Mobile APP ROI Estimation Calculation Model
GOOGLE DEVELOPER
Suspendisse faucibus mauris massa, ut scelerisque est ultricies nec. Aliquam quis dictum tortor. Nunc tincidunt gravida ornare. Sed pulvinar faucibus dolor ac efficitur.
1. The role of ROI calculation
When we establish a product project and formulate a promotion plan, we need to calculate the estimated product profit margin, which is also what we often call ROI (Return On Investment). ROI is basically the direction indicator of the entire team and the fundamental guide for the decision-making of each product team. The delivery team uses it to determine the delivery plan and price setting; commercialization adjusts the aggressiveness of advertising realization and in-app purchases based on it, and operations can plan active activity plans based on it.
2. ROI calculation method
To calculate whether the recovery ROI can make a profit, it is also an important reference indicator for the boss to decide whether to start and continue the project. There are two methods that people usually use in the market:
Image: https://uploader.shimo.im/f/mswNlgLakiR1xJg6.png!thumbnail?accessToken=eyJhbGciOiJIUzI1NiIsImtpZCI6ImRlZmF1bHQiLCJ0eXAiOiJKV1QifQ.eyJleHAiOjE3MTc4NjE3NDcsImZpbGVHVUlEIjoiMTZxOHhaOHdlSmkyUTFxNyIsImlhdCI6 MTcxNzg2MTQ0NywiaXNzIjoidXBsb2FkZXJfYWNjZXNzX3Jlc291cmNlIiwidXNlcklkIjoxNjI0MDA2NX0.vm_wPqZOtEnS4MHWRHn6HtTtuCK6s-Kfm0GxH-ewIRM
Profit level within a cycle. This calculation method is suitable for products that can be quickly recovered within a short period of time, or for old products that have been online for a long time. For example, the total revenue to total cost within 1 day or 1 week.
The second method is to calculate the revenue value generated by each user in his life cycle. It must be higher than the installation cost of each user. Only then can UA marketing promotion be profitable. LTV is the abbreviation of the three English initials of "Life Time Value". The Chinese translation is "user life cycle value, the revenue that users will generate in their lifetime, and LTV is a predicted value.
Image: https://uploader.shimo.im/f/KNjhIA3KYqYSZ5JV.png!thumbnail?accessToken=eyJhbGciOiJIUzI1NiIsImtpZCI6ImRlZmF1bHQiLCJ0eXAiOiJKV1QifQ.eyJleHAiOjE3MTc4NjE3NDcsImZpbGVHVUlEIjoi MTZxOHhaOHdlSmkyUTFxNyIsImlhdCI6MTcxNzg2MTQ0NywiaXNzIjoidXBsb2FkZXJfYWNjZXNzX3Jlc291cmNlIiwidXNlcklkIjoxNjI0MDA2NX0.vm_wPqZOtEnS4MHWRHn6HtTtuCK6s-Kfm0GxH-ewIRM
For example, when we say LTV 30, we mean the revenue that users are expected to contribute to your application within 30 days. The LTV indicator is the most important in all commercial work.
Let's focus on the second ROI calculation method, because this is more commonly used and more difficult to calculate accurately. Most products have a long recovery cycle, and developers cannot really wait until the user's life cycle ends before calculating ROI. Therefore, after about a week of advertising, it is necessary to calculate and predict the future recovery of various channels in key countries.
3. CPI: Cost per installation
Image: https://uploader.shimo.im/f/h2tQSW8IY62lAyaX.png!thumbnail?accessToken=eyJhbGciOiJIUzI1NiIsImtpZCI6ImRlZmF1bHQiLCJ0eXAiOiJKV1QifQ.eyJleHAiOjE3MTc4NjE3NDcsImZpbGVHVUlEIjoiMTZ xOHhaOHdlSmkyUTFxNyIsImlhdCI6MTcxNzg2MTQ0NywiaXNzIjoidXBsb2FkZXJfYWNjZXNzX3Jlc291cmNlIiwidXNlcklkIjoxNjI0MDA2NX0.vm_wPqZOtEnS4MHWRHn6HtTtuCK6s-Kfm0GxH-ewIRM
First is CPI, which stands for cost per install. This is a direct result on Google Ads or user acquisition platforms like Facebook and Mintegral or any other platform.
