Analysis of bidding promotion in 2B industry!

When 2b products are being sold and acquiring customers, they are affected by the high order and low conversion business model, which is often a pain that practitioners in every industry cannot let go of even though they appear to be indifferent.

Because for 2b to close a deal, the way of thinking is completely different from that for 2c business.

It requires full communication and integration with customers to create value for them. This also means that the user life cycle of 2b enterprises is destined to have several more stages than 2c: a long unconscious stage, cognition, interest, preference, purchase, and then entering a new procurement cycle, namely "consideration" and "preference".

In recent years, many people have written about "N ways to acquire customers for 2b products" and "These marketing methods make 2b customer acquisition more accurate" on the Internet.

But in fact, it's still the same old story.

As for 2b customer acquisition, I personally think that the effective customer acquisition channels ultimately come back to search engines and brand building .

Secondly, music like EDM is not very practical in the country.

The most typical example is that Oracle has developed some "high-priced enterprise marketing automation" in which EDM is placed in a very important position. Most of the companies that use it are foreign companies. As for whether it is easy to use, it is really difficult to judge.

Furthermore, 2B enterprises have been placing bid ads on Baidu, and the cost of acquiring customers has continued to rise in recent years. At the same time, many people have asked me, "Why do accounts often spend a lot of money in a day, and there are many visits, but no one inquires about the business?"

We use restaurant franchise chains as cases to study and analyze the above issues. The brand is referred to as "A Catering" below.

The first step is to locate the problem and check the Baidu bidding account.

Display: The display volume for the day was 30,000. The display region, time period, and keywords were no different from those on other dates with inquiries.

By querying the details of various words, we found that there was no problem with the keyword matching and the search terms were basically normal.

In addition, the launch areas are all cities where franchising can be carried out, so there is no problem of not being able to provide services in the launch areas.

At the same time, the results displayed show that the creative content of this education brand and other competitors are similar, so it is possible that users will switch to other brands after viewing the education brand page, but it is impossible to judge the probability, and we can only make continuous optimization and adjustments for this.

Therefore, the problem troubleshooting for this link has been basically completed.

Click link query, the number of clicks on that day was 450, and the click rate was 1.5%. Because I have little exposure to this industry, I am not sure whether this ratio is appropriate.

Let’s take a look at the inaccessible pages that cannot be opened. Through the website log, 98% of the status codes returned are 200, and the arrival rate is 89%, indicating that there are no abnormalities in the click access.

The page dwell time is 3 minutes/the bounce rate is also normal at 22%, which further proves that the matching degree of keywords and search terms is not a big problem.

One point that needs to be noted here is that the page dwell time and bounce rate are average values, so it is necessary to confirm their frequency ratios in various segments, such as 0-60s, 61-120s, etc., to avoid large values ​​affecting the final average value, which may lead to omissions in problem troubleshooting.

Malicious clicks. Here we look at an indicator, that is, the click/UV ratio is about 1.22. As long as it does not exceed 1.5, malicious clicks can be ruled out.

The next step is to check the IP addresses that access the channel. When I first came into contact with SEM about 6-7 years ago, I felt that this function was rather useless because the IP exclusion function was not very practical except for preventing colleagues in the company from clicking at random.

Using IP to locate the user's location can only give an approximate location at best, due to the inaccuracy of the IP geographic information database.

If the user uses a proxy server, the server sees the IP address of the proxy server, and the server obtains the location information of the proxy, not the location information of the user.

Even if we find that some IP addresses frequently visit but have not converted, we will judge them as invalid UVs and exclude them. This is actually an approach of rather killing by mistake than letting go, which is not rigorous enough.

Through the above content, the account side did not find the reason for the unregistered access, so we can only find the reason from the data after the user accessed.

The difficulty you may encounter here is the traditional companies I have come into contact with. Their data statistics are often too simple, so that when analyzing business problems, the lack of data statistics makes it difficult to analyze the problems smoothly.

Fortunately, when A Restaurant’s website was first created, it used a third-party data analysis platform and tracked all page behaviors.

In order to avoid randomness in the analysis results due to a small data base, we selected data from a certain month for analysis, that is, we grouped the unregistered users who visited Baidu's paid search channel on the same day .

The distribution of registration volume in future intervals of days is the number of users who have not registered on the same day. On the 2nd, 3rd, and Nth day, the number of registered users will be generated.

Because new users account for about 85% of the users who visit Baidu through paid search channels every day.

Compared with 2c, 2b business has a long unconscious stage, that is, the behavior cycle of registration consultation and transaction will be longer.

Therefore, it is necessary to determine whether the users who visit the unregistered consultation page on the same day will register/consult at some point in the future.

Given that the third-party data analysis platform is unable to implement user condition filtering such as "unregistered visit on the same day" when extracting user data during user grouping.

Then it can only be extracted from the database by technical personnel.

The final data showed that the first peak was on the 4th day with 1.8% registrations, and the peak reached 1.2% on the 7th day. After that, the number of registrations began to decrease and gradually approached 0 after the 15th day.

This indicates that the consulting cycle for this type of project is generally around 15 days.

