Alex Friedman
Where to hear and read about Alex Friedman1 episode or test
Alex Friedman speaks for Flurry, a mobile analytics platform that tracks 1.2 billion monthly unique users across 500,000 apps. The lesson most worth taking from her is to instrument your app from day one, so you can find your most valuable users and judge every acquisition channel by the quality of users it brings, not by installs.
apps that can maintain over 50% of their users past month four have a much longer shelf life
Flurry: The Key Metrics You Should Be Tracking in Your Mobile App
Alex Friedman: the story and the lessons
1Alex Friedman and the scale of Flurry
Friedman speaks for an analytics platform that sees most of the world's phones. Flurry reaches 90% of smartphones and tablets and tracks 1.2 billion monthly unique users across 500,000 apps.
About 155,000 developers use it, from one-person shops to the largest brands. Shazam and Smule are among its clients.
Flurry records 1.3 trillion events a month with about 120 people. An event is any action in an app, such as a purchase or reaching the next level.
The small team is why the Unity, Corona, Titanium and Cocos2D modules are mostly community supported.
2Friedman's three steps: know your data, find your best users, measure success
She frames analytics as three steps that each build on the last. The plan runs from knowing your data, to finding your best users, to measuring success with tools that keep capturing those users.
A valuable user may be one who pays, or one who brings in more people. The data set up in step one is what reveals them.
She calls the heaviest persona users the whale users. Flurry has over 40 personas, such as business travelers, new moms and hardcore gamers, built from baskets of apps across over a billion devices. A user joins one only with very high recent usage, for example more than 25 times the average sessions in sports apps.
Advertisers use persona data to learn users' interests and to raise return on marketing spend.
Friedman judges acquisition channels by user quality, not installs. In her example one campaign had 683,000 clicks and 15,000 installs, a 2.3% install rate. The web campaign converted best into quality users, so spend moves away from weaker campaigns.
Quality is defined per campaign, such as reaching level three, next to built-in 3-day and 7-day retention. "just getting to the install is one thing getting beyond the install is something else entirely"
She treats the definition of a valuable user as a hypothesis. You may assume users who reach level three are the valuable ones, but the data may show it is those who reach level five.
Goals and progress need ongoing review, not a one-time setup. The aim is "letting the data tell you how to build further into your app".
3What Friedman says to measure
Start with new, active and retained users and sessions. These come built in once the SDK is added. A drop in sessions after a release can signal a new crash, and a jump in new users after a feature can signal a hit.
Tie custom events to business goals and use parameters. A game logs completed levels, and an app with a social side logs Facebook shares. Each event can carry up to 10 string parameters, such as episode number and watch duration, and the platform allows 300 events in total.
Reusing one event with parameters keeps you under the cap. A tutorial step designed to take one minute but taking five points to a confusing step.
Use funnels to find the exact step where users quit. In the hotel app example, 75% saw a hotel list, 35% selected one, 20% entered payment and 17% got confirmed.
The question to chase is where the last 3% went between payment and confirmation.
Use return rate by cohort and aim to keep over 50% past month four. Flurry compared a cross-section of apps and found those holding over 50% of users past month four had a much longer shelf life.
Static return rate is the metric many of Flurry's clients rely on most.
4Friedman's advice on starting early and getting set up
Track from day one, even if you do not look at the data. Collect data passively in the first release. If you do not look for six months, so be it, but the history is there when a question comes up.
Without it you iterate blind. Friedman and Mike Rollins gave this as their top advice, and the host admitted making this mistake.
Plan tracked events before launch, from mockups. Map key steps the way you would map pages to track on a website. Teams that add analytics two days before launch get less value.
Mike Rollins says this planning also improves product design, because it forces the team to name the points that matter most.
Basic integration takes about five minutes. Adding the SDK is a few lines of code and needs no servers on your side. A full advanced setup takes a few days, or about an hour with a roadmap and clear business goals.
Verify the setup by logging an event and checking event logs. Event logs show entries within 5 to 10 minutes, while dashboard data lags by three to six hours because of batch processing.
Checking every hour does not help, since there is no fixed schedule.
