Baidu ocpc delivery operation skills

As the title says, 2021 has passed and we welcome 2022. In the past year, Baidu's search for "ocpc" has remained a key topic of concern. Tagore also popularizes some ocpc knowledge and shares it with everyone from time to time, hoping to help everyone.

This time, the content mainly collects various methodologies about OCPC in the past year and makes a simple data presentation of the practical results. There are pictures and practices, but the effects will vary depending on the industry and model. This content is still for experience sharing and is for reference only.

Next, I will disclose the practical data of OCPC deployment of new accounts in my 21 years of multi-account OCPC strategy.

Answering some questions through daily notes and data presentation will get you closer to the truth. But it is not ruled out that there will be some differences after the function update. For example, the current enhanced mode can be simply understood as a marketing hosting version under the ocpc model; and the target conversion cost: the conventional ocpc running method, there is no doubt about this.

Large-volume mode: A relatively intelligent 0-threshold mode with small conversions in the early stage due to large volume, but stable conversions in the later stage. The updates of these functions are mainly aimed at the current different account situations. The adjustments made in Baidu's backend are really increasingly able to meet everyone's needs, and the effects vary depending on the situation. Overall, ocpc is still the best choice for guaranteed results.

This article is suitable for readers who have just come into contact with ocpc. The maximum limit is in the medium and short term.

The veteran optimizer of ocpc can just float away. Aims to share some operating skills and simple optimization ideas of the ocpc system.

This article cannot guarantee that all the ideas are correct, but they can be said to be effective.

After all, data cannot lie. From the data to the effects to becoming a case, everything really happened. I hope it can help more students.

I have just got the account and have been running CPC throughout the whole process. However, in order to be prepared for any hot products, I have given all these plans the first phase hat. You can see that they are all in the first phase and the data has not been verified yet. There are 2 plans for PC and WAP respectively, and the normal promotion monitoring data is included.

If the conversion effect of the first order is good enough, it would be ideal to directly enter the second order in normal mode. But I know that good things must come with great difficulty. It is impossible for me to pass the second level so easily, otherwise I will not have the materials to present more possibilities to everyone. So students, please continue reading.

The four plans have been running for more than half a month, and have been constantly adjusted in the past 7 days, but we still find that the conversion effect on the mobile terminal is not ideal. The cost is extremely high. But this is the cpc stage, so there is no need to be lenient when making adjustments. Once problems are analyzed, adjustments can be made directly to make the foundation of the model more organized.

In fact, because the WAP side of the previous old user is already very good, as I said in my previous article, there will be competition among the models for volume, so the WAP side data of this new user is not good, but it does not mean that it cannot be optimized.

At this time, Tagore’s idea is that the conversion cost is high and needs to be reduced, but since the conversion cost is so low, the final conversion cost is so high. This indicates that the conversion cost of the mobile terminal is set unreasonably, and the conversion cost should be increased within an acceptable range and observed again. Of course, this is the first stage, and CPC operations are more biased, but the thinking of the underlying logic cannot be ignored. At the same time, considering that old users are competing for market share on mobile devices, if we cannot break through, we should consider starting from the PC side and entering the second-level situation as soon as possible.

It is not difficult for students to imagine here that we can change the target conversion bidding mode to use the conversion bidding in the data accumulation stage. In this way, the system calculates the consumption and conversion costs within a time period. The key to this knowledge point is that since the system will be promoted intelligently in the later stage, we can start from the needs of the system and let the system better understand the entire promotion process. This will help the system better understand your model and the target audience. PS: Modifying the target conversion bid can generally only be done in a delivery package with a non-zero threshold.

As expected, the famous scene came: as shown in these two pictures, PC got the recommendation value because of its better overall effect, and then it quickly entered the second stage and was about to start smart promotion. I felt quite encouraged.

After WAP modified the strategy, it quickly brought out the recommended value. This recommended value can be said to be a good stepping stone for the system to help you enter the second level. With the experience of PC, Tago immediately modified the bid. But one piece of knowledge must be mentioned here: should the bid be completely in line with the system recommended value or can it be slightly lower than this value?

