown in the figure below):
The difficulty of the evaluative indicator system is mainly to consider the comprehensiveness of the problem and the feasibility of indicator data collection. Special attention is paid here: the collectability of data metrics. When many students mention evaluation, they will casually say: NPS, user satisfaction, user intent and other indicators.
The question is:
1. How do you plan to collect these indicators?
2. How many real users can be covered by the questionnaire?
3. Can the situation be reasonably assessed with other indicators?
4. How to predict the predictor of intention?
If you don't answer these data collection questions clearly, even if the metrics sound good, they won't be able to land, and you will find challenges.
Process-based
"How are we selling?"
"How's our production progress?"
"How is our R&D doing?"
This kind of data indicators are used to represent the progress and results of a process, which is a process-based indicator system. This includes the most familiar transaction flow (funnel analysis) model. Actually, as long as it complies
1. Have a clear end point
2. There are several steps
3. There are rolling investment resources for each step
You can use a similar funnel analysis method to build a data indicator system.
Not only the transaction process or user retention, but also production, R&D, and procurement can be considered in this way. It's just that there is no attenuation at every step in these processes, so there is no funnel logic. In these processes, delivery time, quality, and cost are considered (as shown in the figure below).
The process-based indicator system is the simplest of the four categories. Because the link and end point of the process are very clear. When the goal is clear, the indicator system is easy to sort out. The most feared thing about the process-based indicator system is data collection, especially process data collection. The analysis of many toB industries cannot be done because there are too many process indicators.
Containment
The inclusive indicator system generally breaks down a large indicator into the sum of several sub-indicators/analysis dimensions. The DuPont analysis that we are familiar with is this logic, which splits a large indicator downward.
An inclusive indicator system that is generally used to diagnose problems. Because the sub-indicator + classification dimension can be specific to a certain business department and action, it can better find the source of the problem and find solutions.
However, it should be noted that the diagnostic ability of the inclusive indicator system is based on the fact that the main indicator itself can explain the problem. For example, when using the DuPont analysis, it is tacitly assumed that "profit" is the main issue.
If profit does not explain all the problems, but also considers customer experience, market share, etc., you cannot expect a set of inclusive indicator systems to solve all problems, and you need to build a separate indicator system for each problem to answer.
Impact
The reason why it is called impact type is because the operational actions are generally superimposed on the normal process, producing additional effects. Normally, there is a sales rhythm, and doing an activity will additionally stimulate sales, and there is a normal user retention curve, making a membership reward, and additionally stimulating users to retain a period of time. The impact-based data indicator system is to make the superposition and the additional two points clear.
At this point, the metrics to consider are a bit more complex
1. There needs to be a set of data indicators to reflect the normal rhythm of the business
2. There needs to be a set of data indicators to describe the superposition action itself
3. There needs to be standards to judge the performance of the business itself
4. There needs to be standards to judge the additional effect
As a result, impact-based metrics are particularly complex. For example, to make an indicator system for product promotion, it needs to be considered
1. Normal sales indicators (number of goods sold, number of buyers, amount, profit)
2. The operation of the activity (the target number of the activity, the number of people who meet the standard, the number of rewards received, and the investment of the activity itself)
3. Normal sales trend VS activity increment
There may be several algorithms for activity increment here. Like what
1. From the perspective of time, it can be distinguished between active and inactive time
2. From the perspective of commodities, it is possible to distinguish active and inactive commodities
3. From the perspective of the crowd, it is possible to distinguish the active and inactive people
But each method is carefully scrutinized, and there are some inaccuracies. Because even if there is no activity, the product has its own sales trend (organic growth rate problem), and it is difficult to find 100% similar products and people. Moreover, product promotion may also have an overdraft effect (fans and users of the product stock up in advance while it is cheap, resulting in subsequent reductions)
Therefore, when constructing an impact-based indicator system, setting reasonable judgment standards is extremely troublesome, and there are often quarrels. Of course, in principle, what meets the needs of the business is a good standard, and it is not necessary to consider every effect in detail. However, I would like to remind students to be extra cautious in this part of the work, and casually mention a few indicators, which will bring trouble to the follow-up work.
Comprehensive use of four types of indicator systems
The above four types of distinctions can help us sort out our ideas and deal with problems in our work. Because in jobs and interviews, the person who asks the question will rarely take the initiative to distinguish: "What is the scenario, what is the logic of the business?", but generally say: "How do I evaluate this XX? What indicators do I want to look at?"
At this point, it's important for data analysts to stay awake at all times
Let's take a simple example
"Analyze our products"
"Analyze the plans for the revamp of our products"
"Analyze the effect of the revamp of our products"
"Let's take a look at our product redesign campaign"
It is four completely different scenarios, and the supporting indicator system required is completely different
At the same time, you have to take a closer look at what kind of product it is
1. Transactional products: The main process is the transaction, and facilitating the transaction is the ultimate goal
2. Content-based products: user behavior is scattered, there are many forms of monetization, and there are many angles to observe
3. Tool-based products: the function is fixed, but the breadth and depth of user use are different
These are all elements that need to be considered when building an indicator system. It is precisely because of this that I recommend that students do not try to "familiarize themselves with and memorize" a few data indicators, but master the ability to sort out the data indicator system, and then they can respond to all changes with the same.