Friedman gets app ideas from daily frustrations. Her own idea is a "Tinder for Manhattan apartments" for her housing hunt, and she uses the Tayasui Sketches drawing app because it is fun and relaxing.
Any app could be the next hit, so you need a view into it.
Flurry: The Key Metrics You Should Be Tracking in Your Mobile App
Covers which mobile analytics to instrument from the start, how to find your most valuable users with segments, personas and funnels, and how to judge acquisition channels by user quality instead of installs. Worth hearing for concrete limits and benchmarks, such as a 5 minute basic integration and the finding that apps keeping over 50% of users past month four have a longer shelf life.
Friedman speaks for an analytics platform that sees most of the world's phones. Flurry reaches 90% of smartphones and tablets and tracks 1.2 billion monthly unique users across 500,000 apps.
About 155,000 developers use it, from one-person shops to the largest brands. Shazam and Smule are among its clients.
Flurry records 1.3 trillion events a month with about 120 people. An event is any action in an app, such as a purchase or reaching the next level.
The small team is why the Unity, Corona, Titanium and Cocos2D modules are mostly community supported.
She frames analytics as three steps that each build on the last. The plan runs from knowing your data, to finding your best users, to measuring success with tools that keep capturing those users.
A valuable user may be one who pays, or one who brings in more people. The data set up in step one is what reveals them.
She calls the heaviest persona users the whale users. Flurry has over 40 personas, such as business travelers, new moms and hardcore gamers, built from baskets of apps across over a billion devices. A user joins one only with very high recent usage, for example more than 25 times the average sessions in sports apps.
Advertisers use persona data to learn users' interests and to raise return on marketing spend.
Friedman judges acquisition channels by user quality, not installs. In her example one campaign had 683,000 clicks and 15,000 installs, a 2.3% install rate. The web campaign converted best into quality users, so spend moves away from weaker campaigns.
Quality is defined per campaign, such as reaching level three, next to built-in 3-day and 7-day retention. "just getting to the install is one thing getting beyond the install is something else entirely"
She treats the definition of a valuable user as a hypothesis. You may assume users who reach level three are the valuable ones, but the data may show it is those who reach level five.
Goals and progress need ongoing review, not a one-time setup. The aim is "letting the data tell you how to build further into your app".
Start with new, active and retained users and sessions. These come built in once the SDK is added. A drop in sessions after a release can signal a new crash, and a jump in new users after a feature can signal a hit.
Tie custom events to business goals and use parameters. A game logs completed levels, and an app with a social side logs Facebook shares. Each event can carry up to 10 string parameters, such as episode number and watch duration, and the platform allows 300 events in total.
Reusing one event with parameters keeps you under the cap. A tutorial step designed to take one minute but taking five points to a confusing step.
Use funnels to find the exact step where users quit. In the hotel app example, 75% saw a hotel list, 35% selected one, 20% entered payment and 17% got confirmed.
The question to chase is where the last 3% went between payment and confirmation.
Use return rate by cohort and aim to keep over 50% past month four. Flurry compared a cross-section of apps and found those holding over 50% of users past month four had a much longer shelf life.
Static return rate is the metric many of Flurry's clients rely on most.
Track from day one, even if you do not look at the data. Collect data passively in the first release. If you do not look for six months, so be it, but the history is there when a question comes up.
Without it you iterate blind. Friedman and Mike Rollins gave this as their top advice, and the host admitted making this mistake.
Plan tracked events before launch, from mockups. Map key steps the way you would map pages to track on a website. Teams that add analytics two days before launch get less value.
Mike Rollins says this planning also improves product design, because it forces the team to name the points that matter most.
Basic integration takes about five minutes. Adding the SDK is a few lines of code and needs no servers on your side. A full advanced setup takes a few days, or about an hour with a roadmap and clear business goals.
Verify the setup by logging an event and checking event logs. Event logs show entries within 5 to 10 minutes, while dashboard data lags by three to six hours because of batch processing.
Checking every hour does not help, since there is no fixed schedule.
Friedman gets app ideas from daily frustrations. Her own idea is a "Tinder for Manhattan apartments" for her housing hunt, and she uses the Tayasui Sketches drawing app because it is fun and relaxing.
Any app could be the next hit, so you need a view into it.