Talking about Data Products

The Internet is an industry that makes popular concepts, and Data Products is no exception. In fact, the "real" data products have long since come out, only "name" is slowly becoming popular after a few years.

I have seen a lot of articles on data products. We do not have a unified understanding of the concept of understanding there are different places, so I want to simply express my point of view, the main content is not seen in other online text of a talk.

First, what is the data product

To talk about data products, the unavoidable "cliche problem" is the definition of data products. My understanding is that data products, broadly speaking, are a form of product that can play a data value to assist users in making better decisions (or even actions). It can act as an analytical presenter and an enabler of value in the user's decision-making and actions. From this perspective, search engines, personalized recommendation engine is obviously also a data product, the product is already relatively mature because of the form, so few people are divided into the concept of data products, in addition, these products are often worn outside the data one Layer coat, so that non-professional users can not intuitively feel the existence of the data.

In addition, there is a narrow range of data products, such as the well-known data cube of Taobao Data, the CRM platform of Baidu Index, E-commerce, data decision support systems of various companies and so on. A structured classification introduced.

Second, why there is data products

People's daily business activities are the spiral process of "decision-making" and "action" and the intertwined sub-process. The decision-making in the main process means that the heart should decide what to do and what is the goal to achieve Specific implementation process, such as the user to solve the inconvenience of travel, his main decision may be "to buy a car for their own means of transport", and in the specific course of action, they will soon face "buy what car" and " What channel to buy "and other sub-decision-making issues.

All decision-making and sub-decision-making processes in action are based on "some kind of reference." The simplest reference can be its own instinct. It is better to rely on the subjective experience of "coming people." However, it is harder and harder to make decisions, The so-called experts have also been repeatedly faked; and the optimal decision-making needs to rely on "evidence", quantitative evidence of real-time data, with the popularity of mathematics, statistics, computer science, data in the decision-making optimization process is more and more valuable This is especially true in the big data era.

Decision-making process, the value of the data can be reflected by what? No more than three: a. Data itself, b. Data services, c. Data products. For example, if a user wants to know if tomorrow's weather is suitable for travel, he can look directly at tomorrow's temperature data. This is where the data itself is worthwhile; he can also consult the relevant data analyst or consultant, Provide artificial data services or solutions to determine tomorrow's weather; the third way is to use data products that solidify data, data models, and decision-making logic as much as possible into a single software system to be more automated Accurate, intelligent way to play the decision-making value of the data.

Third, the classification of data products

In a narrow sense, from the point of view of users, it can be an internal user, an external business client, an external personal client, or the like. From the initial report form (such as static report, DashBoard, ad hoc query) to multi-dimensional analysis (OLAP and other tool-type data products), customized service-oriented data products and then to intelligent data products, Enabled data products.

Reportable data products are too pale and have limited visualization capabilities. Multidimensional analytical data products are more suitable for professional data analysts than for business or operations personnel, and their use is limited. Future trends may be customized Service and intelligent data products.

The so-called customized service data products are based on the deep needs of users and construct the data model, product design and visualization plan that are most suitable for the current business pain points. Data products here act more like service providers than a common tool.

Intelligent data products will be more intelligence into the big data products, and logic combined with decision making, play a role. For example, you can have a traditional affiliate marketing system that allows you to target your users by their own rules, and you can do so in smarter data products by entering your marketing goals and parameters, such as double-eleven Maternal and child market promotions, the system can be based on historical data to calculate what kind of product should be selected, in what user base, in what form the effect will be better.

Most of the available data just tell you what is going on now or in the future. Where the pain points are, but can not give better advice or even support the implementation of a proposal. What you can do with an enabling data product is such a job that not only tells you which users are losing most of their potential, but also guides the user through the follow-up remediation process, which segments need to be stimulated by promotions and which ones Need service, which needs to provide him with exclusive VIP business, which needs a better interaction and so on.

Fourth, the particularity of the data product needs to grasp

A really good data product must first grasp a core - to find the real core needs of users, pain points. This sentence for non-data product manager is simply a nonsense. But for a data product manager, come is not so easy, have their own particularity.

