Detailed explanation of how the data indicator system from design to landing

Introduction: Almost all of the data analysis work will mention a word - "establish a data indicator system", although this word is not strange to everyone, but what exactly are the data indicators and how to build a specific, many people are still A head of confusion.

01? Overview of data metrics

Before understanding what data metrics are, let's think about this: why do metrics exist? What problem is it designed to solve?

The development of human beings and science is advancing with time. In the early days, in order to make the experiments and results of the natural sciences more uniform and convenient for standardized measurements, some standardized professional indicators came into being. With the development of human society, the social sciences also need more and more statistics to measure things, a series of statistical indicators have been gradually produced. With the development of new information technology, data indicators are gradually recognized by the public as a way to measure goals.

From a social science perspective, indicators are the domain of statistics and are used for descriptive statistics of data. Indicators are a synthesis of concepts and their numerical values that describe the quantitative characteristics of the whole, so they are also called composite indicators. In the actual statistical work and statistical theory research, often directly explain the overall quantitative characteristics of the concept called indicators.

1. What are data indicators?

Data metrics are different from statistical indicators in the traditional sense, it is a summary result obtained by analyzing the data, and it is the metric value after the business unit is refined and quantified, which makes the business objectives describable, measurable, and detachable. Data metrics need to further abstract business requirements, collect data through buried points, design a set of calculation rules, and present them through BI and data visualization, and ultimately be able to explain user behavior changes and business changes. Commonly used data indicators are PV, UV and so on.

The indicators described in this article are ways to measure goals, and the indicators consist of dimensions, aggregation methods, and measures, as shown in the following figure.

Among them, the dimension refers to the angle from which to measure, is the perspective and direction of things, decided to measure the indicators according to different angles. Aggregate way refers to which methods are used to measure, is the way to statistically summarize the data. And the measure is mainly to clarify what the specific goal of things, is the determination of a physical quantity, but also used to clarify the unit of measurement of data.

For example, total playback time is the sum of the hours (in minutes) that a user plays audio over a period of time. Following the above disassembly, the dimension is the period of time filtered, the aggregation is calculated as the sum of the durations, and the measure is the uniform unit-minutes.

Here, we can understand that the metrics are composed of these dimensions, which is equivalent to the English word formation, where prefixes, suffixes, etc. **** together to form a word.

2. What is an indicator system?

The essence of systematization is to systematically organize the data indicators, specifically according to the business model, according to the standard classification of indicators of different attributes and stratification. Of course, different stages of business and different types of business will have different stages of the division of standards.

The data indicator system contains very rich statistics, from a macro point of view, it is a relatively comprehensive organic whole; from a micro point of view, each data indicator has its own specific meaning, reflecting the objective facts of a particular detail. Different data indicators have different definitions and different logics, and these various statistical quantities **** the same constitute the data indicator system, making it indelible value.

In general, the data indicator system is a summary of the systematization of business indicators, used to clarify the caliber of the indicators, dimensions, the logic of the indicators to take the number of information, and can quickly access to the relevant information of the indicators.

02? Principles for building a data metrics system

1, build a metrics system to have a focus

Can't just list the metrics, which is a common problem many data analysts make, come up with a large number of metrics listed first, and don't state the priority, which to look at first, and then look at which, and the business is simply unintelligible.

2, build the indicator system to have a goal

Many people are used to listing indicators, own a set of indicators split routine, regardless of what we want to solve the business problem, anyway, according to the time, channel, region and other latitude split, split to split there is no specific standard, the last but also entangled in the end, how much change in the indicators is the problem.

3, the indicator system is not the more all the better, and the business is the most relevant is the best

This previous indicator system article repeatedly emphasized the writing of the article will be in order to attract attention, the title is written XXX industry indicator system Daquan, although to organize the indicator system to everyone as much as possible to summarize the multiple business scenarios, the indicators listed in detail, but the different companies, business complexity is not the same, there is no set of indicators, the system is not a system of indicators, and the indicators will not be used. The complexity of different companies, business is not the same, there is not a set of indicators system is able to generalize, only the most relevant and business is the best to use.

