What exactly is being optimized when optimizing information flow ads?

Is the following scenario the daily work of most optimizers?

The optimizer's daily work is to open the backend in the morning, download data, analyze it for about 2 hours, prepare to eat, prepare materials in the afternoon, pass the review, and wait for the judgment of fate.

Understand the traffic distribution mechanism, develop optimization strategies based on user conversions, and understand the three factors that affect creative click-through rates. In fact: information flow advertising is simpler than you think.

1. Understand the traffic distribution mechanism

Optimizers can view the media they deliver as "friends". After knowing a friend for a while, you will definitely discover some of his or her preferences and shortcomings.

Take WeChat MP as an example:

① Why is it easier for a new plan to gain volume on the second day?

The advertising mechanism of Moments is to push ads twice a day, with peak traffic periods being 7-9 a.m. and 12-14 p.m. New plans will enter the new advertising pool, where traffic is limited.

Plan review is an irregular review. If the traffic peak period has passed, the planned data delivery will be relatively slow and will not receive effective data support. The next day, the system will think that your plan is not liked by users and will reduce the push of your ads.

② What should we do in the new advertising pool?

It can be considered that the new plan at this stage is a crying baby. Now no one has the skill of "quality score". At this time, we can only rely on bidding to strive for the distribution of exposure traffic, based on the principle of high first and low later. When the "advertising diagnosis" of the plan has scoring parameters, it will enter the regular advertising pool.

③ What to do when entering the “old advertising pool”

Keep the bid unchanged for now, try to control the planned traffic by controlling the "time period and budget", and then gradually adjust the bid after 24 hours, with each bid adjustment increasing or decreasing by 5-10%

④ Traffic is particularly high during holidays

The circle of friends during holidays is definitely a battleground for all mine owners. Every month, Tencent social advertising will push the marketing hot spots of the next month. We can refer to it, do the maintenance and update of the plan in advance, and try to accumulate the plan data, so that the impact is relatively limited (except for major holidays such as 618 and Double 11)

2. Develop optimization strategies based on user conversion

Take page registration as an example: exposure-click-H5 page browsing-button click-submission success-crowd screening

Some users will be lost as they go through each step, and in the end, there will probably be only a few who can become our users (even high-quality users).

The step that optimizers can control is the registration stage. The subsequent payment behavior is actually more closely related to the content and functions of the APP.

So how can we optimize this conversion funnel and turn more of the hard-earned traffic into our registered users? In fact, we can use a method called the "Four Quadrants Rule"

① High cost and high registration: We usually observe this plan for about 24 hours. If the cost is too high, we will suspend it immediately. But don’t ignore it. You can collect this plan and review it in time to find out which selling points are easy to impress customers and reorganize the selling points.

② High cost and low registration: This plan will be suspended directly. On the one hand, it is easy to increase the historical conversion data level of the account. Secondly, it is also a waste of money.

③ Low cost and high registration: One of the plans that optimizers want to test most. Generally, optimizers will conduct large-scale operations, but how to do it?

At this time, it is recommended not to increase the volume too much at once, which will disrupt the system model. For example, if I increase the budget by 10% today and 20% tomorrow, the positive feedback mechanism system will think that the quality of this plan is good and distribute the traffic.

④ Low cost and low registration: The cost is good, but there is no volume. This kind of plan is also common. So how should we scale up this plan?

1. Increase the budget in steps by 10% on the first day and 20% on the second day.

2. Copy the same plan and modify it to run volume for testing

The "Four Quadrants Rule" uses terminal data to push back to the front end. This optimization idea will be more convincing

3. Understand the three factors that affect creative click-through rate

1) Understand the type of media to place the advertisement , such as consulting, social media, vertical websites, etc.

2) Creative level (taking Toutiao Loan as an example): The process of the audience seeing the advertisement is: advertisement image + copy (click rate) - enter H5 (conversion rate) - download - install - activate - register - loan amount

Taking the headline industry data in September as a reference: Android's cpc: 2.55, ctr: 1.4%, iOS's cpc: 3.55, ctr: 0.86%

If our creative CTR is lower than the market level, many optimizers may change the creative at this time, but I wonder if anyone has realized that we may not have found the right person (targeted setting).

3) Landing page level: Make sure it is relevant to the creative idea

Three-stage layout: Use three screens to express “who you are, what you can provide to users, why choose you/the finishing touch”

Take home decoration as an example: the first screen is mainly composed of large pictures, and try to set up two forms to fill in. I’m sure some students will do this, and the two forms are set up exactly the same. If it were you, would you trust something so quickly when you first saw it? The first form should be as easy as possible for customers to fill out, even if they only need to fill in a mobile phone number.

4) Effective guidance: The button color should match the entire layout, and the copy should use a series of action words such as "immediately, right away, try it" to encourage users to take immediate action.

Da He has something to say: Optimizers are actually dealing with data all the time, but they need to see the essence through the phenomenon. Certain data can cause a series of chemical reactions. Don’t be afraid of trouble, learn more and test more.

Source: Information Flow Weekly