Why do you always get dissed for your data analysis?

Data fluctuation analysis is a key task in business operations, but it is also often the most vulnerable to criticism and complaints. This article explores the common problems in data fluctuation analysis and proposes a systematic analysis method to help analysts more accurately identify and interpret data fluctuations, thereby providing strong support for corporate decision-making.

Data change analysis is the most frequently done part of our daily work, and it is also the most frequently complained about part. Not only leaders and business departments complain about it, but also the analysts themselves complain about it:

Problem 1: Overexertion. If the stock price rises by 1%, they will have to do an "in-depth analysis". If the stock price falls by 1%, they will have to do an "in-depth analysis". This makes analysts miserable. Often, before they submit a report on a 1% increase, the stock price falls by 1%. They are exhausted every day by the fluctuations in numbers.

Problem 2: Blind analysis. When analyzing changes, people only analyze a bunch of dimensions, and the final results are like: North China dropped by 7%, male users dropped by 5%, etc. They don’t know how to draw the final conclusion. Or they just leave it all to the business.

Question 3: In vain. After analyzing the abnormal changes, the data “7% decrease in North China” is missing. So what? Should the business launch an event? Should the business organize a promotion in North China? The data is too thin to support business actions.

Essentially, only a dead person's electrocardiogram is a straight line, but a normal person will have fluctuations. Only by grasping the business rules, judging the reasonable fluctuation range, and understanding the means of dealing with abnormalities can we promote reasonable decision-making. To achieve this effect, it needs to be completed in several steps.

1. Understand the basic trend

First, in actual business operations, data indicators come from business, and the business itself is affected by three factors:

1. The business rules itself, such as ice cream is popular in summer and down jackets are popular in winter
2. The impact of active investment, such as continued store openings and continued additional investment costs
3. The time when an accident happens, such as it suddenly rains on the weekend, or hot news suddenly appears

Among the three influences, unexpected events cannot be predicted in advance, but the business rules and the impact of active investment can be traced. Therefore, if you want to do a good job of abnormal analysis, you must first review the historical data, split the trend of overall performance, DAU, sales volume and other indicators, observe whether there is a natural cycle, and observe which active behaviors will bring obvious impacts. These basic data accumulations are important references for judging "what is abnormal fluctuation" (as shown in the figure below).

Note! When the company is not big enough or the business is still in the growth stage or undergoing business transformation, the seasonality may be masked by the impact of active investment. For example, although our company is engaged in down jacket business, we have only opened 10 stores in total. If one of the stores has a bad location and poor business, it will affect 10% of the total performance.

In this case, it is important to distinguish between business actions and focus on the impact of business actions rather than natural laws.

2. Differentiate business actions

Note! Not all business actions will have a huge impact on overall performance, overall DAU, and overall sales. Therefore, it is necessary to distinguish the key business actions that affect the indicators.

Common key business actions include:

• Channel: Continue to open offline stores and continue to invest in online customer acquisition
• Products: Hot new products are launched, and the prices of existing products are greatly reduced
• Users: Subsidy for users, special subsidies for VIP users
• Product: Major APP version update, important function launch

These key business actions are often accompanied by a large amount of marketing expenses/R&D expenses, which have an impact on all users and the effect is obvious. At this time, it is recommended to take a small notebook to record as shown in the following figure:

1. How many key actions occurred?
2. What changes have occurred in total performance indicators/DAU indicators during the period of key actions?
3. When internal and external factors occur simultaneously, how much effect can the internal factors have?

Keeping more records in this way will facilitate subsequent evaluation of the effects and prediction of live broadcast trends.

In contrast, some small-scale promotions/small-scale store openings/small-scale product updates may have a small impact or insufficient investment and will not produce a huge effect. In this case, you should first record the impact of these small-scale promotions/small-scale store openings and closures/small-scale product updates (for example, 5% of total users, 3% of stores), and then evaluate the effect.

Theoretically, the maximum impact of these small activities on the overall fluctuations is only a few percent.

3. Prioritize

By understanding the underlying trend, we can free ourselves from the daily 1% or 2% fluctuations and focus on real issues, such as:

• Continuous shock decline
• Breaking the original seasonal pattern
• Business actions did not meet expectations

As shown in the figure below, these are signs of real problems rather than daily 1% fluctuations and should be paid more attention to.

After capturing the real problem, do not blindly disassemble it in various dimensions, but focus on: "Are there any problems with the key actions that affect the business?" For example, to diagnose the fluctuation of performance indicators, first ask: What is the key action that drives our company's performance this year? If our company is a toC business, and the main means is to promote new products, then we should first focus on the performance of new products (as shown in the figure below).

The same is true for diagnosing fluctuations in performance indicators. If our company is a toB business, and the key action to drive performance this year is to "stabilize the share of major customers," then the logic of diagnosis will be different (as shown in the figure below).

What is the really serious problem?

The key actions that our company takes to promote business development are no longer effective! Moreover, it is omni-channel/all customer types/multiple attempts have failed/the investment-to-production ratio continues to decline. This is the real key problem.

On the contrary, if the abnormal fluctuation of indicators only comes from: an emergency/an activity that was not carried out well/a promotion that was not effective, or a bug in a version update, it means that the problem can be solved with a slight adjustment later, and this is a "small problem".

4. Provide implementation suggestions

After the above three-step diagnosis, it will be much easier to provide executable suggestions. If it is a "small problem" that occurs locally, it means that you can treat the symptoms and solve them in a targeted manner, such as:

• Sudden BUG → Fix the BUG and it will be fixed naturally
• Poor performance in a certain region → Order rectification and copy the benchmark practices
• A channel is not doing well → Adjust the approach and move some resources to a good channel
• Rain in the short term will affect sales → If there are still resources, then we can promote sales

However, if you find that the problem is a "big problem" of key business action failure, then it will be difficult. The failure of key business actions means that some fundamental changes must be made to reverse the situation, but it is not easy to make fundamental changes in a short period of time!

Many companies will continue to struggle with the old methods for a while until the indicators are completely ruined, and then they are forced to think of new methods. Before that, the business indicator curve faced by data analysts will be a downward trend with no end in sight. At this time, it is impossible to find a solution based on data alone. You can't go on the old path and expect a new result.

However, at least we can improve our analytical ability. At least when someone asks us in an interview: How do you analyze indicator changes? We no longer have to shake our heads and say "Indicator changes are nothing more than disassembly..."

Instead, it is systematically stated:

• We can identify regular business trends and avoid misjudgments and nervousness
• We can identify key business actions and predict the trend of indicators
• We can distinguish whether the abnormality is caused by the failure of key actions and remind the business to adjust the countermeasures.

This way we can find better jobs ourselves and get away from companies that only follow the rules.