03 ? How to design and implement a metrics system?

The construction of the indicator system is divided into two major steps: the design of the indicator system and the implementation of the indicator system, which can be split into some small steps, we first look at the indicator system from the design of the landing of the overall step-by-step chart, the following and then according to the chart subdivided into how to dismantle the landing of each of these steps.

1. How to design an indicator system?

1) Demand sources

The main demand sources change with the product life cycle. Build data indicators according to the current state of the data is divided into three stages after the middle school. The first thing to be clear is that there is a target program first and then data indicators, rather than creating some indicator system out of thin air and then set to the product.

At the beginning of the data indicators to build the product strategic objectives of the main priority to build a full range of Polaris indicators to monitor the indicators; medium-term business-driven to build indicators to measure the existing business, business-driven direct access to the indicators is generally a secondary indicator that needs to be integrated into the indicator model; to the late stage, this time, the data indicators have been built more or less, it's time to check the model according to the defects To fill in the gaps, build a closed loop of indicators for the product, through the data to reverse push the product iterative optimization.

2) Determine the first level of indicators

The first level of indicators is actually a reflection of the product in various important aspects of the operation situation, the operation of the user as an assembly line, around the user life cycle can be tapped into some important first level of indicators and naturally form a closed loop.

Among the many indicator models, the AARRR model summarizes the user lifecycle very well, but the shortcoming is that it omits the part of user churn. I personally feel that the AARRR model summarizes the user lifecycle in a more complete way, i.e., Acquisition, Activation, Retention, Revenue, Referral, and Recall.

Around these six aspects, you can expand the following level of indicators (just to give some examples of general indicators, the specific level of indicators can be defined according to the specific business):

3) to get the second level of indicators

The second level of indicators from the first level of indicators derived from the first level of indicators, in order to achieve the first level of indicators, the enterprise will take a number of strategies, and second level of indicators are usually associated with these strategies. It can be simply understood as the realization of the first level of indicators, used to replace the problem of locating the first level of indicators.

The role of the secondary indicators is to implement the rise and fall of the first level of indicators to the specific business sector or the responsible person, through the composition of the disassembly we can get from the first level of indicators corresponding to the second level of indicators. For example, the revenue of this first-level indicators, through the composition of the demolition can be divided into advertising revenue and internal purchase revenue.

4) Get the third-level indicators

Through the analysis of the second-level indicators, we can find the responsible party for the corresponding problems, and the role of the third-level indicators is to guide the responsible party to locate the specific problem, and then repair the problem.

By dismantling the path of the secondary indicators, we can get the tertiary indicators, and the frontline staff can quickly make corresponding actions through the specific performance of the tertiary indicators, so the requirement of the tertiary indicators is to cover the key actions on each critical path as much as possible.

Here we continue to take the example of the internal purchase revenue indicator, through the path decomposition, the final key behavioral path that leads to the internal purchase is: browse the product, add to the shopping cart, submit the order, pay successfully.

According to the above process, we constantly check and fill in the gaps to determine the first level of indicators and gradually dismantle them, so that we can build a set of effective data indicator system.

2. How to implement the indicator system?

Finally, it's time to get started, and after the goal is achieved, the next step is to bury the planned indicators in the ground.

The landing indicator is not like the design of the indicator first focus on the first level of indicators, but should first focus on the second level of indicators, because the first level of indicators is composed of the second level of indicators, the second level of indicators buried point after the first level of indicators can be calculated naturally.

Buried point is not a person's thing, need to cooperate with all departments, the following figure is buried point of the whole design to the landing process:

I do not know after reading this picture there is a doubt, the responsible party why still need to understand familiar with the needs of the demand-side is not to give the indicators, according to go to the buried point on the good ah. If you think so, then you are destined to only do a tool man.

First of all, the indicators are closely related to the specific business logic design. If you do not familiarize yourself with the business, it is not possible to refine the design of the buried point for the indicators in multiple dimensions, and the final design of the buried point program is bound to be missing three or four loopholes.