CPI is the most critical metric for evaluating UA campaigns. It means how much your user acquisition campaigns cost for one install.
CPA stands for cost per action. CPA is similar to CPI, but here you pay for actions taken by the user. For example, an event can be something like a person who made at least one in-app purchase or a user who reached level 5.
4. LTV: Lifetime Value
What is Lifetime Value? For example, WeChat Reading gains a new user, who recharges 50 yuan on the same day; one month later, the user recharges another 50 yuan; two months later, the user spends another 80 yuan; then this user contributes a total of 180 yuan in revenue to WeChat Reading. 180 yuan is the 90-day lifecycle value that WeChat Reading obtains from this user. From the definition of LTV, it can be seen that it consists of two parts, "lifecycle retention" and "contribution value", so it is a long-term indicator. It combines the two dimensions of "retention" and "value" to help developers or companies measure the true value of users from a long-term perspective, so it is more comprehensive and scientific.
Image: https://uploader.shimo.im/f/CFdVO3ZliJMeXC1M.png!thumbnail?accessToken=eyJhbGciOiJIUzI1NiIsImtpZCI6ImRlZmF1bHQiLCJ0eXAiOiJKV1QifQ.eyJleHAiOjE3MTc4NjE3NDcsImZpbGVHVUlEIjoiMTZxOHhaOHdlS mkyUTFxNyIsImlhdCI6MTcxNzg2MTQ0NywiaXNzIjoidXBsb2FkZXJfYWNjZXNzX3Jlc291cmNlIiwidXNlcklkIjoxNjI0MDA2NX0.vm_wPqZOtEnS4MHWRHn6HtTtuCK6s-Kfm0GxH-ewIRM
For example, if a game has an ARPU of $0.5 and a LT of 5, then LTV = 0.5*5 = $2.5; ARPU (Average Revenue Per User) is the average revenue per user, which can be considered as the revenue brought by each user within the time window.
In the actual work process, LTV data needs to be predicted more. As mentioned earlier, developers cannot really wait until the end of the user life cycle to calculate the recovery. After the product is launched, data can be collected for a period of time, and then it can be calculated and estimated.
How to make a prediction? The prediction is divided into three specific steps.
1. Predict LT
LT is Life time, i.e., life cycle. For example, according to monthly statistics, it means the cumulative number of active days that players stay in the game in the next month.
The LT curve and retention rate curve basically conform to the law of power function. Through this graph, we can find the changes of users in the entire life cycle. As time accumulates, LT gradually accumulates and increases; but the subsequent LT growth slows down significantly, and there is almost no growth in the end.
Image: https://uploader.shimo.im/f/5YgX4m7vLzCAfVMO.png!thumbnail?accessToken=eyJhbGciOiJIUzI1NiIsImtpZCI6ImRlZmF1bHQiLCJ0eXAiOiJKV1QifQ.eyJleHAiOjE3MTc4NjE3NDcsImZpbGVHVUlEIjoiMTZxOHhaOHdl SmkyUTFxNyIsImlhdCI6MTcxNzg2MTQ0NywiaXNzIjoidXBsb2FkZXJfYWNjZXNzX3Jlc291cmNlIiwidXNlcklkIjoxNjI0MDA2NX0.vm_wPqZOtEnS4MHWRHn6HtTtuCK6s-Kfm0GxH-ewIRM
LT is predicted by the retention rate curve. LT is essentially the embodiment of user retention. The calculation formula is: LT = 1 + rr_1 + rr_2 + rr_3 + … + rr_n. Among them, rr_n can represent the retention rate on the nth day. (rr is the abbreviation of retention rate)
For example:
For a game APP, the retention rate on the first day is 40%, and the retention rates on the second, third, fourth, fifth, sixth and seventh days are 32%, 30%, 28%, 25%, 22% and 20% respectively. Then LT = 1 + 40% + 32% + 30% + 28% + 25% + 22% + 20% = 2.97 days. It can be written as LT_7 = 2.97 days, which means that the user's life cycle within 7 days is 2.97 days.
