Super practical Baidu information flow account volume skills

Many people have this question: the bids are the same, the targeting is the same, the creativity and the landing page are the same, so why do some plans perform better? The good thing here is that it can be any dimension such as click-through rate, conversion rate, display, consumption, or these combined dimensions.

Quality score is the solution I hear most often, but what exactly is quality score and what does it relate to? I’ll talk about that later. Here, let’s talk about ad ranking first.

The same advertising plan may perform well in some cases and poorly in others. The reason behind this is the problem of traffic. But why does the same plan get different amounts of traffic? This involves the issue of ranking.

Based on the ranking method of similar programs, I divided the media I know into 3 categories: elementary, intermediate, and advanced.

Primary media: advertising algorithms are the most primitive

The bidding method is mainly cpm and cpc

From the perspective of advertising ranking, it is relatively simple and direct, just look at the bid. The bids for this type of media have hidden bids, that is, there will be a slight difference between your bid in the optimized account and the actual bid in the competitive ranking. On the primary platform, this difference is basically random. For example, if the account bids 3.99, the actual bid may be 3.991. The difference is mainly in the number of decimal places. The number of digits and the difference are all related to the core algorithm of the media and are kept confidential.

ECPM is not mentioned here just for the sake of convenience. The final ranking of ECPM of primary media is the same, whether it is ECPM=bid*click rate or ECPM=consumption/display . Most of this type of media have just started commercialization, and many of them do not have their own independent platforms and directly rent third-party advertising service platforms.

Intermediate media: The account has a preliminary level of quality

Quality is only related to the money in the account

The quality index is added to the ranking of advertisements and the primary randomness. This quality index is for the entire account and is related to the money in the account.

It is related to the budget, balance, historical consumption, etc., but only a few people in the media know how to calculate the specific quality, and the changes should be recalculated based on weekly, monthly, quarterly, and annual cycles. If the budget, balance, and historical consumption of an account are all large, then the quality must be very high. For plans with the same conditions, the system will have a clear traffic bias. In terms of judgment criteria, there are framework accounts and port accounts similar to Baidu.

Advanced Media: Quality at many granular levels

Related to historical consumption, landing pages, creative materials, etc.

Take Toutiao as an example. Ad display has to go through the process of rough sorting, fine sorting, and other filtering. The process is shown in the figure below:

The rough ranking and fine ranking here are related to the quality. From the perspective of ecpm, the main quality of the headlines is related to the bid, estimated click-through rate, and estimated conversion rate. How to estimate the click-through rate? This requires calculation based on multiple dimensions such as historical delivery data and click data of the entire industry.

Knowing the ranking problem of the media, the next step is to optimize

Primary Media:

As we said above, on the primary media platform, there are two plans with the same bid on the delivery side, but the bids in the actual traffic segments are slightly different, which means that the actual rankings are different despite the same bids. This difference is random.

Because of the existence of randomness, when testing, in addition to materials, orientation, time periods, etc., I will also test accounts and plans. The specific approach is to test the account, and deliver the same materials, bids, targeting, and budgets on different accounts at the same time; to test the plan, directly copy the same plan and do the test.

However, primary media generally has little to do with ports and qualifications. Optimizers focus more on time periods and bids, because the traffic of such media is generally not very stable, and brand advertising or fine-tuning of algorithms will have an impact; and this kind of response can only be adjusted in time periods and bids. If the backend effect of this media is good, I will generally copy a few plans that are currently running well and wait for tiered bidding. Different bids will get different traffic, and different businesses will have different effects.

Intermediate Media:

Mid-level media are essentially the agency ports and accounts that bring the most revenue to the media. The media will have a certain amount of traffic tilt in subsequent placements. Of course, in order to balance the profits among agents, relatively small ports will also have a certain proportion of traffic taken care of. Therefore, for this type of media, the agents often say, "If it doesn't work, I suggest you change your qualifications, port, or account."

For this type of media, I usually get accounts from 2-3 agent ports. In addition to comparing rebate policies, I will test the accounts on each port myself before deciding which agent to get the account from.

Premium Media:

Optimization revolves around 4 dimensions and 3 stages. Optimizers need to be familiar with the business they are promoting and understand at each stage that what they are optimizing is the review and how to modify it. Optimize and improve step by step, and finally achieve the effect required by the customer (boss).

There are many small details in high-end media, and the account cycle is generally very short, so optimizers need to keep making plans. Many times we make plans just for the sake of making plans. In addition, with the intelligence of Toutiao, a lot of creative ideas can be generated just by entering keywords. Many people ignore their own optimization methods and optimization rhythm.

Every optimizer should have his or her own method and rhythm on what to do during the cold start, growth period and stable period. Finally, I would like to remind you to read more media documents and familiarize yourself with every part of the account background. I read them often and they often give me a lot of inspiration and make me feel like I have suddenly realized something.

Summarize

Some small habits that I usually optimize

1. For primary and intermediate media, when running a plan, it will generally be turned off after 3-5 minutes of operation, paused for a while, and then turned on again; I can't explain the reason, but after I personally do this, the effect is better than not doing it.

So far, I have encountered this phenomenon on three media platforms. At the end of 2018, I ran an information platform in the fan-adding industry. Before this operation, the cost was around 90, and the consumption speed was very fast at the beginning. Once a plan just started, the customer called a halt and replaced the server. Later, it continued to run, and the consumption became very stable, and the cost dropped directly to more than 50. The attribution analysis finally found that only the plan was suspended.

All the accounts of this media were planned to be operated in this way for more than two months, at a cost of about 30 million; the highest cost was about 60 million, until the media backend was upgraded and rectified.

2. When bidding, all the numbers after the decimal point are generally set to 9, such as 0.29, 71.99.

3. For advanced media, in addition to changing ideas and plans, the landing pages will also be changed and fine-tuned from time to time.

4. Check your own advertisements on the media every day to see the display, and you can also check out competitors’ advertisements.

5. In order to increase the volume, sometimes people will post their own ads to improve the click-through rate.

Author: Jiang Mosheng

Source: Jiuzhilan