Data analysis from the perspective of product operation

What is Data Analytics?

Data analysis is a goal, a tool, and a way of thinking. Just like narcissus reflecting itself in water, data analysis is more like a mirror for product operations . It can reflect performance, motivate work goals, and test operational results, but most importantly, it is a working ability and a way of thinking.

If data analysis also has a life cycle, it would be “believe in data – doubt data – reshape data – reflect on data”. This is the business process of data analysis and the mental journey of data analysts. It also corresponds to the career stages of data analysts, namely rookie, entry-level, intermediate and advanced.

1. Cainiao’s Startup: Trusting Data

To do data analysis, you must first believe in the data. As a qualified product operation rookie, you must first follow it as a criterion and respect it as a benchmark before you can further explore the data. Specifically, you can start with daily, weekly, and monthly reports, familiarize yourself with the data indicators of KPIs, and radiate to related data with key data as the center.

For example, if the key data indicator of the KPI is sharing data, then with it as the center, you can focus on segmented data dimensions such as the number of sharers, number of shares, and sharing return UV. Because active data is closely related to sharing data, you can also focus on related data such as activeness and attracting new users.

However, "believing in data" is a working misunderstanding of many people, that is, taking data as data, seeing data as data, and doing data for data. Such data interpretation is superficial, staying only on the surface of things without going deep into the core essence.

The most common situation is: working around the work goal KPI data and being driven by data is like being a debtor of data, carrying a heavy burden on the long KPI road.

A slightly more advanced approach is to break down the work KPI, breaking it down from large to small and from small to micro, and carry out work through more detailed data dimensions to achieve the KPI goal.

For example, the data of "number of friends" in WeChat group can be further divided into sub-indicators such as the number of friends who start a camp, the number of friends who end a camp, the number of friends who have 1V1 contact, and the number of paying friends. Just like climbing a mountain, if you take it step by step, you will eventually reach the top. Although the process is a bit slow, a snail can also reach the pyramid.

2. Initial transformation: doubting the data

From belief to doubt, this is a very difficult process, which is no less difficult than reshaping the worldview. But this is also the necessary transformation period for entry-level personnel to become mid-level personnel, just like metamorphosis into a butterfly. Although the process is difficult, the prospects are bright. After this period of product operation, I felt a sense of sudden enlightenment.

Doubting the data mainly means doubting the authenticity and validity of the data. Are the data sources and statistical results true? Are all the final results valid? What is the effective sample size? Why does this happen?

Take the APP for example. Due to historical issues left over from the product, the fields that registered users need to fill in have changed. There are many missing data in the user detail table. These missing fields are invalid data.

For example, when we apply for a credit card, the "personal annual income" column may not be filled out completely accurately, partly because the income of the person filling it out is not fixed. In addition to the basic salary, there may be quarterly bonuses, performance bonuses, year-end bonuses and various subsidies;

On the other hand, the salary of the person filling out the form is floating, and will change with the increase of length of service or job hopping. In addition, some users fill in higher salaries out of vanity or selfishness in order to obtain a higher credit limit.

The above situations illustrate the authenticity of the data. After the sponge absorbs enough water, it will of course "gain weight".

So what is data validity? The opposite of validity is invalidity. The situations that easily generate invalid data mainly include: the data sample capacity is too small and not comprehensive enough. For example, using the survey data of 100 college students to summarize the learning status of tens of millions of college students is tantamount to seeing the leopard through the tube; the data sample does not meet the preset conditions. For example, the salary situation of the white-collar class is investigated but the blue-collar class is asked to fill out a questionnaire survey; there are also complex factors such as inappropriate research methods and incorrect analysis methods.

In addition to doubting the authenticity and validity of the data, we must not let go of any doubts, and anchor our goals and feed back to the business by getting to the bottom of the doubts.

Questions like: "Is it necessary to do this kind of data analysis?", "The results of the data analysis don't seem right?" Such data questions often require us to communicate and collaborate across departments to find the answers. Through such "data investigation", we can discover the truth behind the data and serve business growth.

3. Intermediate Leap: Reshaping Data

After things get better, you need to enclose some land and run your own business. It should be noted that self-operation does not mean self-indulgence or self-entertainment, but self-reflection. Improve the operational level of data analysis through introspection, reconstruct your own data thinking logic, and establish your own data analysis methodology.

Product operations that have reached this stage can filter and present data according to their own needs or purposes, and are fully aware of the main factors that affect certain data.

Take data filtering as an example. You can choose data of different granularities (such as daily, weekly, monthly, etc.), and you can also select key items of the data (such as gender, region, years of experience, etc.).

So how do you present the data? The most commonly used are absolute numbers vs relative numbers. For example, if you find that the absolute value is not prominent when making a work report, but the relative number is very prominent (month-on-month, year-on-year, etc.), then you can choose the relative number (proportion, ratio, etc.) as the data to be presented.

Data presents results, but how to analyze the reasons behind the results is the key. Find the truth from business scenarios, find answers from common sense and logic, and dig deep into the causes behind data results.

For example, the overall number of shares, from a business perspective, is accumulated from the number of shares of each part, which can be preliminarily summarized as "overall = part A + part B + part C +..."; in addition, based on logical common sense, active, new and offline business scenarios all have a certain impact on sharing data, so they can be considered comprehensively and summarized as a whole.

4. The peak period of high-level: reflect on the data

Entered advanced data analysis and reached the peak of his career. Product operations at this stage mainly reflect on data. In other words, it summarizes experience with a more comprehensive approach and turns previous data analysis cases into reusable and transferable knowledge assets.

Looking back, which data can be better classified, organized and cross-analyzed? Which data can feed back to the product and promote its iterative optimization? Which data can be made into automated data reports?

What data can be made into a data dashboard? What data reflects the real business situation and drives the business? …

By deeply reflecting on data issues such as these, product operations organize data experience and compile a set of data analysis methods of their own, including writing data requirements documents, promoting BI tool optimization, using data analysis tools, sorting out data dimensions, collecting and screening data, breaking down and segmenting data, analyzing data causes, and summarizing data conclusions.

The process of reflecting on data is actually all-encompassing. Business, activities, operations, brand, market and other factors are all key items. Reflection on data can be said to be a review, induction and reuse of work, internalizing past data experience into valuable personal experience savings.

Believe in data, doubt data, reshape data, and reflect on data. For product operations, this is the natural process of doing data analysis work, and it is also the career growth path from rookie, beginner, intermediate to advanced. To master the essence of data analysis and achieve transformational growth, one needs the accumulation of experience and a consistent and tenacious attitude.

The above are just my personal opinions, and I welcome your criticism and correction.

Author: Li Dudu

Source: Li Dudu