How to use zero-threshold oCPC delivery?

Baidu's zero-threshold oCPC has been online for some time. The original low threshold is gone. In fact, the low threshold is still more humane, especially for those small advertisers.

Today's zero threshold seems to be more humane, but in fact the effect is not as good as the low threshold. Although both have data accumulation, the low threshold is based on the data accumulation of the model after you build the package. After entering the second stage, the effect will be relatively obvious. The zero threshold is directly based on the accumulation of data based on your account historical data for 7 days. After 7 days, if you pass the learning period, it is considered a package that has run the model.

1. Scope of construction

Generally, you just choose "Mobile" and then select the entire account. There is no need to separate it into computer and mobile, because generally speaking the quantity is not enough. Unless your quantity is very large, a few plans or computers and mobile alone can get you through the learning period. Otherwise, just select all.

Originally, I also built packages for computers, mobile phones, and large-data plans, but they all failed. Anyone who is interested can give it a try. However, it should be noted that the consumption of each attempt is not low, because when the system is exploring, the CPC will often exceed your expectations. If you cannot pass the learning period, you can try it if you can accept this part of the cost.

2. Goal conversion settings

If you are connecting to an online consulting software, it is recommended to set up "one-sentence consultation". The author has also tried three sentences and leaving clues. One time, "leaving clues" passed the learning period, but the CPA was twice as high as expected, so I gave up in the end.

Because the CPA is relatively high when the package is created, the system will make a prediction based on the account's historical data. The CPA must be higher than 80% of the system's recommended bid, so in the end it will not deviate too much from this cost. Now the system is relatively accurate. The system will run according to the CPA you set. It will not be too much higher or lower. If it is too high, the learning will fail, and it is impossible to be too low. The current oCPC is no longer the oCPC of the past, so it is still recommended to build a package with "one-sentence consultation". The chance of success is still relatively high. Judging from the running data, leaving a message when the customer service is not online is also considered a successful one-sentence consultation, and there will not be a lot of spam traffic.

3. Target conversion bidding settings

This system requires that the bid cannot be lower than 80% of the system's recommended bid. The suggestion is to increase it by a little bit more than 80%. If the system's recommended bid is lower than your expected bid, you can try to increase the space within your expected bid range. This can obtain more traffic and give the system more room to expand. This way, the model has a greater chance of success.

If the system bid of 80% is far higher than your expectations, then you can use CPC to stabilize the account. The system bid is dynamic. Check back in a few days and build a package when it is relatively stable.

1. Optimization strategy before package building

The system has become more and more intelligent, and the control over data has become more and more precise, so it is still necessary to optimize and stabilize the account as much as possible before building a package.

First of all, expand the account as much as possible. For example, add as many keywords as possible, optimize the creativity as much as possible to better grasp the user's pain points, and improve other basic settings. In short, before building a package, do as much as possible to improve it, and don't wait until the package is built to make modifications.

Secondly, after completing the above basic operations, you need to run the data for a period of time, such as 7 days or 10 days. The specific situation needs to be based on the account situation. It is best to run until the data is basically stable.

2. Optimize strategies after package creation and before the end of the learning period

During this period, the system runs based on historical data and the data of the package you created, so it is best not to touch the account at this stage and keep your hands to yourself.

If the account performs poorly for three consecutive days, for example, the CPC is very high but the conversion is very low, then you can adjust the CPA appropriately, or delete the oCPC package and build another package after running for a few days.

Of course, you still need to pay attention to the negative words during this period. It is best to negate those irrelevant words in real time, and pause or delete those weakly related keywords. Only in this way can the accuracy of the account's keywords become better and better, and the quality of the account can become higher and higher.

3. Optimize strategy after the learning period

Generally, the study period ends in 7 days. Either the study is completed, which means you have passed the study period, or the study fails, which means you have not passed the study period.

1) Optimizing strategies through a learning period

This means that the system has already run the data according to the data set by the advertiser and has captured the right user group. The system will continue to run the data based on this captured user group. Do not touch the account at this time. If you spend a lot occasionally for one or two days, don't worry too much. As long as the conversion within the cycle is within expectations, the account can continue to run. If the CPC exceeds expectations, you must keep it stable. Generally, one-fifth or two-fifths of your CPA is possible. Do not adjust the CPA when the CPC is occasionally high, otherwise it will be useless.

2) Optimization strategies for learning period failure

This shows that the system has not captured enough target users and your model has failed. You need to delete the oCPC package and go back to CPC. Continue to optimize the account for a while and try again. Summarize the situation of the account in the past 7 days and apply it to the account for continuous optimization and improvement.

The above is all the content of this sharing. In short, oCPC traffic is a process of balancing advertisers and the system. Everyone needs to find the balance point and obtain the maximum traffic at the lowest cost.

Author: Li Gongzi

Source: Aichisem