Indicators, as a combination of business and data, are the basis of data statistics and the basis for quantifying business effects. How to avoid the confusion of indicator names and non-unique indicators in the process of using indicator statistics? At the same time to do a good job of data orderly and structured classification organization and storage, to avoid the duplication of the underlying data construction, data statistics source is not unique. This article focuses on how to define atomic indicators based on business processes, define relevant derived indicators based on the combination of atomic indicators and modifiers, and at the same time do a good job of standardizing the naming of indicators.

1. Background

Indicators are quantitative metrics that quantify business units after they are subdivided, which makes business goals describable, measurable, and disassembled. The construction of the metrics system is the most important, which will directly affect the whole process of "generation→processing→storage→calculation→application". In the actual use of the process of the following problems: ① for the same business scenarios, each department will use different indicators to measure; ② for the calculation of the same indicator, different points in time to statistics, statistics sources and methods will be different.

Specifically, the following should be done: 1) standardize the naming of dimensions and measures to ensure uniqueness, and the naming rules should be clear, common, and easy to understand as much as possible; 2) for confirming dimensions or measures, unify the calculation caliber to avoid ambiguity; 3) cover as many core dimensions and measures as possible, and use this as a basis to promote the construction of data to ensure that the dimensions covered by the indicator dictionary can be differentiated, and that the indicators are all statistically distinguishable; 4) based on the indicator dictionary, the indicators are all statistically distinguishable. Statistics; 4) based on the indicator dictionary, the core dimensions and measures into the metadata center, access to the indicator extraction tool, the subsequent realization of no need to write sql to complete the self-service query and analysis needs, to build the basis of self-service analysis of data.

Therefore, to provide standardized, *** enjoy, data service capabilities, reduce the cost of data interoperability, release computing, storage, manpower and other resources to eliminate the pain of business and technology. The data system architecture is as follows:

Business segments: If there are more business products and fewer indicators or overlaps between them, they can be separated separately.

Specification definition: design a set of data specification naming system.

Model design: based on dimensional modeling theory, construct consistent dimensions and facts based on dimensional modeling bus architecture.

Data System Architecture

2 Specification Definition of Indicators

Specification Definition of Indicators: take dimensional modeling as the theoretical basis, construct bus matrix, divide and define data domains, business processes, dimensions, metrics/atomic indicators, modifier types, modifiers, time periods, derived indicators, and so on.

Canonical Definition of Metrics

① Data Domain

Refers to a collection of business processes or dimensions that are abstracted for business analysis. Among them, the business process can be summarized as a non-split behavioral events, under the business process, you can define the indicators; dimensions, is the environment of the metrics, such as the buyer order event, the buyer is the dimension. In order to protect the vitality of the whole system, the data domain is required to be abstracted and refined, and maintained and updated over time, but not easily changed.

② business process

Refers to the business activities of the enterprise event, such as, order, payment, refund are business processes. Among them, the business process is not detachable.

③ Time period

Used to specify the time range or time point of the statistics, such as the last 30 days, the natural week, the cut-off date, and so on.

④ Modifier type

is an abstract division of modifiers. The modifier type is subordinate to a business domain, such as the log domain's access terminal type covering wireless, PC, and other modifiers.

⑤ Modifiers

It refers to the business scenario-qualified abstraction of indicators other than statistical dimensions. Modifiers belong to a modifier type, such as the modifiers wireless, PC, and so on, under the access terminal type of the log domain.

⑥ Metrics/Atomic Indicators

Atomic indicators and metrics have the same meaning, based on the behavior of a business event under the metrics, the business definition of the indicators can not be split, with a clear business meaning of the name, such as the amount of payment.

⑦ Dimension

Dimension is the environment of the metric, which is used to reflect a class of business attributes, and the collection of such attributes constitutes a dimension, which can also be called an entity object. Dimension belongs to a data domain, such as geographic dimensions (which includes countries, regions, provinces and cities, etc.), time dimensions (which includes annual, quarterly, monthly, weekly, daily and other levels of content).