Furthermore, the indicators given by the demand side are not necessarily comprehensive, and the demand side often has a poor sense of data, and is not able to gain insight into the details of the current business that can be analyzed by the data.

So this requires the data product manager familiar with the business to understand the product to understand the user, in order to hit the nail on the head to design a set of guiding significance of the buried program, rather than according to the book to draw gourd to get out of some ice-cold data to look at on the good to remember that every buried is a deep meaning, the data is also a soul.

Clearly buried points of the workflow, the next thing to determine is the choice of self-research data portal or the use of third-party tools, such as: God's curator, Growing IO, Zhuge IO and so on. The two have the following main differences:

Self-research workload, build a long cycle, the third party to provide ready-made models, build a short cycle.

Self-research is more flexible, relative to the buried implementation of the data reported by the more friendly, without too much unnecessary logic records, in the latter part of the calculation of the indicators can be arbitrary, such as some of the time-consuming as long as the hit point, the self-study can be calculated by the time difference between the two events time-consuming, while some third-party does not support.

In short, self-research is a pain in the ass in the early stages, and third-party is a pain in the ass in the early stages. From the realization of the difficulty of self-research requires far more manpower and resources than third-party services, the vast majority of small and medium-sized companies will choose third-party services,

the following buried point of introduction based on third-party services to explain the way.

The old rules, before explaining the overall flow chart:

1) Buried specification document

As mentioned earlier, the construction of the indicator system requires the cooperation of various departments, a buried specification document can standardize the workflow to improve efficiency, but also clear demand specification to reduce the cost of communication to avoid the understanding of bias. The buried specification document includes workflow specification, naming specification, requirement document specification, etc. These should be stipulated at the beginning of the indicator system.

Of course, due to the lack of experience at the beginning and some of the problems in the subsequent work will be exposed, the initial version of the specification document may not be so detailed, but the general framework is still to have, and then add some of the minutiae.

2) Get the requirements prototype

That is, the product functional prototype or activity prototype.

3) Define the page, element names

After getting the requirements prototype, first of all, the page inside the prototype and the page in the name of the elements defined in advance, in order to follow up on the use of unity to avoid inconsistent naming of pages in different indicators.

If it is a page, it is recommended to name all of them, the elements inside the page may be a little bit more, you can pick some important elements on the critical path to name, and other elements depending on the follow-up work required to carry out the buried points (of course, if you have the energy to name all the monitoring is better, after all, the data is better, to avoid the subsequent need to use the data to find out that there is no buried points of the situation).

4) Define the event name

Why standardize the event name? Let me give you an example, one day you want to check the user path, when you use the user path analysis found that there are a large number of display events interspersed in the user behavior events, this time you are not very annoying.

If you had standardized the naming of events when you buried them, you could have easily filtered out all the events that had nothing to do with user behavior by filtering out events prefixed with the name of the event in the filter conditions.

In addition to the above benefits of event naming, there is another benefit is that it is convenient for the demand side of the use, the user can easily know the specific meaning of the event through the event name to improve the efficiency of the use of event naming can be composed of the following parts: behavior, object, result, type.

Behavior: the specific behavior of the event, there are mainly 4 categories:

Click - click a button or element of a class of events.

Go - a type of event that enters a page or function.

Show - An event that displays a page or element.

Exit - An event that exits a page or feature.

The event behavior is mandatory, and additional behaviors can be added later.

Objects: The event behavior can correspond to a page or a feature, and the event object must be filled in.

Result: The final result of the behavior on the object, there are 3 main categories:

Success - The result of the behavior on the object is success.

Failure - The result of the behavior performed on the object is failure.

Result - The result of the behavior performed on the object is either success or failure, when the specific result is stored in the event's dimension, the event result must be filled in.

Type: This parameter is an expansion parameter, such as displaying the event may display a page, or may display a pop-up window, this time in the event of a page after the suffix or pop-up window suffix, the subsequent use of the event can be very easy to distinguish between the specific types of events. The event type is optional, depending on the situation.

The above is the naming standard for events, and you can name some of them from this standard as follows: register_indicator_success, enter_recharge_page_success, and so on.

