作者：Vasiliy Sabirov

在上篇文章中我们讨论了付费用户的划分以及RFM分析方法。

而这次我们将基于完全不同的原则继续描述用户划分。你是否考虑过你的收益结构？谁能创造更多收益—-是新用户还是旧用户？新用户和旧用户各自的收入比例是多少？这种情况会随着时间发展发生什么变化？这便是我们将在本文中讨论的内容。

整体的用户结构

首先，我们将根据用户的注册时间将其划分到不同时间段中。如何选择不同部分完全取决于你，即基于你的业务性质以及参与项目的时间段。

不管怎样我们建议最好延伸到5至7个部分。

例如：

第一个部分—-从注册那天起到现在不足14天

第二个部分—-从注册那天起到现在有14至30天；

第三个部分—-从注册那天起到现在有1至2个月；

第四个部分—-从注册那天起到现在有2至6个月；

第五个部分—-从注册那天起到现在有6个月至1年；

第六个部分—-从注册那天起到现在已经超过1年。

通过选择特定部分，你便能够基于时间分析创造一份有关用户结构的报告。

这份报告显示了什么：

如果新用户明显占据主导地位—-你便遭遇了用户留存问题。你的项目不能长久地留住用户。这也意味着你需要着眼于提高用户留存或考虑如何从新用户身上赚取利益。

user retention(from tuicool)

如果旧用户明显占据主导地位—-这也不是什么好事。新注册用户是否出现什么问题？也许是时候购买一些流量？你需要牢记的是你将会迎来更多用户。旧用户是不可能一直支撑着你前进—-迟早你的应用会开始下滑。

也许下一步你将审查你的用户结构和动态—-该结构是如何随着时间的发展发生变化。通常在这个阶段会出现一些最有趣的事。

付费用户的结构

最后，基于同样方式分析收益：通过付费用户的注册时间进行划分。

在关于收益结构的报告中，我们能够更清楚地看到旧玩家和新玩家的利益差。实际上（游戏邦注：在基于长期用户留存的项目中），新玩家的平均消费单价较低，而旧玩家的平均消费单价较高。

就像在我们的例子中，收益呈现下滑趋势（需要清楚的是在这个例子中付费用户的数量是稳定的）。而出现这种情况主要是因为旧玩家所创造的收益减少。

所以关于该项目我们的诊断是，这个项目在注册了3个月以上的付费用户方面存在问题。所以必须优化项目的长期用户留存从而避免最后一部分用户的流失。

数学模拟

基于上面的报告，你将能够创造一个数学模型让自己可以提前几个月预测到收益。

需要什么：

估算每个选择部分的规模；

面向所有部分估算从部分N转向部分N+1的可能性（即用户在一个月内以及在接下来一个月内保持活跃的可能性？）；

评估每个部分中每用户平均收益（ARPU）。

通过结合我们模型中的所有数值，你便能够创造有关你的用户结构和收益在一个月，两个月，三个月以及六个月内如何发生改变的模型。

此外，这一模型将让你能够基于流量和盈利评估各种实验。

它将能够回答这样的问题：

如果我断开付费流渠道而只留下病毒性传播方式的话会怎样？这是否会影响我在12个月内的收益？

如果我优化了2%的用户留存（如30天的用户留存），这将如何影响用户和收益的结构？我将改变游戏平衡并因此提升10%的80级用户的平均消费单价。这时候我的收益比会发生怎样的改变？

通过本文我们想要传达一个简单的理念：根据用户注册时间去了解你的用户和收益结构是非常重要的。这能帮助你做出更有效的决策—-不管是关于市场营销，盈利还是游戏设计。

（本文为游戏邦/gamerboom.com编译，拒绝任何不保留版权的转发，如需转载请联系：游戏邦）

How to analyze paying users. Part 2, The structure of the revenue by time

by Vasiliy Sabirov

Last time we were talking about the segmentation of paying users, reminisced RFM-analysis, as well as whales and dolphins.

