What is the exposure mechanism of information flow advertising?

When doing information flow advertising, you may encounter such problems: your creativity and your plan suddenly run out of volume, or there were conversions one day, but suddenly there was no exposure at all the next day.

When you encounter this problem of "no volume and no exposure", excluding some objective factors such as channel flow restrictions, you should consider the exposure mechanism of information flow advertising!

Coarse sorting & fine sorting

First, once we set the bid and targeting, your creative and account setup are complete, the ad will go online for review.

After the system review, it will give you an estimate, which we call "rough sorting". For example, it is 8:00 p.m. now, and 100 new ads have been uploaded to a certain ad space in the Toutiao APP. The system will perform a rough sort based on a series of factors such as your CPC bid, estimated click-through rate, etc. After the sorting is completed, the system will compare them. If an ad wins, the system will display the best and highest-scoring high-quality ad.

If there is a click during the display process, OK, add 1 point! The reward is that you will get higher performance!

But don’t be happy too soon, because if no visitors click on you after the display, OK, deduct 1 point, and then go back to compete again! "The law of natural selection and survival of the fittest is most appropriate in this circle."

In-depth exposure mechanism

Here we combine the case analysis, taking the promotion of Douyin advertising business as an example

Step 1: Filter out ads that you are not interested in or that do not fit your creative style (the remaining 15%)

Assuming that 1 million new ads are posted at 8 pm, the system will exclude other ads that are not within the targeting range.

Let me put it this way, if the advertising space is for investment and franchise, the creative labels, creative content and industry account qualifications of the user's advertisements are for the investment and franchise industry, accounting for about 15%. This means that these 150,000 advertisements are in line with the targeting, and the system will filter out the remaining advertisements from other industries.

Step 2: Filter out user requests that are not within the ad targeting range (the remaining 70%)

The second wave of the system will filter out users who have seen the same type of ads too many times. This is why we have been encouraging everyone to use original creativity. If the system determines that yours is original, it will give you about 70% exposure opportunities. Even if you change a character, you will get more exposure.

Of course, if you copy other people’s ideas, the system will filter them based on the ads that have already been run among users. If these ads are already being run or are similar, then I’m sorry, they will be filtered out directly. The result waiting for you is less exposure or no exposure at all!

Step 3: Filter out ads that you are not interested in or that do not fit your creative style (the remaining 99%)

The system will evaluate users based on their attitude towards ads. Approximately 1% of users will be uninterested or will directly click “×” on ads. Congratulations to the remaining 99% for entering the next level.

Step 4: Filter ads that do not have enough available budget and balance (100% remaining)

Continue filtering. Ads without sufficient available budget and balance will not be delivered!

Step 5: Estimate the probability of being successfully placed after entering the rough ranking, and enter the rough ranking (remaining 100%)

If you have passed the screening process without any problems, it means you have the opportunity to be shown. Your budget, presentation, quality and type are all good, and the system will give you higher exposure.

Step 6: Those with higher ECPM after rough sorting are qualified to enter the refined sorting stage (the remaining 10%)

In the stage from step 1 to step 5, you will find that 90% of the ads are useless in the rough ranking stage. Basically, they are still exposed in the first one or two days, but the effect begins to decline on the third day. This is because the ads have not entered the fine ranking stage.

As for the rest, you’ll find that the creative ideas that run well usually have a continuous upward, wave-like curve for several days, with exposure increasing day by day.

If it has been running for two days and the performance in the rough ranking stage is good, and the exposure has been increasing for two days, then on the third day it will face two directions. One part is a decline, and the other part is to get greater exposure, which is the rare 10% we mentioned in step 6.

As shown in the figure above, the degree of fluctuation will have a life cycle. The 10% that survive after the refined sorting mentioned earlier is the maximum range of the deletion.

Endless competition mechanism

If you make it through the "fine ranking" stage, you will have about 5,000 ads that can participate in the final exposure competition.

The ads at the front will move on to the next stage, which is basically another competition process.

In the end, there may be only 1,250 ads left out of the 1 million. For the ads that can survive, if we assume one step per day, basically no creative ideas will survive more than three to five days. You must be asking, it's just a casual vote, why is it so complicated?

Because the competitiveness of one advertising pool is different from that of another, during the competition process, competition among the same type of advertisements and user interests will affect core competitiveness, and it is also restricted by factors such as "advertising freshness, category freshness, remarketing, etc."

Stop bidding

After talking about the competition mechanism, let's give an example about bidding: "If the conversion cost on the first day is 50 yuan, but the budget cost is 70-80 yuan, you think this is suitable, so you want to reduce the cost a little bit, so you reduce the cost by 5%. As a result, the model changes, which directly affects the system's bid for the entire tag population, and the cost instantly soars to more than 200.

From 50 yuan to 200 yuan, this is a typical example of 'the greed of the people leads to the snake swallowing the elephant'!" Therefore, due to the algorithm mechanism of smart bidding, we must be cautious when bidding when placing information flow accounts.

The system must have a stage for collecting information. When the conversion cost is appropriate, it must not be easily modified. "Contentment brings happiness."

Don't rush to throw away the original plan, you can use it for future account analysis

The so-called "degree" refers to when the cost of the new planned unit is twice the expected cost, you can choose to suspend it.

For example, friends who run Baidu information flow have experienced that you spent 200 yuan in two hours, and the estimated cost was 80-100 yuan, but you found that it has doubled. In this case, you can directly choose to change the word package. If this happens to Toutiao, you can change the creative idea;

After we made the change and ran a run, we found that the cost was another 200 yuan, but there was still no conversion. Then you can pause and create a new plan. Never make multiple changes to the original plan, because smart bidding has a model. It is automatically optimized based on the system conversion tags and modeled according to the tags. Therefore, the early model is very important, but it is difficult to modify later.

Therefore, when a plan (OCPC/OCPM) has a cost of two to three times in a row and the effect is still not good, and optimization is of no avail, you can give up! The original plan for a newly created plan that has not been running for long should be kept for one week and then deleted after one week for final data sorting. For a plan that is running well but suddenly has no volume, you can keep it for one month for data analysis.

I hope that this article will help advertisers and information flow practitioners gain a relatively clear understanding of the internal mechanisms of ad exposure, and truly achieve “prescribing the right medicine for the right disease”, and ultimately attract a large number of target consumers!

You must understand that only by properly controlling the exposure can you be the upstream source for subsequent high clicks and high conversions! Only through real user and data feedback can we continuously optimize our delivery methods.

No information flow advertising can last forever.

Related reading:

1. Information flow advertising algorithm mechanism: one formula applies to the entire industry!

Author: Quail Egg

Source: Houchang College