Data analysis method of experts 4: benchmark analysis

"The key to data analysis is the way to benchmark analysis." In the journey of data-driven business, benchmark analysis is like a key. It seems simple, but it actually contains many key points. How to correctly use this method to discover successful benchmarks and effectively replicate their experience? Let's explore in depth and open the door to accurate data analysis.

Many students are confused: "We always talk about data-driven business, but how can we achieve it?" To achieve data-driven, the company needs to have a work environment that respects data and standardizes management, and also needs corresponding data analysis method guidance. Simply shouting "year-on-year, month-on-month, higher, lower" every day will not drive business.

Among them, benchmark analysis is the most basic method, and it is also a typical method that is easy to understand but wrong when used. Today I will share it with you systematically.

Basic ideas

Q: How can we make our business better? People will naturally think: "Who can do it? I will learn from them!" Benchmark analysis is the application of this idea. Find a replicable successful benchmark from a bunch of businesses, and then directly copy its practices. This method sounds simple, but it is very useful in data-driven!

Let me give you a simple example: if a data analyst talks to sales about how to achieve sales performance, they will only be laughed at, "Have you ever sold a single order? How dare you teach me!" But if a data analyst summarizes and shares the secret recipes of sales masters, other sales people will stretch their necks to listen, fearing that they will miss something. This is the intuitive advantage of benchmarking.

The biggest advantage of benchmark analysis is that it breaks the barrier of "data analysts have never worked in business", uses business cases to convince the business, leverages the power of others, and thus better realizes data-driven.

The specific operation is divided into 5 steps.

Step 1: Set benchmark indicators

Setting indicators is the first step in benchmarking analysis. Generally, the indicators that others can expect to improve after copying the benchmark are used as benchmark indicators.

for example:

When evaluating the sales team, we mainly look at the sales amount and the proportion of payment collected.

When evaluating user operations, we mainly look at the activity rate and conversion rate.

When evaluating user growth, we mainly look at the number and quality of new users.

The so-called success benchmark refers to being very successful in terms of a specific indicator and can be used as a benchmark. If you want to improve the performance of your sales team, you can focus on sales amount and payment collection ratio, and other indicators can come later. This leads to the first point to pay attention to.

Point 1: Success in a specific indicator does not mean success in other indicators, nor does it mean perfection. A salesperson with good performance may have poor attendance, and an operational activity with good benefits may be very labor-intensive. If you demand perfection, you will either not be able to select a benchmark at all, or the selected benchmark will be too special to be copied.

Step 2: Benchmarking indicator stratification

Benchmark indicators can be used to stratify the target objects involved in the screening and distinguish between high, medium and low. The simplest method is the decile method. That is, the target objects are sorted according to the benchmark indicators, and each 10% is divided into 1 group, a total of 10 groups, and then the differences between the groups are observed to decide which group to use as the benchmark (as shown in the figure below).

Point 2: The benchmark does not necessarily have to be the best one! It is very likely that the best one has the best time, location, and people, which others cannot copy. Blindly pursuing the best may discourage others from learning. Therefore, the benchmark should be selected according to local conditions.

Step 3: Clean up the anomalies

The so-called abnormal situation refers to the fact that the objects distributed at the top level are not the best achieved by their own efforts. This step is the most easily overlooked step in data analysis.

It is likely to perform well because:

1. Good luck: Actually, it is not very good at ordinary times, but it just happens to be lucky.

2. Good timing: This year is just the right time for the company to be successful. If the time passes, the company will collapse.

3. Good season: It is the peak season recently, and it will be gone after this period.

4. Good resources: The investment is very large, and once the investment stops, it will be gone.

For example, the performance of a sales team often has the following situations, so you must be careful when setting benchmarks (as shown in the figure below).

The reason why we need to clear abnormal situations is that these situations make the benchmark unconvincing, so we need to analyze them clearly in advance. This is very important. Because many benchmark analyses do not analyze these abnormal situations at all, but just flatter whoever is popular recently, which naturally cannot drive business in the end. This is the third point to note.

