In recent years, with the concept of Growth Hack, lean operation, data-based operation and so on becoming popular, the name of data product has been mentioned more and more times.

But what exactly is a data product? How can data products solve business problems? How did the hottest business concepts like Growth hacking get off the ground? How to design a data product that can meet the needs of users? This article will share these questions with you.

What is a data product?

Simply speaking, it is to take data as the main automated output of the product form. The concept of automated output is emphasized here to distinguish data research and consulting firms like Gartner from real-time Internet analytics products like GrowingIO. Obviously, their reports can also be understood as a product whose main output is data, but which does not have the characteristics of an automated output.

Once we’ve defined the concept, we can break it down and refine it. In terms of user groups, it can be divided into three categories:

  • Data products for internal use, such as self-built BI and recommendation systems;
  • Commercial data products for all enterprises, such as Google Analytics and GrowingIO;
  • Google Trends and Taobao indexes are available to all users.

In the examples above, recommendation systems can be a little confusing. In fact, like algorithms like user portrait and search sort, they are essentially a set of rating tags based on user data and the corresponding data model. Therefore, in the division of many enterprises, it also belongs to the category of data products. However, limited by personal experience, this article does not involve such products.

Why do we need data products?

Zhang Ximeng from Silicon Valley highly praised drucker’s saying: “If you can’t measure it, you can’t improve it.” This coincides with the core concept of Growth Hack — data-driven Growth.

Growth is the obsession of business operators, and the curve of practice is hidden in data products.

For example, in Facebook, the Growth Team, which reports directly to Mark Zuckerberg, has two Data teams, Data & Analysis and Infrastructure, to collect, calculate and display Data. They monitor all of Facebook’s data and continue to optimize it for performance.

How serious is Facebook about Data Driven? A VP led a team of 30 people to do the homepage revision for a year, but in the process of grayscale launching within three months, due to poor data performance, they were directly rolled back. In contrast, the domestic renren network copy that revision, follow up to now. It’s safe to say that Facebook’s data products have contributed a lot to its rapid and steady growth.

Facebook Growth Team

(Photo by Qin Chao, former Facebook engineer and Technology partner at Fengrui Capital)

How to design data products?

For product design, some set of steps are essential. After clarifying these contents, product planning at system level and product design at function level will be much clearer conceptually. We have abstracted it into five steps:

  • For what users and scenarios
  • What problem/value does it solve
  • What is the analysis of the problem
  • What kind of metrics do you need
  • How should these indicators be combined

3.1 For what users and scenarios

Any product design needs to be clear about the audience and the scenario, because different users will open your product in different situations in different ways.

  1. Different users have different values. This approach is mainly for the first category of enterprise internal products. I do not advocate job discrimination here, but from the perspective of the value that data can produce, a correct decision by the top management can save countless costs below.
  2. Different levels of users are concerned with different granularity. Always provide the next analysis of granularity and the entry point that can be refined to the finest granularity. Data analysis is all about segmentation and tracking down change.
  3. Different types of users use data in different scenarios, so design around those scenarios. For example, for Sales type customers, they are more likely to glance at data quickly on the way to meet customers, so mobile and automation are key. In design, the principle is to present key metrics through the mobile interface, without detailed analysis. And in some indicators of change can be timely notification through the mobile phone. The office data analyst, on the other hand, must provide the PC interface with more detailed analysis and comparison capabilities.

To understand your audience, you need to communicate with them over a long period of time. For example, our PM is in the habit of communicating with sales and customers every week, and every PM must take a part-time job in customer service for a period of time after entering the job. Only in this way, PM can better understand users and their usage scenarios and design better products.

3.2 What problem is solved/what value is added

This is essentially about figuring out what the product satisfies the user’s needs. Every need has value and priority.

  • First, determine what the core requirements are, and use the Demand/Want/Need method to analyze. People come to you and ask for Coke. If you don’t have coke, you can’t satisfy them. But he really only wanted to quench his thirst. All he needed was a drink.
  • Secondly, PST method can be used to analyze the value of demand. P: On the X-axis, how painful is the user? On the Y-axis, how many users are suffering; Z-axis: How much the user is willing to pay for this. The value of the requirement is multiplied.