The first one is the hierarchy of needs. The users of data products often have many internal users. They have different understanding and proficiency in data, data sharing and data processing. Therefore, various levels of Requirements, the outline of the broadly includes: 1) business / management needs; 2). Analysis of the demand; 3) data requirements. For example, if an e-commerce company wants to improve the order conversion efficiency of all users, this is the first type of demand. To accomplish this goal, a lot of work needs to be done and many analytical needs arise, such as analyzing product details Page out of the trend, it is the second column of demand; and specific statistics of certain data items are data needs. The most horrible thing is that the demand side has encountered the problem (the first type of demand), the wrong type of analysis (the second type of demand), and the explicit data requirements (the third type of demand) have been raised. As for the data product manager, it is a required course to guide the analysis needs from the data requirements and further involve the business / management needs in response to specific pain points.

The second particularity is the particularity of the demand side of the internal data products. The users of the data products within the enterprise are both the users and their colleagues, friends, leaders and subordinates. They themselves have a certain decision-making power over the product manager The ability to intervene, the need for product managers to balance the "ideal and reality," you understand. This situation is particularly serious for data teams that are not directly under the top decision makers.

Five, data products, the three key elements

In my opinion, there are three key elements that need to be focused on achieving a data product: 1) data, 2) decision logic, and 3) action processes.

The value of data, no doubt. It's like the blood flowing inside the whole product. With what kind of data types, data cycles, data granularity, often will decide what your data products can provide services.

The logic of decision-making is lacking in many mediocre data products. They are simple, reactive reports that show the demand side. Good data products should be able to help users think, especially in the usual user experience is the business logic of decision logic, part or all of the integration into the data products, you can visualize, dynamic and convenient process of explicit decision-making, Improve the user's decision-making efficiency.

Just staying focused on problem discovery and problem analysis is not enough. We also need to be able to solve the problem, which involves the third key element of action flow. For example, when a data product analyzes a segmented user base, it finds that activity has dropped significantly in the most recent month, whether it triggers a marketing process automatically, personalized a "process of action" based on the user's characteristics, and In all aspects of the process to play the value of the data.

Sixth, the relationship between data products and big data

I really do not want to write such explanatory texts that steal the trivial concept because big data is a concept that is brought up by everyone but barely understood by all. I am here to write what the concept of explanation is wrong, imposing search engine's "load."

Therefore, it is still back to the core value point of view: Mentioned earlier, the greatest value of data products is to help users optimize the decision-making, as well as supporting the realization of the value of decision-making. If you compare a data product to a machine, that data is like the raw material the machine is running on. "Raw materials" + "process" + "result display and application" ≈ data products.

While big data, of course, also falls into the category of data, it is like a more efficient raw material that delivers more efficient value (more perspectives, deeper, more real-time information and knowledge, especially predictive knowledge) , "Efficient raw materials" + "advanced processing" + "advanced display and application" results are also data products, of course, you can also territories called "big data products."

Another non-professional example to understand: the familiar "weather forecast" is a typical data products, its raw materials may have a long period of temperature, humidity, wind, solar intensity, UV intensity, PM2.5 value , Location information, various data collected on satellites, various professional meteorological data of other ground equipment (examples are for professionals only); a series of data collection, cleaning, analysis, Process "can get the data values ​​and probabilities (temperature, wind, rain and snow, etc.) of several core meteorological features in the next few days; and we see the weather forecast of this data product, it is the above core information integrated Together, giving the video + GIS a presentation, as well as replicating the public's "action" advice (travel advice, clothing index, car wash index, etc.) made.

There are many examples of this kind of big data combined with the greater value of data products. Imagine, if you can accurately predict the trend of a stock tomorrow, more than the effort to summarize the laws based on part of the historical information to be more valuable; if you can know what users will suddenly next month, the mother and baby products Large purchases, will save a lot of "thousands of people side" of the traditional advertising costs.

And a data product generation process, the most reasonable is exactly from the value-driven point of view. Instead of simply starting from the data or from the technology.

Finally: The six aspects of data products are written down, with concepts and examples of vernacular. The original intention was to look at the so-called "new things" from the basic common sense of life and basic logic, rather than various terminologies, Writer high above, the audience foggy feeling.