5) combing indicator dimensions

This is the time to grandly introduce the new 4W1H analysis method mentioned in the previous "indicator system construction flow chart". Why called the new 4W1H, because the traditional 4W1H for the new interpretation, in the new interpretation can be more reasonable plus I summarized in the actual work of experience.

According to the usual summary of the buried points, the event dimension is mainly composed of the subject and the event cause and effect of several large dimensions. The subject is the user, device and application, and the cause and effect is the source and result of the event. By adding the cause and effect dimension it is easy to see where an event comes from and where it goes.

Let's use a diagram to understand how the new 4W1H analysis defines dimensions:

Who: the subject that triggered the event, is the only distinction between the user's logo, if the user is logged in, then use the user ID (device ID also needs to be recorded), not logged in, then use the device ID.

When: When the event occurs, use the UNIX timestamp is fine.

What: Specific information about the subject that triggered the event, usually three subjects, the user, the app, and the device. If we use third-party services, except for the user information, we need to set up a buried point, the other third-party SDK will automatically collect, so this part of the parameter is not the focus of our work.

Where: The physical location where the event occurred, which can be determined by GPS, LBS, or IP, depending on the user's authorization. Location information is also automatically captured by third-party SDKs.

How: the specific description of the event, this piece is the focus of our work, the lack of experience will often miss some important dimensions, resulting in subsequent analysis of the support is not. According to the personal summary of the causal analysis method can be divided into the description of the event of the source and the result of the description, the source of the event to go there are only two categories: multiple acts caused by the same result, an act caused by different results.

For example, if you enter the top-up page, you may come in from different entrances; and if you click the top-up button, you may top-up successfully or fail to top-up.

The result of the event is a description of the specific information about the event. Through causal analysis to enter the recharge page to recharge the success of this series of behavior we can do the following events buried (the following event dimensions are only listed in the causal analysis of the relevant dimensions, other parameters depending on the specific business free to increase).

By burying points in this way, we can clearly know the distribution of the various entrances to the recharge page, and also know the distribution of recharge successes and failures after clicking the recharge button.

6) Define the timing of reporting

The timing of reporting events is specifically determined by the definition of the event. There are mainly the following three categories:

Show: show when the report, you need to specify whether the repeat display repeat report, such as the kind of automatic rotating banner does not need to repeat the repeat display repeat report, because such a repeat report is not meaningful, while the user repeatedly slide the repeat display can be repeated report;

Click: click on the report, this is the simplest!

Interface: This involves interaction with the back-end interface, such as the previous example of the purchase_coin_results event, the timing of the report is reported when the recharge succeeds or fails, i.e., the client gets the specific results returned by the back-end report.

7) Output data requirements document

When the above work has been done, you can output the requirements document, which mainly contains the following information:

8) Enter the indicator dictionary

After the buried indicators on the line, in order to facilitate the use of the business side, the indicators can be divided into different topics according to the business to facilitate the user to quickly find the indicators that are needed.

04? Data indicator system building methods and experience

How can we build an effective indicator system, I would like to share the following experience:

1, master the basic thinking model, a comprehensive understanding of the business

Data analysis can not be separated from the business, to understand the business is the prerequisite for us to build the indicator system

The data analysis of the business, to understand the business is the prerequisite for us to build the indicator system, to master some of the basic thinking model, can be used to understand the business, to understand the business is the prerequisite for us to build the indicator system.

Mastering some of the basic thinking models can help us quickly and comprehensively insight into the business.

1) 5W2H model

The classic data analysis thinking model. Ask questions with five English words starting with W and two English words starting with H, and discover clues to solve the problem from the answers, i.e., why, what, who, when, where, how, how to do, and what price (How much).

5w2h can help us develop a rigorous and comprehensive thinking mode, so that the process of analysis is more comprehensive and more organized, and will not produce confusion and omissions, when you feel that your indicator system has been perfect, you can use this model to help you Ken find the loopholes in thinking.

