Tracking Unique Active Users for App Growth

Facebook continues to use Daily Active Users (DAU) and Monthly Active Users (MAU) to monitor and report growth. The company had reported 1.18 billion daily active users on average for September 2016 and 1.79 billion MAU for the same month. Mobile growth has been significant and was reported as 1.09 billion DAU (average for September 2016) and 1.66 billion MAU. 

Measuring unique active users over a period helps web and mobile app publishers monitor and report on growth and compare growth across channels (mobile versus total, in the above example). The period could be a specific day, specific month, specific week, specific hour, last seven days, last 30 days, today, yesterday, last 24 hours, and so on.

Daily Active Users (DAU)

Daily Active Users (DAU) is the number of unique active users of your app on a specific day.

You can report DAU for a specific day (i.e., December 7, 2016) or as an average of days in a quarter. For instance, Facebook reports DAU as an average over the number of days in the reporting quarter.

Monthly Active Users (MAU)

Monthly Active Users (MAU) is the number of unique active users of your app in a specific month. The month could be a specific month, i.e., “February,” or the preferred “previous 30 days,” which excludes today.

It is much harder to manage unique users for an app across “previous 30 days,” but the data is much more reliable and not calculated as an estimate.

Weekly Active Users (WAU)

Weekly Active Users (WAU) is the number of unique active users of your app in a specific week. The week could be a specific week number with Monday as the starting day or the preferred “previous seven days” which excludes today.

A month is a lifetime in an app’s lifecycle. Most product managers look at DAU, WAU, and MAU to catch and act on changes early.  

Facebook monitors but does not report WAU.

MAU and WAU vs. DAU

It is important to state that WAU and MAU are not an aggregation of daily active users. Many analytics packages double and triple count daily active users. The right way to count WAU and MAU numbers is by not counting the same user multiple times. This is how Facebook and LinkedIn report their DAU and MAU numbers. Note that MAU is always less than or equal to DAU.

This is best explained with an example. Consider that the following users visited a website or used an app. It may seem obvious, but be sure to note that Sam used the website or app three times on Tuesday the 9th.

Calculating DAU, MAU, and WAU

• Three unique users used the app or visited the website on Monday the 25th — that is, Sam, Sarah, and Kim. The DAU is 3.
• Kim and Sarah were the two unique users on Wednesday the 10th. The DAU is 2 for that day.
• For Week A, the unique users were Sam, Sarah, Kim, Nick, and Adrian. That makes 5, assuming the week starts on Monday.
• For Month N, the unique users were Sam, Sarah, Kim, Dickey, Charlie, and Michelle. The MAU for Month N is 6.
• On Monday the 1st, the last 7 D active users would be the number of active users in period Tuesday the 26th to Monday the 1st. This is 5, i.e., Sam, Sarah, Kim, Nick, and Adrian. The example assumes zero users from 27th to 31st for simplicity.

Typically, app publishers and growth marketers always look at the following numbers on a daily basis.

• Yesterday active users: The number of unique active users of your app yesterday.
• Today active users: The number of unique active users of your app today as of now.
• Previous 7 days active users: The number of unique active users of your app in the last 7 days, excluding today.
• Previous week active users: The number of unique active users of your app in the previous week.
• This week active users: The number of unique active users of your app this week as of now.
• Previous 30 days active users: The number of unique active users of your app in the last 30 days, excluding today.
• Previous month active users: The number of unique active users of your app previous month.
• This month active users: The number of unique active users of your app this month as of now.

Choosing the Right Mean

The various metrics (installs, activations, active users, sessions, usage, and retention) can be compared against mean, median, mid-range, tri-mean, and the winsorized mean along with the deviation. Winsorized mean and tri-mean are comparative statistics that take temporary surges out of the math.

