Kill them! This is how data empowers operations

In the digital age, data has become the core driving force of business operations. However, how to transform data into actual operational results is a challenge faced by many companies. This article explores the key role of data analysis in operations and how to improve operational efficiency and effectiveness through data empowerment.

Students who do data analysis often serve operations, and they are also most afraid of operations entanglement. Because operations work is highly related to data analysis, six out of ten data analysis articles you see online are written by operations.

Operations are so deeply involved in data analysis that they often clash with data analysts over analysis ideas, methods, and conclusions.

Today we will look at the biggest problem. There are many types of operations (as shown below), among which event operations is the most strategic, most closely related to data analysis, and also the most criticized position. Let’s use it as an example today.

Please listen to the question: For a certain game APP, the user activity rate showed a slight decline in May. The activity operation team decided to hold a sign-in and clock-in activity to increase user activity (the specific increase was not specified). The data before and after the activity are shown in the figure below.

The event operation team insisted that if there were no events, the natural growth in May would have declined.

The boss thought this was bullshit.

The event operation team said: Data analysts should use artificial intelligence big data to accurately analyze the natural growth rate.

Q: You are a data analyst, what should you do?

01 Absurd things under the banner of science

First of all, what is the key point of this question?

A. Decreased user activity rate

B. Natural growth rate

C. Artificial Intelligence and Big Data

Think for a second

Let's ask a question: One day, someone comes to you with a bow and asks you, "Please use artificial intelligence big data to accurately analyze how much higher my hit rate is than natural." What will you do? Will you pick up the keyboard and start writing code? - No! You will first ask him, "What are you shooting?"

If he says: I don't know what I'm shooting, can you help me analyze it? What would you do? Would you use artificial intelligence big data to analyze what he wants to shoot? - Of course not! If you are polite, you will ask him to find the arrow he shot first; if you are impolite, you can just curse. Because even kindergarten children know: before shooting arrows, you have to set a target first.

This is common sense.

So the overall point of the question is: it doesn't say how much it will increase. Even the question itself is problematic. Please note that the source of the problem is the decrease in the number of active users. As a result, when setting goals, the operation turned it into user activity. The difference of one word changes the meaning from clear to vague.

• What is a “degree”?
• It means the number of active users. Why not use it directly?
• What does it mean? Is the formula the same for everyone?
• The meaning is comprehensive calculation, what are the calculation formula and weights?

The indicators themselves are not clear, and there is no clear indication of how much to improve, which is a disaster for post-activity analysis. It is simply a replica of the archery story. The question is: why did such a weird thing happen?

02 The difficulties behind the absurdity

If you have actually worked in a company, you will know that not all decisions are highly rational, for example:

• This is a regular activity. The operation just modified the template and launched it online.
• This was done under the personal instruction of the boss. I don't understand it and I don't dare to ask.
• This is a business intuition: I feel there is a problem! It is too late to change it by the time you really observe it.
• This is a kind of superstition: Don’t you have artificial intelligence and big data? You should have it in no time.

In short, in real enterprises, probably:

• 30% of the activity plans were not clearly written, with only a vague statement: "Increase consumption/increase activity"
• 30% of the activity plans use imprecise terms, with a lot of fanciful concepts such as activity, activity value, and activity power.
• 30% of the activity plans have not been calculated, and are decided on the spur of the moment. Some even set a target of 100 million just to make the slogan catchy.
• Only about 10% of the plans can be written clearly: increase the number of active users, and maintain DAU above XX level in May

Of course, in large companies with standardized management, this kind of chaos is much less common. However, similar problems exist in most companies. They do not write clear goals in advance and rely on big data for analysis afterwards. They even try to get away with it by artificially creating a very low or negative natural growth rate. What should we do if we really encounter such a thing?

First, we must not talk about "natural growth rate". Especially in this kind of business with intensive short-term activities. If we must talk about it, we should adopt a buy-and-leave-it mechanism: everyone should agree on the natural growth rate in advance, and then we will look at this number afterwards and not adjust it. This is the same as refusing to regret a move when playing chess.

1. Make it clear in advance what your goal is, and increase it from X% to y%;
2. Find clear, intuitive goals and avoid fantasy/complex concepts;
3. Find goals related to core KPIs and avoid miscellaneous interference.

These are the three principles for solving the problem thoroughly.

Of course, there are two challenges in doing this:

Challenge 1: Some operations just don’t know how to set goals, can you help?

Challenge 2: Some activities simply don’t have goals set in advance, how can we remedy this?

03 Basic Methods of Setting Goals

There are three basic approaches to setting goals:

1. KPI Decomposition Method
2. KPI reverse calculation method
3. KPI Scenario Method

Corresponding to three scenarios:

1. Activity goals are KPI indicators
2. Activity goals are KPI process indicators
3. Activity goals are KPI-related indicators

Some students may ask: Why are they all linked to KPI?

Answer: If what you are doing has nothing to do with KPI, then you also know the importance and urgency of this matter.

Making a big fuss about things that have nothing to do with KPI may itself be the reason that affects KPI.

KPI decomposition method example:

KPI reverse calculation example:

KPI scenario method example:

It is very necessary to maintain good communication with operations at ordinary times, so that data analysts can intervene in the early stages of planning. It can not only help operations clarify their ideas, but also help operations calculate their goals, and also prepare for post-launch monitoring and post-event review, achieving three goals at one stroke. The best state is to have a good work done beforehand and no need to quarrel afterwards. Cooperation and win-win for everyone is the best state.

04 Basic methods of post-event remediation

If you don’t set a goal beforehand and must make up for it afterwards, remember that the core is not the natural growth rate, but “what indicators the business needs to achieve.” Especially in the case at the beginning.

When the overall goal has been achieved, worrying about the natural growth rate afterwards will often turn into a blame game.

This can be done in three steps:

Step 1: Set the direction

Step 2: Find a way

Step 3: Look at the details

Through such operations, we can at least end the confused state and clarify: what kind of curve we want to make, and determine which method to change this time. In the details, we can find the optimization direction for the next iteration.

Note that this is more based on the judgment that "it has failed so far". This is not a scientific way to evaluate the effectiveness of activities. If you want to scientifically evaluate activities, you must design experiments in advance, divide them into test groups and reference groups, and test the user response effect. Again, the more preparation you make in advance, the less you will worry about afterwards.

Many students will say: Even if we do this, our company's leaders are still superstitious, the operations are still brainless, and when problems arise, they still blame others. What should we do? Even so, Teacher Chen also recommends that everyone first understand: how to do this.

In this way, when we encounter problems, at least we can judge whether it is my problem or someone else's problem. At least we can know which direction to work towards. This is also the difference between real business scenarios and scientific laboratories: you have to dance with shackles, walking a tightrope between limited data, various types of colleagues, and too high or too low expectations.

Some students may ask: Huh? Why is there only half of the fight missing? What is the other half? Answer: This is another eternal question in the post-event analysis of activities: What if you do this! Why can't the number of active users be higher? What if it is higher? If one indicator is high, and the others are low, what should I do!

This is what we often call the comprehensive evaluation problem (Teacher Chen usually refers to it as: so what question). If you are interested, we will share it later.