Step 4: Extract benchmark features

The above three steps ensure that we have found a true master. After finding a true master, we can start studying the master. There are three aspects to be refined:

1: The basic characteristics of masters - differences in innate traits that cannot be changed in the short term.

Taking sales as an example, a salesperson’s: personal appearance, manners, work experience, resources, customer relationship closeness… cannot be replicated in a short period of time, and may never be made up for - some people are just born to be likable, some are just born to be handsome, and these are difficult to replicate.

If you find that the basic characteristics of experts in a certain field are highly similar, it means that only such people can succeed in this field! Then the idea of ​​subsequent replication changes: not copying behavior, but copying people, looking for people with the same basic characteristics as experts!

2. Behavioral characteristics of experts – actions that can be learned, imitated and copied.

Let’s take sales as an example, for example:

Visit customers in the morning or visit customers in the afternoon

When visiting customers, should we talk about products first or about feelings first?

Give the product a trial first or give the price discount first

…

These actions, words, and behaviors can be copied. They need to be extracted separately. If such behaviors really exist, they can be copied to others immediately to improve performance.

3: Characteristics of the targets of experts - are experts able to kill in all fields, or do they target specific targets?

Taking sales as an example, some salespeople may be able to deal with all types of customers, some are only good at dealing with bosses in traditional industries, and some are good at dealing with young people in the Internet industry.

If you find that an expert can kill in all fields, you can directly copy the person/behavior; if you find that the expert is only useful in a specific field, you need to set up several benchmarks and expand the applicable fields of the benchmarks as much as possible.

This is the fourth point to note, and it is also an important point to avoid benchmark analysis from becoming chicken soup. Many people like to follow the successful people and be a yes-man. After all, as long as they are successful, their farts will smell good. This approach of not grasping the key points will only make people superstitious about the so-called "the sun at four in the morning" and "only sleeping four hours at night" and other chicken soups, ignoring the real key points.

Step 5: Verify the replication effect

After the first four steps, a feasible solution has been output to the business:

Either copy people and expand the proportion of high-quality groups

Either copy the action and increase the number of high-quality actions

In the end, it’s up to you to see whether it’s a mule or a horse. If the replication is successful, it means it’s really effective. If it’s not, then look at how to extract the real business success factors. It should be noted that it is very likely that the business team cannot replicate because of poor execution, not because of poor data analysis. Therefore, data monitoring of the execution process should also be done well.

for example:

Cloner: How is the recruitment progress, whether the recruitment quality meets the expected requirements, and whether the work is carried out as scheduled after joining the company.

Copy actions: whether business training is in place, whether a sufficient proportion of people attend the lectures, and whether subsequent work is carried out as required.

This is the last step of benchmark analysis. Only after completing this step can we conclude: the benchmark is effective/the benchmark is invalid and other methods are needed.

summary

The reason why benchmarking analysis is “easy to understand once you hear it, but wrong once you do it” is mostly because: the data analysis work is not done carefully and the above five points are not paid attention to.

1. Some analysts don’t even define what “success” is, but start talking about experience

2. Some analysts talk about "eight experiences" and "ten advantages", which seem to be comprehensive, but in fact they are all empty talk.

3. Some analysts do not clean up anomalies, and are often challenged as “special cases!”

4. Some analysts do not refine features and always say, "We need to get higher!", but are ridiculed, "If you can do it, go ahead! Get one and let me see it."

Especially the fourth point above. Some analysts set benchmarks and just shout: "If you can get 5 new customers every day, you will have sales!" This is typical nonsense. Don't salespeople want to do it? What methods to use, what words to say, and where to get customers are the experiences that really need to be copied.

The real difficulty of benchmark analysis lies in the fourth step. It is very likely that the winner will do well in almost every aspect. When doing analysis at this time, it is easy to cause the problem of "cause inversion", that is, is it because he did this behavior that he performed well, or is it because he is good that he did this behavior? This question is a very profound question and needs to be discussed separately.