Tips for optimizing information flow APP delivery!

In the actual delivery, how do we find the problems in each link and how do we solve them? Here we will talk about it in four parts.

1. Display

First of all, it’s about presentation. In which aspects should we improve presentation? When there is a problem with the display, how to solve it? To sort these out, we must first understand the logic of presentation.

Display logic:

• Ads loaded - users need to see ads when swiping their phone

• Filter ad requests from users who are not in the ad targeting range - select ad groups that meet the targeting

• Make sure you have enough budget and balance to bid

• Filter out people who are clearly not interested in the ad and click Dislike

• Filter out people who have seen the same type of advertisement too many times and enter the rough sorting

• Estimate the ECPM of the ad, ECPM = bid * estimated click-through rate * estimated conversion rate, set a rough threshold value

• Multiple PK, select the best according to ECPM; the roughly sorted population is then sorted according to the ECPM value

• In the absence of brand advertising, ECPM bidding wins and article mixing

• Successful exposure

From the above, we can see that the real factors affecting the display are targeting, budget balance, market environment and ECPM value . Let’s talk about them one by one.

1. Orientation

There are two types of targeting methods. One is broad targeting in the early stage , which helps to increase sales and quickly get through the cold start period. The other is precise targeting in the early stage , where the system targets precise groups of people, and the planning model is stable in the later stage and then gradually released.

These two methods have their own advantages and disadvantages. Method 1 may not have accurate running volume in the early stage, and method 2 may not be able to survive the cold start period or the cost may be too high. In the actual delivery process, it is recommended to use both in combination and configure a reasonable ratio according to different products and businesses.

For example, when investing in a product with a broad target audience, such as short videos, 80% of the plans can be loosely targeted, leaving only a small number of plans for precise targeting testing. If you are investing in animation or children's early education products, it is recommended to use 80% precise targeting and 20% loosely targeted to increase the account volume. The above ratios are not absolute and can be considered based on actual conditions.

2. Budget balance

If the consumption of an account or plan reaches the limit, the system will limit the flow, so make sure that the account has sufficient balance and the plan has sufficient budget.

So does it mean that the more balance in the account, the better, and the larger the budget set in the plan, the better?

Actually, it’s not. I saw a question and answer in Today’s Academy before. It said that if a plan can run 10,000 yuan a day, it is enough to set the budget to 15,000-30,000. If the plan starts to gain volume, the optimizer can gradually loosen the budget of this plan. Some students may ask, if I set the planned budget to 100,000, what will be the consequences? No one can say for sure. The effect of information flow delivery is determined by many factors. We just need to do our best.

3. Market environment

There are some large-scale event nodes every year, such as 618 and Double 11, or other advertisers pay high prices to grab volume frantically, consuming more than 10 million a day. At this time, it will become more difficult for your ads to be exposed.

We generally cannot control the overall market environment, but we can adjust our own delivery strategies. For example, if you want to complete the new tasks in Q4, you need to advance the budget and seize the golden period of 7 days of National Day to gain volume, or put the budget in other channels that are less affected.

4. ECPM value

ECPM value is the core of optimization and is mainly determined by bid, click-through rate and conversion rate.

The bid can be controlled manually. Considering the input-output ratio, we will try to set it as low as possible, so the core is the click-through rate and conversion rate, both of which depend on the material.

There are generally two directions for materials: one is to summarize the selling points of your own products and update them iteratively; the other is to learn from the running materials of competitors. The channels include App Growing or directly asking media operations. For Kuaishou information flow, you can go directly to the Kuaishou account of competitors to view it. We will discuss the specific details of optimizing click-through rate and conversion rate in detail in the click section below.

So if we do the above points, will the account display be good? Not necessarily. As shown in the ECPM analysis above, the estimated conversion rate and click-through rate are determined by the media, which are uncontrollable factors. While we are doing a good job of the above points, we need a large number of new plans to increase the probability of increasing the volume. Multiple plans and multiple accounts are also beneficial to the stability of the delivery.

In the implementation of delivery, we may also encounter many display problems. Let us make some simple examples:

Q: The entire account is not displayed?

A1: Check if you have been locked up in a black room or restricted. Check if there is any internal message or contact the media for investigation.

A2: Squeeze out volume like squeezing toothpaste, the conversion rate is getting lower and lower, the account ECPM value is getting lower and lower, the account display is getting smaller and smaller, until there is no display, abandon the account, and open a new one immediately

Q: Single plan is not displayed

A1: The bid is low and uncompetitive, and it cannot survive the cold start period. The bid is higher than the industry average in the early stage.

A2: The crowd is too narrow to run out - if it is not necessary, try to relax the crowd.

A3: The estimated CTR and CVR of the creatives are too low. Create and launch them based on the running volume of competitive products and the core selling points of your own products.

2. Click

After talking about display, let’s take a look at clicks.

Kuaishou has cover clicks and behavior clicks, and other information flow channels without cover clicks can also refer to it.

1. Cover click

There are three points to note when it comes to cover clicks: first, it is best to make a separate cover, and avoid mindlessly drawing a frame from the video; second, it is best to have a copy to guide the cover. Taking iQiyi Suike as an example, the copy can focus on the drama or membership activities; third, there should be multiple covers for one video . Avoid having one cover for each video and constantly copying the plan, as this will make it difficult to increase the volume.

