Channel promotion for APP user growth!

A successful app always relies on user growth. If there are no efficient means of user growth, the approach of just spending money would have already fallen behind.

This article discusses the channel management aspect from the data level.

For normal APP companies, if they have a budget, it is best to spend 90% of the money on major application markets (mobile phone manufacturers). In addition, Guangdiantong, Chuanshanjia and Baidu can be appropriately allocated for delivery.

The advertising industry has been in a downturn recently, so you can give it a try. The prices of Guangdiantong and Baidu are not high. You can find a business first, and then customize your needs. For example, if you only want a certain position of a competitor, the price-performance ratio is still acceptable.

Don’t invest in an alliance without establishing a channel anti-cheating system, as it is easy to be brushed with a lot of fake data. (If the retention rate is less than 30%, don’t pay for promotion)

If you don’t have money, you can only increase the volume step by step . Don’t have too high expectations, especially if you are an individual developer. First, seize the main battlefield of the app market and include popular keywords in comments, subtitles, and introductions.

It is better to release more frequently, such as once a week; don’t count on free resources, such as OPPO’s first release, they require the top 500 apps to be given the first release opportunity.

If all else fails, spend some money on ASO. It is still worth it in the early stages. Some channels can charge fees based on the profitability of your APP, so be sure to get in touch with them more often. (The volume is basically on these application markets OPPO, VIVO, Huawei, Xiaomi, Samsung, Meizu)

In addition, it is worth mentioning that you must seize the traffic entrance. For example, use short videos, WeChat, QQ, etc. to introduce products and guide users to download.

Finally, you can write a script to batch-post comments on Baidu Tieba and Weibo messages to guide users to download by themselves; if you have accumulated data, you can do email marketing, and you will still get surprises by combining it with your own products.

Although channels are very important, you must first figure out the core points and selling points of your products, and then find your target audience based on your selling points.

For example, our product is a novel reading APP.

So where are the users who like to read novels? They will definitely go to major pirated novel mobile websites to read, right? For example, Biquge. Then, we buy advertising space from these webmasters and guide users to our h5 website or download the APP directly. The cost of a single new user is less than 80 cents.

For example, FaceApp’s traffic initially came mainly from various fun testing websites.

People found it interesting (the product had a selling point), and then they put it on Facebook and Twitter, attracting the attention and sharing of KOLs, which was then promoted by celebrities and the media. Gradually, the world's top KOLs followed suit, and eventually the number of downloads reached tens of millions per week.

Hey, it looks easy, right?

In fact, they made this APP in 2017, with 28 functions, 25 of which failed. Any rapid growth is essentially about finding your own user base quickly and effectively. Channels are just ways, and there are definitely better and cheaper ones; but the premise is whether your APP is interesting, good-looking, and useful.

Since I am engaged in the data analysis industry and have been working for 11 years, I have listed several effective solutions for your reference:

1. Use Python to write crawlers (introduce crawling your own Umeng - taking channel data as an example)

Find the URL in the load_table_data request in F12-network

Copy this URL and open it to find a surprise. Then it is a simple crawling. The core is to complete the header, especially the cookie. Please note that it is in JSON format, which can actually be directly entered into the database.

2. Use MySQL to establish a channel evaluation system

Purpose: We spent a portion of our money on online alliances in the early stage, and we need to evaluate them;

Data source: crawled from Umeng. Key fields include new additions, DAU, retention, user behavior (number of ad exposures and percentage, ad clicks and percentage, number of playbacks and percentage). Umeng’s total exposure and GuangDianTong should be checked regularly.

Processing: Cost accounting is performed based on the conversions (ad exposure and clicks) brought by each channel, that is, deduction of volume;

Feedback: There is a channel promotion background where channels can see their own daily data.

Industry channel anti-fraud can be roughly divided into the following stages:

When the application market was just becoming popular, anti-fraud simply judged a few dimensions such as retention, duration, region, and device. It was obvious that it was easy to be inflated, and even if you set up your own reporting points, it was easy to be cracked.

The second stage is to build your own tracking points, encrypt field reports, and increase assessment dimensions.

The most important evaluation dimension is the association with advertising, such as how much GuangDianTong exposure this channel brings to me, how much CPM it generates, and whether the totals match; if they do not match, then the conversion will be based on GuangDianTong exposure, etc.

At this stage, you can basically prevent channel inflating. Even if you can’t, this channel still brings in advertising revenue, so it doesn’t matter.

The third stage often occurs in large companies with a lot of money and a lot of stupidity. They build their own anti-fraud teams and use machine learning to combat fraud.

The idea is as follows: 1. At the hardware level, report encryption (including report fields, unique user calculation, etc.); 2. Software: perform machine learning based on existing rules (general algorithms are logistic regression), which is essentially a classification problem;

This model is expensive, but useless.

Because this algorithm needs to be developed. If you need to feed the data, by the time you finish feeding it, the time has basically passed. Based on my limited understanding, this direction is a failure in the application of big data. Don’t fight against cheating just to deduct the volume, it still has to be used for business.

3. Relationship between new channel additions and DAU

Judging from the format of the derivation, if you want to increase DAU tomorrow, the new additions required depend on the sum of yesterday's new additions, past new additions and the retention difference.

This model means that today's DAU is the sum of yesterday's new additions plus the retention of the day before yesterday's new additions on the nth day, plus the accumulation of previous new additions.

Similarly, tomorrow's DAU is tomorrow's new additions plus today's new additions multiplied by the retention rate plus yesterday's new additions multiplied by the 2-day retention rate. . . Accumulated.

In this case, if you want your DAU to increase tomorrow, then tomorrow's new additions must at least be greater than the sum of the products of the past new additions and the retention difference (the retention curve is very important, and Excel can do it with sufficient accuracy if it is fitted). If it diverges further, you can consider the slope of the retention curve.

In fact, the 30-day retention curve can estimate how many new users need to be added every day to reach a DAU of one million; and this model can be applied to each channel.

Some additional experience:

1. The quality of Alibaba Cloud OS is average;

2. There is no need to use App Store, it will capture packets by itself. Using App Store is not good for the product because it requires users to download App Store before they can download you. If it captures the packet by itself, the most it can do is not be able to search.

3. All major application markets have free resources (which are scarce and difficult to apply for). It is recommended to apply for products that are stable and good-looking. Otherwise, it will be difficult to provide resources, especially for users who have just launched their products with little volume. Many APPs have volume requirements.

4. When upgrading the APP, it is recommended to jump to the user’s mobile application market to upgrade, so as to increase the download volume. At the same time, guide users to leave more comments.

5. When creating a hit product, the most important thing to consider is the channel. Any hit product needs the support of social platforms. Any meaningless product is unlikely to skyrocket; you can only do every detail step by step.

So I want to say to all the students who are making decisions: even your dreams have to be based on reality.

Before making a product, count your resources first to see which ones can be used; then combine your own resources to design the product. When designing the product, you need to consider whether you are a vitamin or an antibiotic, and which direction to go. At the same time, you need to plan how to play your product.

In the next issue, let’s talk about the pitfalls and mistakes we have made in data products.

Author: Kaka didn't wake up

Source: Kaka didn't wake up