Optimize the allocation of advertising resources
|
After I took over the client, I proposed several solutions... Last year, I led an Internet service client to promote an APP on a certain app store , with the amount of advertising reaching hundreds of thousands per day. The problem that was always a headache at the time was that the media did not allow monitoring of the activation of each ad spot , and it was difficult to judge the quality of each ad spot based solely on the front-end exposure, downloads, and consumption. We knew clearly that conversion rates for certain ad positions were poor, but because the client’s KPI had requirements for quantity, we had no choice but to continue investing based on experience. After I took over the client, I proposed several solutions: Solution 1 Pause 3 of the 6 currently running ad slots, and add the natural volume, for a total of 4 sources of downloads. After a week, pause another 3 of the 6 ad slots, open the paused 3, and so on. The principle is: by reducing the number of variables (i.e., download sources), the difficulty of evaluating the impact of different variables (i.e., download sources) on a certain indicator (i.e., total activation volume ) can be reduced. The idea of this solution is very good, but it is not suitable for the actual situation of this customer: the download sources are highly concentrated. The following figure shows the distribution of the proportion of each download source in the past month. It can be seen that: The three download sources of natural volume, premium products, and essentials for installation have accounted for more than 70%, and other resources can account for about 10%. The test plan of pausing advertising positions can easily cause fluctuations in account traffic, so consider plan 2. Solution 2 Based on the quantitative analysis of advertising data by statistics, we study whether there is a significant correlation between each download source and the total activation volume. The more relevant the download source is, the greater its contribution to the activation volume, that is, the higher the download activation rate is. Start analysis. Here we take IBM SPSS 23 as an example to import the data of download volume and activation volume. The results of the correlation analysis are as follows: The data shows that the correlation between activation volume and different download sources is: Strong correlation (>0.8): Activation-Boutique, Activation-Search, Activation-Natural volume Strong correlation (0.6~0.8): Activation-Ranking, Activation-Other resources, Activation-Red envelope special Generally correlated (0.4~0.6): Activation-Must-have for installation Because there is no obvious third variable (detailed explanation is at the end of the article), the correlation here can be regarded as a causal relationship. In other words, the data shows that the ranking of the download activation rate of the App in the App Store is: Boutique, Search, Natural Volume > Ranking, Other Resources, Red Packet Special > Must-have for Installation Subsequently, we controlled bids, budgets, time periods, etc., with a focus on improving ad placements with better conversion rates. The following figure shows the distribution of the proportion of each download source during the optimization period. It can be seen that: The download share of high-quality products has increased by nearly 10 percentage points, the share of other resources has gradually decreased to 0, and the share of searches has also increased. The proportion of products that are essential for installation has fluctuated greatly. The main reasons are: the traffic of high-quality products is unstable due to activities organized by competitors; KPI has requirements on quantity; and the rate of new customers for products that are essential for installation is relatively high. The following is the most important data, which can prove the significant optimization effect of Solution 2 (activation cost continues to decrease, download activation rate continues to increase) Finally, there are some tips for you to learn more. Correlation does not imply causation! Correlation does not imply causation! Correlation does not imply causation! Let me tell you a classic statistical story about ice cream and crime. In a small town in the Midwest of the United States, people discovered a very interesting illogical phenomenon: the higher the ice cream consumption, the higher the crime rate. At this time, some people were wondering whether the crime rate could be changed by artificially controlling the sales of ice cream. The answer is no. Common sense tells us that ice cream has nothing to do with crime, and the reason there is a statistically positive correlation is because of the weather. The sales of ice cream are closely related to the weather. The hotter the weather, the higher the sales. At the same time, the hotter the weather, the more likely people are to be outdoors and open windows (increasing the probability of theft), women are more likely to wear revealing clothes (leading to an increase in sexual crimes), and people become more irritable (leading to an increase in impulsive crimes). In the story, the two seemingly related variables (ice cream sales) and (crime rate) are highly correlated, but in fact they are both affected by a third variable (weather). In the data analysis above, there is no obvious third variable between the download sources such as premium products and essential installation products and the activation volume, so it can be determined that the correlation between the variables is a causal relationship.Mobile application product promotion service: APP promotion service Qinggua Media advertising The author of this article @齐云疆 is compiled and published by (Qinggua Media). Please indicate the author information and source when reprinting! |
<<: How to write the copy for New Year gift boxes? Look at what Philips wrote.
>>: Where is the Zhihu advertising backend? Where can I connect to the delivery?
Recommend
How to formulate SEM bidding promotion strategy?
The same applies to SEM bidding. Don’t use dilige...
How to run a community well?
In recent years, the concept of community operati...
How to operate new media in the medical and health field?
The essence of creating a doctor's IP is to t...
A practical guide to writing copy for educational information flows to effectively increase click-through rates!
Information flow advertising copy is one of the k...
New policy on rural veterans’ subsidies in 2022: How much per month after age 60?
Soldiers are the defenders of our country. Since ...
Pay attention! To promote your product, you need to have these 5 product selection ideas
A cosmetic product costs a lot of money in every ...
One event converted 100,000 yuan! How can promotion take advantage of traffic to new heights?
Today I want to talk to you about how community f...
Zhao Honglin - A Hundred-part Introduction to the Chaos Theory
Zhao Honglin - Introduction to the resources of t...
Teach you how to write high-quality brainwashing copy
"I won't accept any gifts this year, but...
20 ways to promote and attract traffic on the Internet, save them!
Hello everyone, today I will share with you the “...
The new version of App Store may have a severe blow to the ranking business, but there is still hope
On September 20, Apple officially pushed the iOS ...
The summary of marketing cases in the first half of 2017 is all in these four words
This is the third article in the mid-year review ...
Common activity forms and user attraction logic
There are often some popular event cases on the m...
Analysis of Zhihu content operation and promotion strategy!
While the number of registered users of Zhihu has...
China's APP active user rankings in November 2016 (TOP450)
iiMedia Research, the world's leading third-p...