• Mean (average): In statistics, the arithmetic mean, or the mean or average, is the sum of a series of numbers divided by the number of numbers in the series.
• Median: Median is either the middle value of an odd dataset or the average of two middle values of an even dataset.
• Mid-range: Median is a measure of central tendency and hints at the middle. It ignores the entire data set, except for the smallest and largest values. It is calculated as an average of smallest and largest values in the dataset.
• Tri-Mean: Tri-mean is a better estimator and measure of the center of data in a way that it is resistant to skewing outliers. By incorporating the first and third quartiles in the calculation of the tri-mean, we include some information about the rest of the distribution of the set of data, as opposed to the median that always drops non-central data and the mean that always includes it.
• Winsorized Mean: When data is skewed or has outliers, an obvious technique most analysts use is to remove or trim out the outliers. Mean with trimmed data is known as trimmed mean. In a 10% trimmed mean, the largest and smallest 10% of the values are removed, and then the mean is taken on the remaining 80%. Note that median is a type of trimmed mean; it is simply the 50% trimmed mean. Winsorized mean is similar to the trimmed mean, except that rather than deleting the extreme values, it is set equal to the next largest (or smallest) value, preserving the divisor count in the mean.  Product Managers prefer winsorized mean over trimmed mean to eliminate outlier data.

Hourly Active Users (HAU)

Hourly Active Users (HAU) involves the number of unique active users by the hour.

In an analysis of hourly active users compared with the tri-mean (an advanced mean that ignored skewness) along with high and low standard deviations, represented by dotted horizontal lines.

We have found two useful scenarios for hourly active users:

• Aggregating by the hour of the day and day of the week gives you an idea of when to manage server upgrades for apps with a global audience.
• Aggregating by the hour of the day and day of the week and across time zones gives you an idea what time your app has the most active users globally. In the chart below, most people use the Best Reads book reader app around 8 p.m. to 9 p.m. in their timezone and Thursday is the most popular day of the week.

Stickiness Factor

Stickiness is a ratio of DAU and MAU and indicates the depth of engagement. The closer the ratio of DAU and MAU to 1, the better the stickiness. This means that more of your users are returning to the site every day.

In overly simplistic terms:

Stickiness = (Daily Active User for a specific day) / (unique and active users in a month or week)

However, here are some enhancements:

• The numerator should not be the active users for the current day since active users increase over the duration of the day and you may be running install campaigns that run at a specific time of the day making your DAU sporadic.
• The denominator should not be calculated for the previous past whole month, so you are using a more recent and accurate number as a divisor.Also using previous month introduces inconsistencies for months with 28, 29, 30 or 31 days. It is much harder to manage unique users for an app across previous 7 and 30 days but the Stickiness Factor is much more reliable, not an estimate, and does not deteriorate at month end. For instance, it does not make much sense to use Feburary’s MAU number as a divisor on March 26th.

7D Stickiness Factor

The 7-day stickiness factor is a ratio of yesterday’s active users versus unique and active users in the previous 7 days (excluding today). In general:

7D Stickiness Factor = (Daily Active User for a specific day) / (Previous 7D unique and active users).

30D Stickiness Factor

30 Day Stickiness Factor is ratio of yesterday’s active users  versus unique and active users in the previous 30 days (excluding today).  In general,

30D Stickiness Factor = (Daily Active User for a specific day) / (Previous 30D unique and active users).

Measuring change in Stickiness Factor is useful when you make a change to your app with your hypothesis being that it’s going to increase engagement and retention.

Stickiness Factor Over Days

It is well observed that implementing push notifications for retention marketing (i.e. to retain users) and in-app messages for engagement marketing (i.e. to engage active users) increases both retention and engagement considerably.  Stickiness Factor can help determine the factor of growth.

Conclusion

• DAU is used for reporting growth and comparing growth across channels.
• Stickiness factor is a ratio of daily active users over monthly or weekly active users and is useful to indirectly track growth.
• Metrics have to be calculated correctly for you to rely on them.
• Most analytics packages are configured by default to provide mostly reports on vanity metrics.
• Activations is a much better alternative to installs, as it is actionable.
• Loyalty is a much better alternative to daily active users, as it indicates how loyal users are.

I will be covering vanity metrics, activations, and loyalty in the next blog on DZone.