This time we will also use segmentation, but on an entirely different principle. Have you ever thought about the structure of your revenue? Who brings more money – the new ones or the old ones? What is the ratio of the revenue from the new and the old users, how it changes over time? This is what we are going to talk about.

The structure of the audience as a whole

At first we divide our entire audience (both paying and not paying) into multiple segments of time from the moment of of their registration. How to select segments – the decision is exclusively yours and depends on the nature of your business and the period of engagement in the project.

Anyway, we recommend to go down to 5-7 segments.

Example:

1st segment – less than 14 days from the moment of registration;

2nd segment – from 14 to 30 days;

3d segment – from 1 to 2 months;

4th segment – from 2 to 6 months;

5th segment – from 6 months to 1 year;

6th segment – more than a year from the moment of registration.

By selecting custom segments, you may build a report on the structure of your audience at the time of analysis.

What does this report show:

If the newcomers clearly dominate – you have a problem with retention. The project can not retain user for a long term. And that means that you have to either work on retention, or think about monetization of the newcomers (for example, make the application paid one).

If the oldies clearly dominate – this is also not good. Is everything OK with the new registrations? May be it is the time to buy a bit of traffic? Remember that the more users, the more users. And oldies do not go far – sooner or later, the app starts to loose rating.

The next step may be to examine not only the structure of your audience but its dynamics – how this structure changed over time. Usually at this stage the most interesting things show up.

The structure of the paying audience

Let’s perform the same manipulations but now only for the paying audience. For example, by report “Users & Gross structure” from devtodev.

This example shows how the stability of the size of your paying audience hides the pitfalls, and the growth of one segment is offset by a decrease of other segments.
We see that the percentage of newcomers (less than 30 days from the moment of registration) is increasing, and the percentage of oldies (6 to 12 months from the moment of registration) decreases. Without the consideration of the structure we would not notice this.

A sign of the healthy application is that the segment of the oldies should be slowly, but growing – more and more users should reach this segment and stay there.

The structure of the revenue

Finally, in a similar manner revenue may be analyzed: by cutting it into segments by the time from the moment of registration of users that make payments.

In the report on the structure of the revenue all distortions for the benefit of oldies and newcomers are usually more vividly pronounced. The fact is that usually (in the projects based on long-term retention) the average check of the newcomers is small, while the average check of the oldies is large enough.

As we see, the revenue in our example has a downward trend (remember that the size of the paying audience in this case was stable). And a decrease in this trend is primarily due to a decrease in revenue from the oldies. Up to the green segment, inclusive, there is some stability, and then decrease occurs.

Our verdict on considered project – the project has problems with payments from users who registered 3 months ago and earlier. It is necessary to optimize the long-term retention of the project so that the natural flow of users in the last segment exceeded the natural outflow.

Mathematical modeling

With the above reports, you will be able to create a mathematical model of predicting your revenue for a few months in advance.

What is needed:

estimate the size of each of the selected segments;

for all segments calculate the probability of transition from the segment N to the segment N+1 (what is the probability of user being active during the month, remains active in the next month?);

calculate the average revenue per user (ARPU) of each segment.

By combining all calculated values in one model, you will be able to model how the structure of your audience and of revenue will change in a month, two, three, six.

Furthermore, this model will allow you to calculate the various experiments with traffic and monetization.

Examples of the questions it will be able to answer:

What if I disconnect the channel of paid traffic and remain only on the virality? How will this affect my revenue in 12 months?

What if I optimize retention (eg, 30 days retention) by 2%, how it will affect the structure of the audience and revenue?
I’m going to make a change in the balance of the game and thus raise the average check of user of 80’s level (which is reached after an average of six months of the game) by 10%. By what percentage my revenue will change?

And so on.

By this article we would like to convey to you one simple idea: it is important to study the structure of your audience and revenue by time from the moment of users registration. This will help you to make more informed and effective decisions, whether it’s marketing, monetization or game design.（source：gamasutra）