mmf data toolkit (presented health refactored 2014)
DESCRIPTION
Addresses data infrastructure and how to scale efficiently. As a platform that is more than doubling in user base (and activity) YOY, MapMyFitness has build a sustainable platform that provides more than 22 million users worldwide access to the information they want in a simple, rewarding way. In the company’s early years, health care providers such as Humana and Discovery were the first major partners to recognize the importance of fitness data and built a community on top of our software that opened the door to their members as well. Now every major brand from Coca-Cola to Purina Dog Food is seeing the value of this data and how they can tap in to promote a healthy and active lifestyle through interactive digital tools. As a new addition to the Under Armour family, the company will look to move past just steps and miles, and more into your daily story of activity. Kevin Callahan will address what challenges lie ahead for presenting data in a compelling way, and the opportunities as a brand to be a leader in Connected Fitness.TRANSCRIPT
Dealing with Health Data Op#mizing Data Toolkits
2
THE MAPMYFITNESS BRAND AND PRODUCT PLATFORM
Our integrated fitness plaGorm consists of iOS and Android apps, a suite of websites, hundreds of third-‐party devices, and API and SDK solu#ons, all synchronized with the cloud. It’s the most flexible and robust pla9orm in the world of Connected Fitness.
3
WHO IS MAPMYFITNESS?
We are Connected Fitness — we are obsessed with using technology to make fitness social, fun, simple, effec#ve, and rewarding.
We are Innovators – aYer 8 years we have many “firsts” we are proud of: • First online fitness mapping and tracking system on web and
subsequently mobile • First connected fitness media plaGorm reaching over 24M
consumers • First fitness tracking plaGorm used for rewards programs by
major health insurers • First to support wireless Bluetooth 4.0 sensor connec#ons on
iOS and Android
We are a Business -‐ we have been genera#ng revenue since day 1 and were recently acquired by Under Armour.
4
LOTS OF DATA
1 New users per month
24 MILLION
Registered members
over
avg MILLION
5
LOTS OF DATA
1 New users per month
24 MILLION
Registered members
over
avg MILLION
24M < NYC 9M
6
LOTS OF DATA
15 MILLION
Workouts in the last 30 days
over
350 Workouts each minute
7
LOTS OF DATA
2M lbs fat burned Last 30 days…
8
WHAT IS CONNECTED FITNESS?
9
CONNECTED FITNESS IS EXPLODING
170 million wearable devices projected to be shipped by 2016*
$3.9 billion es#mated market for fitness and
wellness devices in 2016*
MapMyFitness syncs the data from over 400 fitness devices and brings it to life in a way that’s mo#va#ng, inspira#onal, and useful.
The world is on the cusp of a Connected Fitness Revolu7on.
* IMS Research
10
PEOPLE ARE ADOPTING LIKE CRAZY, BUT THERE’S A CATCH…
Users have as many interfaces as they do devices & services. It’s a fractured mess.
Too much to keep track of!
11
WE ARE THE API, THE PLATFORM AND THE DASHBOARD
MapMyFitness synchronizes with over 400 third-‐party wearable fitness devices & services
12
Device integraWon
MAPMYFITNESS PLATFORM: THE UNIFYING HUB
• Comprehensive online solu#on for tracking fitness
• Mobile & web plaGorm using everyday devices (smartphones)
• Cloud-‐based data management for tracking fitness
• Users can upload results, share progress, and mo#vate one another
• Reference framework for connec#ng 3rd party devices to web, social sharing & data
Web FuncWonality
Mobile Apps
Core Members
Extended reach
Social engagement
13
API AND SDK SOLUTIONS
Access core MapMyFitness features with our SaaS plaGorm. With both push and pull func#onality, you can integrate our tools into your digital experiences or join our ecosystem to access fitness data at scale.
Our licensed solu#ons include: • Routes Widgets • Device Integra#on Widgets • Nutri#on Widgets • Users API • Routes API • Workouts API • Ac#vity Feed API • Groups API • Events API • iOS SDK • Android SDK
14
MAPMYFITNESS PLATFORM: THE UNIFYING HUB
MapMyAPI.com
15
SCALING AND WORKING WITH DATA
16
EXPONENTIAL GROWTH, TIMELY SCALING
0
40M
80M
120M
160M
2006 2007 2008 2009 2010 2011 2012 2013
Users Routes Workouts
2005
We’ve evolved our scaling approaches as usage has grown
Premature scaling would have been wasteful.
Scaling solu#ons enable
business growth!
Scaling problems got interes#ng.
2014?
17
EXPONENTIAL GROWTH, TIMELY SCALING
0
40M
80M
120M
160M
2006 2007 2008 2009 2010 2011 2012 2013
Users Routes Workouts
2005
2013-‐2014 March Workout Comparison 2013: 9M vs. 2014: 16M
18
EVOLUTION OF THE INFRASTRUCTURE
2005 SINGLE VPS 2007 PHYSICAL SERVERS
Managed hos#ng
2010 HYBRID Public cloud for new applica#ons
2011 HYBRID Hybrid applica#ons with cloud and
physical capacity
19
PREDICTABLE VARIATION IN TRAFFIC
Over years
Weeks
Days
20
EVOLUTION OF THE INFRASTRUCTURE
2012 CONSOLIDATION 2013 OPTIMIZATION
Private / Public Cloud Specialized Data Clusters (MySQL, PostgreSQL, Mongo) Separa#on of Services (SOA)
21
DATA TOOLKITS
• Old Way • Simple SQL • Excel Pivot Tables
• New Way • Amazon RedshiY • Aggregate SQL tables, Data Par##oning • iPython Notebook and SciPy
• Future... • we’ll get to that in a minute
-‐-‐
22
AMAZON REDSHIFT
Amazon Redshid is a fast, fully managed, petabyte-‐scale data warehouse service
23
IPYTHON NOTEBOOK
The IPython Notebook is a web-‐based interac#ve computa#onal environment where you can combine code execu#on, text, mathema#cs, plots and rich media into a single document.
24
EXAMPLE 1: SIMPLE WORKOUTS FROM API
25
EXAMPLE 1: SIMPLE WORKOUTS (CONT)
26
PANDAS: PYTHON DATA ANALYSIS LIBRARY
hvp://pandas.pydata.org/
en#re data analysis workflow in Python without having to switch to a more domain specific language like R.
Google “10 Minute Pandas Tour” By Wes McKinney
27
PANDAS AND REDSHIFT
Connect directly to RedshiY Data Source!
28
PANDAS AND REDSHIFT (ZOOM)
Connect directly to RedshiY Data Source!
df = sql.read_sql(sql_query, conn)
29
EXAMPLE 2: PLOTTING MAP DATA
Trick for UI
30
EXAMPLE 2: PLOTTING MAP DATA (CONT.)
31
MATPLOTLIB BASEMAP TOOLKIT
library for plozng 2D data on maps in Python.
32
EXAMPLE 3: ADVANCED BASEMAP VISUALIZATIONS
33
EXAMPLE 5: DATA ANALYSIS FOR SALES
34
EXAMPLE 6: USER SEGMENTATION
35
FUTURE PLANS
• Develop more robust tools to access our GIS, ac#graphy and physiological data • Explore Interac#ve • D3, MapBox/TileMill, etc.
• API endpoints tailored to data analysis • Incorpora#ng UA retail and consumer data to support UA corporate goals • recommenda#on systems • prescrip#ve training • route genera#on
36
Thank You!
WE ARE HIRING IN AUSTIN, TX