Take an example of a funnel made with our product.

The customer initially said that we Want a funnel analysis function, but the core Demand (Want) is to improve the churn problem in the process of user use of the product. Then users from different sources and at different levels Need to be monitored and optimized in different links at different time of use. The final design is the funnel that can be compared and analyzed according to different links in different latitudes.

Funnel analysis function

3.3 What is the analysis roadmap

In fact, the above two points are still the scope of ordinary product manager, until this part of the data product manager professional courses really started. After making clear the problem, what kind of train of thought should be carried on the analysis? The following principles need to be clarified:

  • Data product managers must have data analysis skills in order to create greater data value
  • Data product design concept should be from overview to subdivision and continuous comparison
  • The overview should be brief and concise so that the user understands what is happening and the general direction of the problem. Don’t let users dive into endless details as soon as they come in
  • The segmentation should provide a rich enough dimension for easy analysis. Each breakdown must go down with metrics, and the results of all analyses must be actionable and relevant to the business
  • The data itself is not meaningful, the comparison of data is meaningful. The core of a data product is to highlight this contrast.

This part is the core of the data product manager to distinguish other product managers, but also very demanding. Both rich experience in product design and deep business understanding and data analysis skills are required.

3.4 Confirm whether the data is accurate and complete

Analysis ideas need corresponding data support, data display products needless to say, even algorithm products of user portrait, also must have enough accurate data to do support. The following two points should be paid attention to during the confirmation process:

  • Data completeness Determines in advance whether all required data has been prepared. Data is like an iceberg on the surface of the water, showing only a small part of it, its collection, cleaning and aggregation is 98% below the surface. Therefore, if the required data is not collected or not cleaned, it will increase the whole project period greatly unstable factors.
  • The accuracy of the data In the age of burial sites, this was one hell of a pit. Many times come to use, only to find that the way of burying the point has been wrong. Or find that the way the index is calculated does not exclude certain factors. This situation is more common in enterprise internal products. Because there are so many departments, if you fall in, you can’t climb out.

Therefore, if an excellent product manager wants to achieve Data Driven like Facebook, he must first achieve complete and accurate Data. Burying point is the pain point that must be solved.

3.5 What product form to choose

After the final determination of the above four steps, you can choose the corresponding product form. Common forms of data products include: mail report, visual report, early warning and prediction, decision analysis, etc., which focus on data presentation; Emphasis on algorithm class user tags, matching rules, etc. Limited space, here select visual report class to share with you:

1) Design of indicators

First of all, it is necessary to clarify what kind of product is suitable for what kind of indicators. For example, the most core of e-commerce is order conversion rate, order number and order amount, etc. For social networking sites, it is daily active users and interaction number, etc.

  • Layer by layer separation, not heavy not leak. That is, the MECE principle (Mutuallyexclusive, collectively exhaustive). If the order amount is split into the average price of the order number, the order number can also be subdivided into the number of users per capita, different users will have different orders per capita, layer by layer
  • Ensure that indicators can clearly express their meanings and provide a basis for the upper level of analysis
  • Clear indicator definition, statistical caliber and dimension

2) Presentation of indicators

To put it bluntly, the presentation of indicators is data visualization. This is extremely important for data product managers. It’s not just a UI designer’s job, because it’s about how others understand your product and use your data. Part of it is reading books on the subject, and part of it is looking at enough products to see how they are implemented. Here are some general rules to share:

  • At the same time, focus on showing no more than 7 indicators, 5 more appropriate
  • When designing the display of indicators, the primary and secondary relationship between indicators should be clear
  • Suggestions on the use of several forms of charts: trend with curve chart, proportion trend with accumulation chart, completion rate with bar chart, completion rate comparison with bar chart, multiple indicators cross action scatter chart. Choosing the right format for the right metrics is important.

The picture is from netease Cloud Classroom

Four, conclusion

Data products are so knowledgeable that we have only scratched the surface of the iceberg. A good data product manager must have a variety of skills, to understand their users, to maintain long-term effective communication with them; Identify the core needs of users, not just the surface; The most important thing is to master Data analysis skills, be able to use Data analysis tools, and always have a sense of Data Driven.

Author: Chen Xintao, Product Manager, GrowingIO