Don't be a cuckoo! These methods can save your boring [Data Daily]

Why is it difficult for some data analysts to analyze data daily? This article helps you improve your basic thinking and provides methods and suggestions for data analysts.

Daily, weekly and monthly reports are what data analysts do most often, but they are also the things they hate the most. Every time they update according to the template, it is extremely boring. Leaders usually don’t look at them, and when they do, they like to complain: "There is nothing to find!" Among these three, the most troublesome one is: daily data report. It appears most frequently, consumes the most work, and is the least useful. Basically, the daily data changes very little. If there is a big change, nine out of ten times it is because the data is wrong, or the business has a big promotion. Nothing can be analyzed at all. So how can we make the daily report more advanced and more effective? Today I will explain it systematically.

1. Basic ideas for improving the quality of daily reports

Essentially, the daily newspaper is boring because there is not much change from day to day.

For example, if there are 20 students in a class, and the monitor calls the roll every day, "19 people came to class today, and 1 was absent."

"18 people came to class today, 2 were absent"

"20 people came to class today, and 0 were absent"

After listening for a long time, I definitely don’t feel anything anymore (as shown below):

The same data, reported in a different way, will give you a better feeling (as shown below):

Because if you only look at the overall situation, it seems normal and not surprising that only one or two students skip classes.

but:

1. When a top student suddenly misses class, there may be something wrong with him, so you should pay attention to him.

2. When a poor student is absent too many times in a row and it touches the bottom line of management, he should be warned.

3. An unexpected incident occurred (it rained heavily), and everyone was absent from work. This is understandable.

4. If nothing happens and everyone is absent, it means there is a big problem of ignorance.

An isolated number cannot explain the problem. If we combine it with the classification of poor/excellent students, influential events such as rain/games, and the bottom line of management requirements, we can find the problem. Especially when people who should not be absent are absent, this kind of problem beyond medical care can attract more attention.

A dog biting a man is not news, but a man biting a dog is. If you want to attract enough attention to the information reported, you have to expose this unexpected problem so that people will not feel nothing when listening. The key here is to expose the difference between expectations and reality. The greater the contrast, the more interest will be aroused.

Therefore, simply stacking data cannot achieve this goal. Because all data is a presentation of the actual situation, but the degree of segmentation is different. If you want to create a contrast, the key is to clarify the expected value.

2. Labeling Historical Performance

Just like the attendance example above:

• I was a good student in the past, so I should come to class.
• I used to be a bad student, so skipping classes was normal

This is an expectation based on historical performance.

Similarly, in business development, the objects that are often taken care of in the daily report, such as stores, advertising channels, business teams, and user groups, can also be labeled based on historical performance. As shown in the figure below, the sales capabilities of the business team can be labeled and classified based on past labels.

After the classification is done, it is easy to have expectations: the strong ones will naturally have higher expectations, and the weak ones can just maintain their performance. When the actual performance does not meet the expectations, the problem can be identified.

3. Subjective Expectation Labeling

for example:

• I hope Student A can be a good student, so he must come to class.
• I think Student B likes to make trouble, so it doesn't matter whether he comes to class or not.
• This is based on subjective expectations.

Similarly, there are also subjective expectations in business development. Common examples include:

• Products: hot items, popular items, profitable items, matching items
• Users: Target users, non-target users
• Stores: A, B, C-level stores

With these expectations in place, comparisons can be made between similar levels to identify anomalies.

4. Sorting out the events and looking at their impact

When internal and external events occur, people have expectations, such as:

• Active behavior: I reduced the homework, so more people should come to class today.
• Seasonal factors: Final exams are approaching, so there should be more people attending class today.
• External environment: There is a heavy rain forecast today, so there will probably be few people coming to class today.

Similarly, there will be various events in business development.

• Active behavior (positive): promotions, new products, and advertising launches
• Active behavior (negative): shutting down business, removing products from shelves, system bugs
• Seasonal factors: holidays, summer and winter vacations, weekends
• External environment: weather, public opinion, and major moves by rivals

As long as we can collect these events, we can know everyone's expectations and discover problems.

5. Find the gaps in KPI

KPIs are bottom-line expectations, such as:

• No matter how bad a student is, he cannot not come to school for a day in a week. He will be punished by the school!
• Even if the weather is bad, no student can come, otherwise I will be punished by the school!

Similarly, many businesses have clear KPI limits, so when the KPI bottom line is reached, an expected value will be generated. For example, the most common sales performance. It is normal to have fluctuations in the short term, but if it is expected that such fluctuations will continue and the target will not be met by the end of the month, then that is a big problem! Therefore, the expected value can be found by predicting the trend.

The prediction method can be simple or complex. A simple one can directly use the current cumulative achievement rate as the cumulative achievement rate for the whole month to calculate the gap. A complex one can use the time series prediction method to infer the trend. In short, it is enough to expose the problem and draw attention.

Once you have expectations and combine them with actual conditions, it will be easy to expose differences and attract the attention of leaders.

6. Tips for displaying results

The above four methods can be prepared together, but there is no need to cover everything when presenting. After all, this is just a daily report. If you give too much information every day, people will be more lazy to read it and important issues will be drowned out.

The general display priority is:

First priority: major events. Short-term fluctuations in indicators are generally caused by major events, so those with event impacts are displayed first.

Second priority: KPI gap. If there is a gap in the overall KPI, it will be displayed first; secondly, pay attention to the situation where the KPI gap continues to widen and gets worse.

Third priority: Problems with subjective expectations not being met. Especially for projects/products/

Fourth: Past performance is lower than expected. This is usually used in conjunction with events and KPIs to provide a reference when interpreting “why it performed poorly”.

In order to make this system work better, you can collect information in advance at the beginning of the month, list major events of the month, and summarize several key daily data focus points. At the same time, when an emergency occurs, you can first remind everyone: "Something happened today, which is expected to affect the indicator trend" and then track it. This will allow everyone to pay full attention to the daily information and improve the experience of reading the daily report.

VII. Postscript

The above ideas for processing daily reports are essentially to prepare the answer in advance to the very tangled question of why daily indicators fluctuate. By labeling, sorting out events, and setting KPIs, the problems behind the fluctuations are fully exposed, triggering in-depth thinking.

Note that this method does not work for all leaders! This approach is actually working on data interpretation and is suitable for pragmatic leaders who focus on the value of data.

But some leaders just don't like to think deeply, but like to list a lot of data, and they want to have 5,000 rows of categories and 100 indicators in the daily report. They feel in control only when they pile up everything. If you really encounter such a person, you can just pile up data for him. If the leader asks for 100 indicators, you can add 20 more, and you can look very advanced.

Author: Down-to-earth Teacher Chen

Source: Down-to-earth Teacher Chen (ID: 773891)