⑧ Dimension attributes

Dimension attributes belong to a dimension, such as the geographic dimension inside the country name, country ID, province name and so on belong to the dimension attributes.

⑨ Derived Indicators

Derived Indicators = 1 Atomic Indicator + Multiple Modifiers (optional) + Time Period. It can be understood as the circle of the scope of the atomic indicator business statistics. For example, if the atomic indicator: payment amount, the last 1 day overseas buyers payment amount is a derived indicator.

Derived indicators are divided into the following three types:

Transactional indicators: refers to the measurement of business activities. For example, the number of new registered members, the amount of payment for an order, this type of metrics requires the maintenance of atomic metrics as well as modifiers on which derived metrics are created.

Inventory-based metrics: refers to the entity object (such as goods, members) of certain state statistics, such as the total number of goods, the total number of registered members, this type of indicators need to maintain the atomic indicators and modifiers, based on which to create derived indicators, corresponding to the time period of the "history up to the current time" in general.

Composite metrics: These are composite metrics based on transactional metrics and stock metrics. For example, the conversion rate of UV-order buyers. The rules of composite indicators:

Ratio type: create atomic indicators, such as CTR, browsing UV-order buyer number conversion rate, satisfaction rate, etc.. For example, "the last 1 day store home page CTR", the atomic indicator is "CTR", the time period is "the last 1 day", the modifier type is " Page Type" and modifier "Store Home".

Proportional: Create atomic metrics such as Percentage, Share. For example, "percentage of wireless payment amount in the last 1 day", the atomic indicator is "percentage of payment amount", the modifier type is "terminal type", the modifier is "Wireless".

Variable volume type: do not create atomic indicators, add modifiers, and create derived indicators on this basis. For example, "Change in order payment amount over the last 1 day", the atomic indicator is "order payment amount", the time period is "last 1 day", the modifier type is "Statistical method". "Statistical Method", and the modifier is "Last 1 Day Change".

Rate of change type: create an atomic indicator, e.g., "Rate of change of overseas buyers' payment amount in the last 7 days", the atomic indicator is "Rate of change of payment amount", the modifier type is "Buyers' geography", and the modifier is "Buyers' location". ", and the modifier type is "buyer geography", and the modifier is "overseas buyers".

Statistical type (mean, quartile, etc.): do not create atomic indicators, add modifiers, and create derived indicators on this basis; add modifiers under the modifier type "Statistical Method", such as per capita, daily average, industry average, etc. For example, "Natural Monthly Daily Average" is "Payment Amount Change Rate", modifier type is "Buyer Geography", modifier is "Overseas Buyers". For example, "average daily UV in natural months", the atomic indicator is "UV", the modifier type is "Statistical Method", and the modifier is "daily average". ".

3 Functional Implementation of Indicator Management

In order to realize the management of indicators, the underlying information to be considered:

Functional Design of Indicator Management

Part of the front-end interface to show:

Atomic Indicator Functional Interface

Derivative Indicator Functional Interface

4 Framework for the Use of Data

On the basis of the standardized definition and management of the indicator On the basis of standardized definition and management of indicators, we can promote the construction of the underlying fact table and dimension table to ensure that the data source of statistical data is unique and the caliber of calculation is uniform. At the same time, through the combination of indicators and modifiers, it is easy for business people to self-analyze and use the data, reduce the efficiency of data acquisition, and thus generate valuable conclusions to assist in decision-making and give full play to the value of the data.

Data Analysis Usage Framework

Summary

This article summarizes how to divide the business process based on the business, which involves the corresponding atomic indicators, and creates derived indicators based on the combination with dimensions, building into a standardized and unified data indicator system. In this way to promote the uniqueness of the source of data statistics, and access to the indicator extraction tool, the subsequent business to complete the self-service query and analysis needs, to build the basis for self-service analysis of data.

Some of the contents of this book are based on Alibaba data technology as well as the product department of the "Road to Big Data Alibaba Big Data Practice" reading notes and expand, welcome to exchange together.