As shown in the picture, we also know from our previous theoretical knowledge that the bid can be based on 80%, but Tagore doesn’t like to stick to the data too strictly, so he sets the conversion bid according to 81%. (Pictures are for reference only)

So we can basically conclude that:

If the delivery package cannot enter the second stage automatically or through recommended values, you can try to modify the target conversion bidding strategy for testing. But again, don’t rush to do too much when entering the second stage. Give the model at least three days to run the volume. I won’t explain the reason, as I have said it too many times.

Special reminder: If the conversion bid is still high after three days, you can lower it, but many teachers or materials also emphasize that each adjustment should not be too large. Some students directly reduced their performance by 30-50% after entering the second stage, which resulted in their model becoming inferior, which was not worth the effort. Tagore recommends adjusting the bid by 10-20% each time. The ideal adjustment is 10-15% each time. In addition, the timing for adjusting the conversion bid is generally chosen after the planned budget is completed. In this way, the model can run according to a higher conversion bid the next day, and the competitiveness will feel better, but it is not necessary, this is purely personal experience.

At this point, I believe that careful students will find that there is still a WAP plan that has not yet entered the second stage and has not seen the recommended value. Yes, that’s right, Tagore encountered the “thorn” plan delivery package. What to do? Let’s look down together!

Optimization continued until November 7, and most plans entered the second stage. However, there was one plan that had not seen any progress. At this time, Tago directly enabled the "0 threshold" mode and edited the plan directly to 0 threshold. It took a few days to complete the plan study. In order to ensure that there was more competitiveness to compete for traffic during the study period, Tago deliberately adjusted the conversion bid to a bid similar to the recommended value of the previous wap plan to ensure the stability and continuity of the 0 threshold traffic.

Fortunately, after the study was completed, there was no prompt that the study had failed. I can't help but feel a little excited! This famous scene is quite exciting.

So the question is, do you think it ends like this? Students, if you look at the progress bar of the article, you will know that things are not that simple! I have always adhered to the concept that "anything produced by Tagore must be high-quality", so how could it end in such a hasty manner?

The next step is to make adjustments based on data analysis and optimize operations to reduce costs as much as possible. Please continue reading.

Let’s take a look at these two pictures again. The 0 threshold was just introduced not long ago on November 7th, so the cost is still quite low, but the most feared thing is “peak at debut”, so Brother Tagore paid attention to the changes in conversion costs. If not, the famous scene continued: when No. 7 was feeling complacent about its low cost, No. 8 was hit by a bolt from the blue, and the cost suddenly increased from 35 to 58, which almost killed Tagore.

After analyzing the reasons, it seems impossible that old customers with similar models have grabbed such a huge amount of volume. This factor only accounts for 30% at most.

Now back to the account itself. The cost has increased and to reduce it, I can think of a series of positive actions such as rejecting words, optimizing creative ideas, modifying the planned budget, adding keywords, and even adding creative (flash project) styles. But I clearly remember that on the 7th, I adjusted the conversion bid from 50 to 40 (I forgot to take a screenshot at the time). This increase was exactly 20%, but I didn’t expect that the cost would start to soar on the 8th, so I changed the conversion cost back to 50. After all, I had just entered the second level for less than 7 days. Decided to observe it a little longer.

Hey! The famous scene really happened: on the 9th, after promoting according to the conversion cost of 50, the cost dropped from 58 to 26. It is indeed correct and effective. The cost has dropped to a level lower than before, and the subsequent costs have basically hovered between 25-35, with an occasional high of 45, but overall it is very stable (I won’t take more screenshots here). After all, there is also an old customer whose mobile ocpc plans are competing for volume, so Tagore doesn’t have too high requirements for this model.

Until early December, the data remained stable. At this point, the ocpc setup of the two accounts is basically completed, and there is no need to worry about various emo due to the fluctuating costs.

At this point, Tagore summarizes a few points to share with the students:

The above is some operation record sharing of Tago multiple accounts for ocpc. There are still some details that cannot be captured in screenshots for every operation. I hope this sharing can help more students open up their minds. As ocpc continues to adjust and change, each time period has different or even opposite operating methods, and everyone needs to practice and summarize more according to needs.