We use the retention rate of the previous period to predict the future retention R(t) through data fitting.
(1) Record the actual known retention number in the Excel table, insert and select the scatter plot, and get the following chart. We can obtain the retention rate through the Firebase platform.
Image: https://uploader.shimo.im/f/9pGR2TZ5Ve1rHjRD.png!thumbnail?accessToken=eyJhbGciOiJIUzI1NiIsImtpZCI6ImRlZmF1bHQiLCJ0eXAiOiJKV1QifQ.eyJleHAiOjE3MTc4NjE3NDcsImZpbGVHVUlEIjoiMTZxOHhaOHdlSmkyUTFxNyIsI mlhdCI6MTcxNzg2MTQ0NywiaXNzIjoidXBsb2FkZXJfYWNjZXNzX3Jlc291cmNlIiwidXNlcklkIjoxNjI0MDA2NX0.vm_wPqZOtEnS4MHWRHn6HtTtuCK6s-Kfm0GxH-ewIRM
(2) Then right-click the point chart and select Trend Line. The fitted curve type (exponential, linear, logarithmic, polynomial, power, moving average) will appear on the right side of the Excel table.
Image: https://uploader.shimo.im/f/cM5IWjxaPzRVLMMS.png!thumbnail?accessToken=eyJhbGciOiJIUzI1NiIsImtpZCI6ImRlZmF1bHQiLCJ0eXAiOiJKV1QifQ.eyJleHAiOjE3MTc4NjE3NDcsImZpbGVHVUlEIjoiMTZxOHhaOHdlSm kyUTFxNyIsImlhdCI6MTcxNzg2MTQ0NywiaXNzIjoidXBsb2FkZXJfYWNjZXNzX3Jlc291cmNlIiwidXNlcklkIjoxNjI0MDA2NX0.vm_wPqZOtEnS4MHWRHn6HtTtuCK6s-Kfm0GxH-ewIRM
(3) Check the Display and R-squared items. Among these curve types, select the one with the R-squared closest to 1. The R-squared value is a number that indicates the degree of correspondence between the trend line and the data. The closer the R-squared value is to 1, the better the trend line fits. This means that the function with the highest degree of fit has been found as the retention function R(t). The displayed value is 0.9909. This is a good fit. Generally, values above 0.75 are generally considered to be a good value, usually exponential or power.
The power I use here is y = 0.4105x-0.367. Substitute the formula into the prediction column. y represents the retention rate to be predicted, x is the number of days to be predicted, and -0.367 is the exponent. The characteristic of the power function is that the exponent determines the growth mode of the function. When the current value is negative, it means that the function shows exponential decay.
Image: https://uploader.shimo.im/f/6K9YxLOysDW20Shl.png!thumbnail?accessToken=eyJhbGciOiJIUzI1NiIsImtpZCI6ImRlZmF1bHQiLCJ0eXAiOiJKV1QifQ.eyJleHAiOjE3MTc4NjE3NDcsImZpbGVHVUlEIjoiMTZxOHhaOHd lSmkyUTFxNyIsImlhdCI6MTcxNzg2MTQ0NywiaXNzIjoidXBsb2FkZXJfYWNjZXNzX3Jlc291cmNlIiwidXNlcklkIjoxNjI0MDA2NX0.vm_wPqZOtEnS4MHWRHn6HtTtuCK6s-Kfm0GxH-ewIRM
After predicting the retention rate of the future cycle, the LT value can be obtained according to the LT calculation formula.
If you don't like Excel, here is a very useful online curve fitting tool: