• Predicting your App’s Monetization Future
  • Ignacio Monereo
  • The Nuggets translation Project
  • Permanent link to this article: github.com/xitu/gold-m…
  • Translator: PTHFLY
  • Proofread by Wangalan30, realYukiko

This article will teach you how to predict the future liquidity of your app

Introduction to predictive analysis and user life cycle value calculation

We all want a magic crystal ball that will reveal how our app will perform in the future: how many users it will attract and how much revenue it will generate. Unfortunately, there is no such crystal ball. But the good news is that we have technology that can give you insight into how your APP will perform in the future and help you build a reasonable and effective revenue strategy.

This is the first of two articles on exploring LTV (Lifetime Value). In this article, I will introduce predictive analysis, point out a simple formula for calculating LTV, and explain how to obtain a value that can be used as a plan.

In the next post, I’ll explore how this formula applies to the monetization strategies of five popular apps, as well as provide some insights from developers on how to optimize these monetization strategies.

Predictive analysis

If you want to know the future of your APP, you can make predictions by looking at the data you collect from your users. The main content of predictive analysis is to extract information from these data by applying various statistical techniques.

Predictive models are used in many businesses, and they can help answer key business management questions: How many paying customers will we have next year? What is the expected transaction amount for next year? When will users leave our service?

These models are widely studied offline. Thanks to the digital revolution, we’re seeing it become more common, driven by improved capabilities to collect, integrate and categorize user data.

In the world of mobile apps, game developers are power users of these technologies, and their use has a positive impact on the monetization of their apps.

From empiricism to mathematical modeling

The way to predict how many users there will be and how they will pay in the future will be very different. Two of the most unusual are:

  • Simple models based on professional experience or benchmarking. For example, a company hires an outside consultant to consult and forecast next year’s sales based on the consultant’s market and industry knowledge.
  • Complex mathematical models, such as Pareto/NBD models (see Counting Your Customers by David C. Schmittlein, Donald G. Morrison, and Richard Colombo: Who Are They and What Will They Do Next? . These models take into account multiple variables, including recency (recent purchase) and frequency or number of orders within a given time, to calculate the likelihood of a customer repurchasing.

For mathematical modeling methods, there are some online resources that can help you calculate, For example, the approach described by Bruce G. S. Hardie in the Implementing the BG/NBD Model for Customer Base Analysis in Excel.

Analyze required data and tools

One of the main drivers of the development of predictive analytics is the increasing use of analytical tools. Powerful analytics tools allow us to collect, collate, and integrate data more efficiently, while quickly sharing it with key stakeholders and decision makers.

Some of the most important app metrics are:

  • Users get data: number of installs, number of uninstalls, and incoming channels.
  • Retention data: Retention (1, 7, 28, 90, 180, 365 days after download).
  • Realisation data: number of paying users, recent purchases, transaction frequency, total purchases, attrition rate and new repeat purchases.

The importance of these variables depends on their calculated weights and the following factors:

  • Application environment: Activation and usage of different types of apps can vary greatly.
  • Business model: Depending on which business model you choose, one key piece of data can have a huge impact. Subscription business models, for example, often care about when subscribers renew.
  • Means of use: Some methods will require certain data to build future predictions. For example, use BG/NBD to request frequency and recent purchase data.

It’s important to note that a powerful Analytics tool, such as Google Analytics for Firebase for in-app Analytics or The Google Play Console for monetization, not only needs to be able to collect this information accurately and timely, Also be able to process and share this information quickly. This is key to responding quickly in a dynamic environment such as a digital ecosystem.

background

Predictive analysis relies heavily on historical data of users and buyers. While this is a good starting point, don’t ignore external information that may influence future projections. This may include the company’s stage of development, technology trends and the macroeconomic environment.

Especially when analyzing the lifecycle value of an app, it can be useful to adjust some key points based on environmental factors such as the trading environment and contractual obligations. For example, a retail app might have to choose between buyers within a month of activation and buyers within a year.

There are several different frameworks to help identify these background factors. In this context, I found a framework pointed out in the article Probability Models for Customer-base Analysis by Peter S. Fader and Bruce G. S. Hardie very useful.

Fader and Hardie’s model differentiated user behavior based on whether users entered into contracts with companies and whether transactions were continuous or discrete. Consider an example of this model:

The top left corner is a company that has no contractual relationship with its users to create a continuous flow of transactions. A practical example is e-commerce applications, where consumers trade repeatedly but can leave and leave at any time.

At the bottom right is an example of the opposite: a company contracts with a user, and transactions occur only at certain times. A practical example is that consumers tend to buy life insurance policies at a specific point in time (such as when they get their first job), which will remain in effect for the duration of the premium.

Life cycle value

One of the last popular indicators of predictive analytics is user life Cycle Value (LTV), one of the most popular predictive analytics metrics, which is a user’s estimate of the economic value of their business over their lifetime. This metric is well known offline and is widely used in the app and game industries.

Because it provides estimates of the potential revenue per user, LTV is useful because it provides a reading of the potential revenue per customer. In turn, this can help determine acquisition costs and analyze which channels, platforms, and user distributions are the most cost-effective.

However, there are some common LTV pitfalls to avoid before we get into computing issues. So please don’t:

  • Use LTV as a target and spend resources optimizing and scaling it up. Just think of LTV as a tool that will improve as other metrics such as contracts, retention, and liquidity grow.
  • Create overly optimistic LTV. For example, a startup may overestimate revenue per user, resulting in LTV inflation that may cost more to acquire users.
  • Allow user acquisition (CAC, the Cost of acquiring a customer) to exceed LTV. Although factors to consider (company stage: startup or maturity; Relationship type: contract or not) is a lot, but an industry rule of thumb tells us that the cost of user acquisition should not exceed the net LTV of an app. Yet many companies prescribe a 3:1 LTV to CAC ratio (CAC will never exceed 33% of net LTV).
  • View high LTV as a competitive advantage. In a rapidly changing industry, such as technology, it’s easy to find an example of high liquidity, high LTV but rapidly losing market share as the technology is phased out and users migrate to newer, more attractive apps.

Calculate the LTV

There are several ways to calculate the LTV of apps and games. These methods vary depending on the complexity of the business model, the data available, and the precision required.

To start, let’s use the following simple formula:

LTV (given phase) = life cycle x ARPU (Average revenue per user)

Now, let’s examine the next variable more closely:

A. LENGTH of LTV time

Most developers calculate LTV in 180 days, a year, two years, or five years. Factors that determine the length of LTV may include average user lifetime or business model-based choices.

For example, imagine a developer using the in-app purchase model. The average user engagement is 15 months. In this case, the two-year LTV will be higher than the one-year LTV. However, a one-year LTV is a more conservative option because the average life cycle (15 months) is longer than the chosen cycle (12 months).

The length of LTV time should be considered as follows:

  • The business environment, for example, may be more profitable for some liquidating models (particularly subscription ones) and, if the incentives for liquidating are right, will continue to grow over a long period of time, which can justify long-term LTV. Telecom companies, for example, have traditionally invested heavily in user acquisition and even subsidized hardware in anticipation of a long life cycle.
  • Business models, for example, for certain monetization models (especially subscriptions) may have higher revenue and longer term growth if the monetization incentives are right (churn is negative), it makes sense to use long term LTV. Telecom companies, for example, have traditionally invested heavily in user acquisition, even subsidizing hardware that requires a very long life cycle.
  • Company stages, such as early comparative maturity. Because of technological advances or lack of historical data, early-stage companies often choose longer, more optimistic periods to calculate LTV. On the other hand, a mature company with outdated technology may want to opt for shorter LTV computing times.

B. Life cycle

The lifecycle is directly related to activation and retention. In turn, these two concepts will help increase user retention and increase the likelihood that they will drive monetization. App developers usually calculate the lifecycle based on app retention.

A simple way to estimate retention is to call a user’s in-app “churn” moment when they haven’t opened an app in the past month. Thus, the average attrition time of at least one month since users stopped using the APP can be calculated.

A more accurate way to calculate the lifecycle is to use the survival curve model: a declining equation based on historical usage data (one curve per user or user group). The total or average retention of each component can be calculated, and the equation for a given period can be solved.

Looking at the following example, after counting all users, the probability of a user remaining active after 180 days is only 23%. Therefore, the average user lifetime per 180 users would be 180 x 23%, or nearly 41 days.

One important caveat here is that the life cycle is always in the same unit as the LTV time period. For example, a 180-day LTV would be based on a 41-day expected life cycle, not months or years.

C. ARPU or average revenue per user

The difficulty of calculating ARPU varies from business model to business model. A SaaS model is simpler and a hybrid model is more complex (mixing different business models, such as subscriptions and advertising).

One way to calculate ARPU would be to split the total revenue over a period of time by active users during that time. For example, if the average daily revenue of $10,000 is split by 25,000 daily active users, the ARPU will be $0.4 / day.

I can now calculate LTV for this application. Life cycle within 180 days was 41 days (23%) and ARPU was usd 0.4 / day. Therefore:

180 days LTV = 41 days x $0.4 / day = $16.40 / user

Optimize LTV calculation

There are several techniques that can be combined with this simple LTV equation to improve usability, including:

  • Discount Revenue cash flows. Consider discounting future information flows by taking into account either the rate of inflation (R) or the cost of capital (such as calculating the average cost of capital ratio, which measures the average cost of capital) when the life cycle exceeds one year. For example, assuming a life cycle of N years, the discount formula appears as follows:

LTV = Revs Year 1 + Revs Year 2 x 1/ (1+ r) + … + Revs Year n x 1 / (1+ r )^(n-1)

  • Calculate net LTV.Net profit can be calculated by calculating average variable profit (VC) per user and replacing ARPU in the formula. To estimate VC, the total variable costs need to be deducted from the total revenue. Variable costs are incurred when each new user joins the app (such as marketing fees allocated to each user). The new formula would look something like this:

Net LTV = life cycle x VC

On the basis of:

VC = Total revenue over time (total variable costs)/average users over time

Another important thing to note is that savvy developers will often differentiate users based on VC level and calculate LTV for different user groups. As is often the case in many businesses, app developers will observe that a small group of users brings in the most revenue and profit.

Because variable costs tend to decrease as a percentage of revenue, VC changes frequently over the lifetime of the user. For example, compare a new user who recently subscribed to a software service and needs more customer support during its use with an experienced customer who no longer needs support.

conclusion

Predictive analytics provides a practical way to predict your app’s future performance: its users and revenue. Among these predictive analytics methods, life cycle value (LTV) is probably the one metric that has been gaining popularity among APP developers lately. It is very simple and provides a useful approach that can be applied to visitor planning.

Now that you know a little bit about LTV, in the second post I’ll examine how the LTV formula applies to five popular app monetization strategies. I’ll also offer some insights from developers on how to optimize these monetization strategies.

What do you think?

Do you have any questions or thoughts about predictive analysis and LTV? Continue the discussion in the comments section below or let us know with the hashtag #AskPlayDev, and we’ll respond to it at @Googleplaydev (where we regularly share news and tips on how to succeed on GooglePlay).

The Nuggets Translation Project is a community that translates quality Internet technical articles from English sharing articles on nuggets. The content covers Android, iOS, front-end, back-end, blockchain, products, design, artificial intelligence and other fields. If you want to see more high-quality translation, please continue to pay attention to the Translation plan of Digging Gold, the official Weibo, Zhihu column.