How to implement Baidu information flow delivery from 0 to 1?

Baidu information flow was developed after Baidu search. Therefore, based on Baidu's big search, the delivery of Baidu information flow is somewhat different from most information flows.

1. Delivery backend

What distinguishes Baidu Information Flow from other advertising backends is that it has a [Promotion Client] , whose main function is to modify advertisements in batches (such as time period, region, bid, targeting, certain advertising copy; starting and pausing advertisements; modifying delivery links, etc.).

In addition, you can also copy advertising ideas in batches, so using the Baidu client can greatly improve the efficiency of advertising delivery. However, the client currently only supports ordinary cpc mode advertising and does not support ocpc advertising. The promotion client can be downloaded for free.

2. The advertising background structure of Baidu information flow

The structural hierarchy of Baidu's information flow delivery backend is similar to that of other information flow advertising backends, and is divided into: advertising plan-advertising unit-creative. The contents that can be set at each level are as follows:

One thing that needs to be noted is that in the cpc billing model, the URL follows the creative , so my personal habit is to put multiple creatives under one unit, with one URL for each creative; while in the ocpc billing model, the URL follows the unit , so only one creative is placed under one unit. This is conducive to tracking the creative effects of advertising.

3. Account Structure Construction

The establishment of an account needs to be based on the structure of the delivery background and is closely related to subsequent advertising data analysis.

Starting with the required data analysis, the data we may need to consider include: the effects of different layouts, the effects of different ad positions, the effects of different targeting, and the effects of different creative ideas.

(1) CPC model

Since the URL follows the creative, my personal habit is to have 1 advertising plan, 1-2 units, 2-5 creatives in each unit, and set a different URL for each creative to facilitate subsequent tracking of each advertising effect.

In the early stage, iOS and Android are launched separately, which makes it easier to control bids and costs, and is also conducive to subsequent data analysis by device.

The structure is as follows: advertising positions are differentiated by plan; targeting is differentiated by unit; creative ideas with good effects are kept within the unit, and those with poor effects are closed.

Subsequent analysis and effects: Under this structure, by vertically analyzing the data of each plan at the advertising plan level, such as exposure level, speed, CPC, CTR, the size, effect, bidding cost, etc. of each ad position can be compared.

Taking Baidu large pictures and group pictures as examples, under the same orientation, the large picture ctr>group picture, cpc<group picture, registration cost<group picture, but the population quality is lower than that of the group picture. Therefore, in pursuit of the needs of high-quality users, we will adopt measures such as uploading more sets of pictures and testing a small number of large pictures; if there is a demand for impulse, we can upload more large pictures. Targeted purchase of traffic.

Through the unit level , the advantages and disadvantages of each orientation are compared horizontally, that is, the quality and fit of the population are defined. Taking interest targeting as an example, different targeting labels can target different groups of people. Interests in travel, life services, and healthcare may have a slightly higher quality population, but the bidding is fierce; interests in cars and games may bring in low-priced male registrations.

At the creative level , compare different creatives with the same orientation horizontally, pay attention to indicators such as CTR, CVR, back-end registration and crowd attributes, and eliminate the weak and the strong. In this way, the creatives of the entire account are basically high-quality, and the system will find it easier to determine that this is a high-quality account.

(2) OCPC mode

Since the URL follows the unit, only one creative is set in one unit. 1 advertising plan, 2-3 units, 1 creative idea in each unit. iOS and an are served separately.

The structure is as follows: advertising positions and targeting are differentiated by plans; and the quality of creative ideas is differentiated by units.

The specific analysis ideas are similar to the cpc model (see above).

Please note: Avoid launching too many units at the same time under the same plan, as units may compete with each other for volume. The same is true for creativity under the same unit. A Baidu information flow account can only create up to 99 plans, so it is necessary to summarize the delivery situation in a timely manner and delete useless plans.

4. Advertising Targeting

Orientation (very important)

The most interesting thing about Baidu information flow is targeting, and there are many things to try.

1. Intent word targeting: Since Baidu started out as a search engine, keyword targeting is the most important targeting method for Baidu’s information flow. It is accurate and low-cost. Compared with keyword targeting, device packages seem to have no advantages.

In the early stage of account launch, we chose to run keyword targeting. The established model is accurate and will hardly deviate after entering the second stage, and the volume is not small . When you find that the amount of material is decreasing, you can gradually increase the interest/interest expansion orientation. Remember to add slowly and don't select all interests at once.

Source of word packages: agent pull, media pull.

Selection of word packages: Try to choose ones with high PV, because the magnitude of keyword targeting is not as large as interest. If the coverage of the word package is too small, it will be difficult to expose it. When placing an ad, first select the core words of the business to be placed, and then gradually cover other types of word packages. Different word packages have different bids. The higher the relevance to the business, the higher the bid.

2. Interest Targeting : Baidu’s interest targeting has something different from other platforms: interest expansion. The amount of single interest expansion is relatively large, so in terms of interests, you can try: interest combination, single interest expansion, interest expansion combination. The comparison of different interest combinations and interest extensions can be carried out based on the account structure in the second part.

3. LBS: LBS is suitable for industries with offline stores, such as catering. Just enter the name of the business district and limit the scope of the delivery. The operation is more complicated and the flow rate is small, but the cost is considerable. LBS and regional targeting overlap, and you can only choose one.

LBS targeting can be enabled by the agent. If it is not enabled, there will be no such option.

5. Delivery and Optimization

The following optimizations are for ocpc ads.

1. Observe CTR: Generally, when the exposure is 5000-10000, CTR will be of reference value. If the CTR is too low, you can set a time period/next day delivery to see if there is any change. If it is still low, turn it off.

2. Observe the front-end conversion: Generally, an advertisement will be observed for 3-5 days. If it has not entered the second stage within 3-5 days, it means that the targeting is too narrow/the conversion cost is too high/there is a problem with the landing page, etc., and it needs to be closed. If you have entered the second stage but the conversion cost is high, continue to observe and make a judgment.

3. Observe the back-end conversion: If the front-end conversion is OK, then observe the back-end conversion. If the backend conversion cost is high, there may be a problem with the ocpc code embedding point on the conversion path. Ask the developer to reconnect it to a more reasonable position. The closer the conversion between the frontend and the backend is, the smaller the error.

6. Personal delivery skills

1. In Baidu's information flow, keyword targeting and interest targeting are theoretically overlapping, but in actual delivery, if you select keyword targeting and then interest, the volume will also increase. Therefore, when the running volume of materials is not large, you can try to increase the interest; there are some cases where pure interest does not run in volume, but interest + word package runs in volume.

I once had a plan. At the beginning, I used word packages to run it. It started to decline after about 5 days. Later, I added a few new interest-oriented ones with large traffic (life services, entertainment, beauty shopping), and the scale was able to continue.

2. No interest restrictions vs. Select all interests: The amount of money you can get from these two options is different. For a plan that allows you to place an order, try Select all/No interest restrictions . The final delivery effect may be different.

3. For the ocpc plan that has started to grow, when it comes to the material cycle and the volume slowly decreases, don't rush to shut it down. You can try to directly replace the picture material in the original plan, and you will find that it can start to grow again. I have a plan that consumes about 20,000 yuan a day, which gradually declines after 1 week, but because a large amount of conversion data has been accumulated in the early stage, I directly replaced the picture and kept it for 1 day. Later, it was able to run at about 20,000 yuan/day, and it ran for nearly 1 month in total.

4. The volume of Tieba is good for some industries, but the volume is small. You can try to place ads , but you can't ignore this advertising position;

5. For Baidu ads, the budget cannot be modified frequently . Especially when a certain volume plan has a limited budget, the budget modification range cannot be too large and should try not to exceed 20%. For the ocpc plan, the bid price cannot be adjusted frequently, and I personally feel that the bid price adjusted on the same day will not take effect until the next day;

6. Many people say that Baidu’s information flow fluctuates greatly, and a good plan today may not work tomorrow. Excluding the influence of time period, the possible reason is that the system will re-evaluate this plan at 0:00 and provide different traffic. Therefore, you can adjust the planned budget around 0 o'clock to avoid excessive consumption in the early morning. When you go to work the next day, there will be no budget and high costs, which will make the whole day painful.

7. Baidu information flow algorithm (personal guess): Baidu labels people based on search keywords + news keywords. These labels belong to medium- and short-term exposure, that is, a person’s behavior in a short period of time, and the underlying data of the labels is updated very frequently (such as by day, week, or month). Divide the traffic pool according to labels. After the ad is targeted, the ad that matches the targeting tag will be displayed.

This may be one of the reasons why the delivery is very unstable. Many of my friends have indeed mentioned that Baidu's information flow fluctuates greatly. The properties of Baidu feeds news make it impossible for Baidu information flow to accurately target gender/age. All it can do is to get the desired group of people as much as possible, that is, to keep trying to target.

For Baidu advertising , please contact Qinggua Media

Related reading:

How to use the reverse method to improve the bidding effect of Baidu?

With these 3 tips, you can triple the effect of Baidu bidding promotion for common words!

How to bid and adjust prices for CPC and OCPC users of Baidu Search in 2020

Baidu search promotion ocpc daily optimization guide

Baidu bidding promotion strategy formulation method