The most comprehensive summary, the standard process of data analysis, save it!

The most comprehensive summary, the standard process of data analysis, save it!

This article systematically explains the standard process of data analysis, including two basic processes and six specific situations, as well as how the results of data analysis should be applied in practice. If you feel that you have done data analysis and the results you get are not satisfactory, read this article.

01 Process starting from the problem

This kind of process typically has three steps: question → data → answer . Generally, business departments will think along this path. For example, if you are a sales manager, you are most concerned about your performance, so you will first look at:

  1. Question 1: How is my team performing?
  2. Data 1: Status of compliance this month... Status of compliance this year...
  3. Answer 1: The target has been met, exceeding the target by 20%.

Of course, you may think more deeply. For example, the company currently has a performance ranking award: the top five teams in the country with the highest year-on-year growth each month can receive a bonus. It is already the 20th, and you really want to know if you have a chance to win this award.

Then, you will further analyze:

  1. Question 2: Based on the current ranking and the expected growth rate in the next 10 days, can I win the prize?
  2. Data 2: As of the 19th, year-on-year growth ranking…Each team’s expected completion status in the next 10 days.
  3. Answer 2: Based on the current industry + future growth rate, can I/can’t get the bonus?

Note! Question 2 is much more complicated than Question 1, because Question 1 only requires statistics of historical data, while Question 2 requires predicting the trend in the next 10 days.

There may be several ways to predict, such as:

  • Simply use the trend of the previous 20 days to simulate the trend of the next 10 days. (Trend extrapolation)
  • Based on the trend of the same period last year, simulate the trend of the next 10 days. (Periodic analysis)
  • Recommended results based on the expected conversion rate of the customers who are still being followed up*. (Business model)
  • Model the data based on sales personnel, sales expenses, etc., and then predict the results. (Algorithm model)

This is what we often call: complex requirements. When the requirements become more complex, the data analysis process will also become longer, mainly in the data link. The more complex the method, the more data preparation is required. So, what are the complex methods?

02 Analysis process under four levels of complexity

Complexity level 1: Understand the current situation. This is the simplest, and you can directly count historical data, such as the number of new users added this year/the cumulative sales performance as of January 3; the number of product inventory at the time point of January 3, etc. Note! Simply listing data cannot explain whether the current situation is good or bad. Data + judgment criteria are needed, such as cumulative sales performance + performance assessment criteria, so that problems can be discovered.

In this case, the data analysis process is: the business wants to understand the current situation → statistical data indicators + judgment criteria → describe the current situation.

Complexity Level 2: Cause Analysis. A typical question is, for example, a business asks: "Why did my performance not meet the target?" Note that the processing flow is different when the business has assumptions or not:

In short, if you want to go deeper, you must make assumptions about the business question. Otherwise, if the data breaks down the indicators, it is very likely that only superficial conclusions such as: "Because the number of people is small, the standard is not met. It is recommended to increase the number of people!" will be output.

Complexity level 3: Optimizing performance. A typical question is, “What should I do to achieve the best performance?” At this point, you need to complete all the questions in the previous two levels of complexity before you can reach a conclusion.

Therefore, the performance optimization process will be particularly long. Many data analysts do not know how to make suggestions for business improvement. In fact, it is because they lack the preparation of the previous steps. They do not understand the situation at all and certainly cannot make suggestions directly.

Complexity Level 4: Predicting trends. In the previous section, we have given an example of prediction. In fact, all prediction problems are very complex. At least you need to understand the current situation, know the problem points, know whether the business has plans to make improvements, and collect a lot of information before you can make a reasonable prediction.

at this time:

  • If business actions are not considered, trend extrapolation can be used directly, and the analysis process will be very short: the business wants to know the prediction results → observe past trends → fit the function according to past trends → directly extrapolate the results.
  • If the business wants to consider its own actions, such as "what will happen if I make additional investments", the process becomes complicated. The business wants to know the predicted results → observe past trends → build a model, quantify the impact of investment → substitute parameters and predict the results.

In short, the more complex the business questions are, the longer the analysis process needs to be and the more preliminary preparation is required, otherwise it will be difficult to output valuable conclusions.

03 Process starting from data

There is another situation where the business side does not actively raise requirements, but the data analyst needs to actively read the business meaning from the data and find business problems. The basic process at this time is: data → problem → answer.

However, this process often fails because many data analysts only see the data and do not understand the business situation. Therefore, they do not know how to interpret the number. For example:

  • I know the cumulative sales volume, but I don’t know the business evaluation criteria, so I can’t interpret whether it is “good/bad”.
  • I know what the sales ranking is, but I don’t know the details of the business ranking rewards, so I can’t see who has the potential to win a reward.
  • We know that sales are poor because a certain product is selling very poorly, but we do not understand the product attributes and cannot conduct an in-depth analysis.

Therefore, this article uses a long section to introduce how to conduct analysis from a business perspective, in order to remind those students who stare at reports every day to communicate more with the business and deepen their understanding of the business background/business status. Fortunately, many companies still have close communication between business and data, so the basic process of "data → question → answer" can be optimized.

  • For example: starting from the abnormal change, data → abnormal fluctuation → business communication → problem confirmation → in-depth analysis/problem end. This is generally the case when the data analyst discovers the indicator abnormal change and then confirms it with the business. If the abnormal change is initiated by the business/has long been known/has taken measures to deal with it, then don't worry about it. If it is an unexpected situation, then conduct in-depth analysis until the cause of the problem is found.
  • For example: starting from the benchmark, data → finding the benchmark → business communication → confirmation of replicability/end of the problem. This is usually based on the data, and it is found that a certain product/region/channel performs particularly well. At this time, the data can actively confirm with the business "whether it is an opportunity point" and "whether it is worth promoting". If the business already knows it, then it's over. If the business is interested, we will conduct an in-depth analysis of the benchmark replicability and promote the promotion of the benchmark.
  • For example, from the perspective of department linkage: data → correlation analysis → information sharing → problem confirmation → in-depth analysis/problem resolution. This is usually done when doing business analysis. For example, if you find that sales are declining, you remind suppliers to pay attention to backlog risks; if you find that marketing is spending a lot of money, you pay attention to cash flow; if you find that the business is planning a big event, you remind customer service/after-sales to prepare. Share information first, then see if the relevant departments have responded. If not, conduct in-depth analysis.

All the above processes are summarized in the following figure for your convenience:

Author: Down-to-earth Teacher Chen

WeChat public account: Down-to-earth Teacher Chen (ID: gh_abf29df6ada8)

<<:  Can “sense of participation” marketing help Xiaomi squeeze into the new energy vehicle “circle of friends”?

>>:  Dong Yuhui's new account premiered, earning over 150 million viewers overnight

Recommend

When WeChat’s “Search” function learned to “Ask”

This article introduces the three major new change...

Give your brand a “personality”

Brands are like people. Only people with personali...

Do products on Shopee need to be certified? How to list them?

Whether you are operating a Taobao store or a Shop...

How to learn how to sell cross-border e-commerce products? What are the skills?

You can actually open a store on a cross-border e-...

What are the cross-border e-commerce platforms in China? How to choose?

Nowadays, you need to find a suitable platform to ...

2023 Shopee Seller Deposit Payment Reminder

Starting from April 13, 2023, the platform will re...

Can Shopee generate sales through natural traffic? How to generate sales quickly?

On Shopee, a lot of traffic within stores is natur...

How to log in to Shopee Seller Self-Service Center? How to use it?

Shopee Seller Service Portal (SSP) is an integrate...