2) Logic Tree Approach and MECE Principles

The Logic Tree Approach can help us break down complex business problems into multiple simple problems, which can help us break down more granular data metrics.

The Mece principle means "mutually independent, completely exhaustive", and one of the important criteria for us to build an indicator system is not to repeat and not to omit, and the use of the mece principle can help us to grasp the core indicators and improve the effectiveness of the indicator system.

3) Business Canvas

Business Canvas is a tool to analyze the value of the enterprise, through the standardization of elements in the business model, to guide our thinking, the business knowledge material archives, in the process of understanding the business, we can follow the following chart to improve the filling, from multiple perspectives, a comprehensive insight into the business

In addition to the above thinking model, the best way to understand the business is to communicate more with the business side, to recognize the key issues of the current business, after all, the establishment of a perfect system of indicators system takes a long time, it is best to start from some of the key points, first solve the problem.

2, indicator system construction methodology

Corresponding to the business scenarios of the indicator system has a corresponding methodology, such as the indicator system based on the user life cycle of AARRR, customer satisfaction indicator system, etc., briefly share a few:

1) the first key indicators

This concept is what I saw in the "Lean Data Analytics", refers to the current stage of the immense importance of the first indicator, which is the most important. This concept is what I saw in Lean Data Analytics, which refers to the immensely important first metrics at the current stage, and also points out that you should and only focus on one important metric at any point in time during the startup phase. This concept is also instructive as we build our data analytics metrics system.

First seize the "first key indicator" of the current stage of the company, and then disassemble this indicator to each department to form the "first key indicator" of each department, that is, OKR, or KPI, and then according to the business of each department, based on the first key indicator, the first key indicator is the first key indicator of each department, and then according to the business of each department, based on the first key indicator of each department, the first key indicator of each department is the first key indicator. Then according to the business of each department, based on the first key indicators to think about what should be concerned about the refinement of the indicators.

2) Indicator system based on the user life cycle: AARRR

3) Customer satisfaction indicator system: RATER index model

In short, the construction of the indicator system can be imitated and then optimized, focusing on solving the business problem, I organized a number of specific business scenarios of the indicator system can be imitated to apply, and then adjusted according to the business model. The first thing you need to do is to make sure that you have the right tools for the job, and that you have the right tools for the job.

05?The value point of the data indicator system

What is the value point of building an indicator system for data analysis? What is the use? Most people may not be able to say. In my opinion, the value of building an indicator system has 3 main points:

1, the establishment of business quantitative measurement standards

The indicator system can establish business quantitative measurement standards, the purpose of data analysis is to illustrate, measure, and predict the development of the business.

Let's say that the measurement of a store business situation, a store monthly net profit of 200,000 yuan, just look at this indicator to feel that the store is quite profitable, the development should be good

But then look at the net profit of the first two months, and found that the net profit of the first two months are more than 400,000 yuan, the increase in this indicator, we have found that the store's business situation may be problematic.

In the process of measuring the business operating conditions, a single data indicator measurement is likely to be one-sided, and need to be supplemented by other indicators to make our judgment more accurate. Therefore, build a systematic indicator system in order to comprehensively measure business development and promote orderly business growth.

2, to reduce duplication of work, improve the efficiency of analysis

With the indicator system, data analysts can do less work on the temporary mention of the number of jobs, the indicator system should be able to cover most of the temporary data analysis needs after the establishment of the indicator system, if the indicator system is built, or there are a lot of temporary analysis of the demand for the emergence of the indicator system, which proves that this indicator system is problematic.

3, to help quickly locate the problem

The establishment of a systematic indicator system, with process and results indicators, indicators of the correlation between the front and back of the relationship can be backtracked and drill down to quickly find the reasons for fluctuations in key indicators, the boss allows you to analyze the reasons, no longer need to worry about the face.

However, the premise of these values is to establish a reasonable and effective indicator system, and data quality is guaranteed, data quality can not be guaranteed, the indicator system is good, the results of the analysis is not meaningful.

References:

7000 words detailing how the data indicator system from design to landing /s/13BoA0lOqYyFF7KNsb_RRQ