2. Behavioral clicks

There are also three points that need to be paid attention to in behavioral clicks. The first is the three-second bounce rate . From what aspects can we reduce it?

• There is a plot conflict in the first three seconds

Take iQiyi as an example. We can capture the most forwarded clips by users as the delivery material, or we can judge the conflict point of the content based on intuition and attract the user's attention in the first three seconds. For example, Xiao Feng ascends the Shaoshi Mountain and Zhang Wuji ascends the Bright Summit, which are climax clips.

• BGM-You can choose the popular BGM on Kuaishou, and make sure it is consistent with the tone and rhythm of the material

• Target audiences based on materials

We have discussed the matching of materials and people in detail in the previous article.

• Scene induction

A sitcom can feature sexy beauties, and a TV drama can feature kissing scenes or sex scenes, but the plot in the first three seconds must be strongly related to the content that follows, and be relatively smooth and coherent. At the same time, it is also necessary to pay attention to the platform's review standards to avoid accidental discharges.

The second is the download button and voice-over guidance . The button can be considered from two aspects. One is the copywriting. For example, for Tencent Video, “Watch “Young Travel” for free” or “Sign in to get a free Tencent membership” are more attractive than downloading immediately.

At the same time, the size and color of the button can also be adjusted according to the overall tone of the video; in the last part of the video, adding artificial voice-over guidance can also appropriately increase the click-through rate of the behavior.

The third is materials and targeting . In the previous article, we talked about personalized material delivery. Here we will focus on the overall material design.

To make an overall design, we need to decompose the overall material. Taking the situational drama material of Suike as an example, it can be decomposed into material scene, product selling point, BGM, material length, and ending version . These five elements can roughly form a video material.

The material scenario can be triggered by colleagues in the office watching a popular drama and needing membership, or by a girlfriend being out of drama. The summary can be summarized as different characters in different locations triggering scenarios because they need membership to watch a drama or don’t know what drama to watch;

The characters and places in the play should be familiar to the target audience and easy to accept. There are many forms. The product selling points are composed of the core content value of the APP and the membership gold coin activities.

The BGM can refer to the most popular songs in the media recently, but it should be consistent with the overall tone of the material;

The length of the material is generally controlled between 35s and 50s, which is concise and not procrastinated;

You can make several copies of the copywriting, and the copywriting should highlight the main interests of the material.

All of the above points can be used to control variables for AB testing and qualitative analysis.

3. Installation

Now let's talk about installation.

The process from clicking the download button to installing it on the desktop is also determined by multiple factors, such as material quality, orientation, and network environment. These have been mentioned in the click part. I will mainly introduce a new feature - direct access to the store, which can effectively improve the conversion rate of this step.

Direct access to the store:

Previously, the Android version directly downloaded the APK link with a progress bar. After the download was complete, the phone would prompt that the app was risky and asked whether to download it from the app store.

After completing the direct store link, you can click the download button to jump directly to the App Store for silent installation. This has two advantages. First, it reduces the user operation steps. Second, some users will be scared away when they see that this application is risky. This is especially true in fourth-, fifth- and sixth-tier cities. Direct store link can effectively avoid this situation, provided that click attribution is done. If it is channel package attribution, all new additions will basically be hijacked by the store, and the data falling into the channel package will be basically zero, which will affect the normal effect evaluation.

4. Activation

In the fourth part, let’s talk about activation.

An application is already on the desktop, why not open it? If you have done a good job in display, clicks and installations, there seems to be no particularly good solution to this problem, but there are three points we can pay attention to:

First, data delay and window period

Students who have operated the account will definitely find that we have suspended the delivery today, but there will still be a lot of activations tomorrow. The reason is that the user clicks to download today and activates it tomorrow. This is what we often call "balance". When will this balance end?

The media generally has a feedback window period. For Toutiao and Kuaishou, it should be 7 days. If the user installs the app more than 7 days ago but does not activate it, then this conversion will not be returned. It also depends on the window period set by the advertiser. If my window period is set to 3 days, then if more than 3 days pass from the installation of the app to the activation, this conversion will not be returned to the media.

Second, the converted people do not return

He was an old user, but uninstalled the app for various reasons. Today he downloaded and activated the app through our ads, but this activation will not be counted as a conversion. The larger the app, the greater the impact. The placement at this time also has the effect of recalling old users. Some companies have strong data capabilities and can split out this data, but some companies cannot, and they will not be included when evaluating the effects.

Third, users are not willing to do so/delayed

For example, if we see a group of people, their download completion rate is okay, but their download completion activation rate is relatively low. This is the case with multiple account verifications. We can screen out this group of people and improve the accuracy of the volume obtained.

The above summarizes some issues and ideas that need to be paid attention to in the actual delivery of display, clicks, installations, and activations. Combined with the new user retention sorted out last time, the delivery of information flow apps is basically covered. Each industry and product may be slightly different, and everyone can supplement it.

Author: Dai Shaohua

Source: Dai Shaohua