The final registration/consultation ratio was 5% registration, and the overall click volume for that month was approximately 12,000+.

It is calculated that the number of registrations/consultations is about 600, and on average, the average number of registrations/consultations per day is about 20.

In addition, I found that a large proportion of the registration/consultation behaviors that will occur at some point in the future will be generated by entering the website through other sources.

This can actually explain the cause of the problem we mentioned at the beginning, that is, there are many visiting users coming in during the day, but there are no inquiries.

This is because some users delayed registering for consultation, and their behavior was recorded on other channels again.

The account's promotion expenses for the month were 180,000+, and the comprehensive cost of acquiring a single business opportunity was around 300 yuan.

As for the restaurant franchise business, since I have limited exposure to this industry, I am not very clear whether the cost is high or low.

But what needs more attention is that 95% of these users still have not completed the registration/consultation for various reasons.

From the perspective of user growth, this may be the driving force for business growth.

After tinkering with it for a long time, I still didn't understand how this third-party tool could filter the users I wanted, so I could only extract the qualified users through the database again.

The population that needs to be randomly selected from the database is an equal number of registered and unregistered user IDs who enter the official website through Baidu's payment channel.

Analyze the user volume distribution of the two user groups under specific behaviors (clicking on catering items, investment returns, franchise intention, downloading white papers, etc.).

Among the contents presented in the data, among the unregistered users, the click distribution ratio of modules such as working environment, recruitment positions, and career development is much higher than that of the registered user group.

It can be judged that the visiting users introduced through Baidu's paid channels include a certain number of job applicants.

In addition, corporate news, brand monthly, marketing activities and media reports are also modules that unregistered users pay more attention to than registered users.

Therefore, it can also be judged that the visiting users introduced through Baidu's paid channels also include most of the industry investigators and visiting users of competitors.

Based on the data proportion, it can be roughly calculated that among the users introduced through Baidu's paid channels, the number of job applicants, industry investigators, and competitor users accounts for about 68% of the users who visit through this channel on a daily basis.

Finally, there are demand users, accounting for about 32%.

Here we can further establish the nodes that drive business growth.

1. How to block 68% of invalid users, or how to avoid the introduction of such traffic to reduce customer acquisition costs?

2. As mentioned above, there are only 5% registered users, so the proportion of visiting users with actual business needs is about 32%. How to improve the registration consultation conversion rate?

For the first part, we observe the registration conversion rates under different characteristics of the visiting users introduced through daily omni-channels based on user behavior characteristics (data indicators: / volume), identify the characteristics of low registration rates, and establish exclusion rules based on these characteristics.

For example, by analyzing the click behavior of users who access the site through all channels, we can analyze the user characteristics, such as the registration rates of user groups with different click times for work environment, recruitment positions, career development, etc., such as 0-2 times, 3-5 times, and more than 5 times.

Screen out people with low registration consultation characteristics and exclude them from Baidu bidding and other advertising accounts.

Because it is not connected to Baidu's bidding promotion API, it is necessary to manually update and exclude these users regularly.

At the same time, without reducing the budget, the price of core words will be raised to ensure that core words are online all day and remain in the TOP3.

The purpose of doing this is to maximize the industry demand in a day and avoid large fluctuations in the cost of unregistered consultations during the day due to problems such as insufficient display time/budget, insufficient ranking, and delayed consultation.

In the above figure, after the adjustment in late March, we can see that by the end of April, the consulting costs were able to ensure a relatively stable delivery trend.

For the second part, it is necessary to observe the registration conversion rate under different characteristics based on user behavior characteristics (data indicators: / quantity). Here, the above two parts of user ID data need to be integrated.

Find out the high registration features and find out the value of improving registration based on these features.

For example, by introducing visiting users through Baidu's paid channels, we can analyze user characteristics by looking at the number of pop-up boxes, and the registration rates for different numbers of pop-up boxes, such as 0-2 times, 3-5 times, and more than 5 times.

We observed characteristic dimensions that have a significant impact on registration consultation, such as the registration rate of users who were introduced to visit through Baidu's paid channels with 3-5 pop-up windows, which is much higher than the registration rate of users who were introduced with 0-2 pop-up windows.

Then the standard for increasing user registration rate should be to pop up the window 3-5 times for users of this channel.

Based on the above analysis results, the website content strategy and operation strategy are adjusted accordingly:

- Introduce 3-5 pop-up consultation boxes from visiting users through Baidu’s paid channels;

- The industry analysis white paper was moved to the third position in the navigation, and the number of white paper downloads by users was increased to 5 times;

-….

The strategies adopted were implemented one by one, and the data were compared. It can be seen that the number of registration consultations in April has been relatively stable compared with March, and has maintained a certain growth.

The number of registration consultations for the whole month of March 2019 was 536, and the number of registration consultations for the whole month of April increased to 875, an increase of 339 registration consultations from the previous month.

Summarize:

This project is constrained by the limited data collection dimensions, single product form, and non-standard data. Although the overall effect has been improved, it is not obvious. This case is only used to serve as a starting point for the placement of single advertising channels.

Author: Seventy Two

Source: 72nd Street