detecting and building habits: a data-driven approach - robert moore, ceo, rj metrics - 2016 habit...

62
Detecting and Building Habits: A Data-Driven Approach Habit Summit 2016 | Stanford University | March 22 nd , 2016 Robert Moore CEO, RJMetrics

Upload: habit-summit

Post on 20-Jan-2017

725 views

Category:

Business


0 download

TRANSCRIPT

Page 1: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

Detecting and Building Habits: A Data-Driven Approach

Habit Summit 2016 | Stanford University | March 22nd, 2016

Robert Moore CEO, RJMetrics

Page 2: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Before we begin…

Disclaimer

Page 3: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

“During digestion, your body breaks down the nutrients in food. This causes a rise in blood sugar that triggers your pancreas to release insulin into the bloodstream. Insulin increases the amount of serotonin and melatonin that flood the brain, two chemicals associated with drowsiness.” -Science

Hope You Enjoyed Lunch

Page 4: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Willpower Depletion “The longer we work on mentally strenuous tasks, the more mental energy we expend, and eventually we’ll run out and start falling back to these easy — and often wrong — default decisions.” -Also Science

Page 5: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Willpower Depletion “The longer we work on mentally strenuous tasks, the more mental energy we expend, and eventually we’ll run out and start falling back to these easy — and often wrong — default decisions.” -Also Science

Page 6: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

If you feel yourself getting sleepy during this talk…

Disclaimer

Page 7: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

If you feel yourself getting sleepy during this talk…

If you feel, with each minute that goes by, you are somehow dumber than before I started talking…

Disclaimer

Page 8: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

If you feel yourself getting sleepy during this talk…

If you feel, with each minute that goes by, you are somehow dumber than before I started talking…

blame science!

Disclaimer

Page 9: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

Detecting and Building Habits: A Data-Driven Approach

Habit Summit 2016 | Stanford University | March 22nd, 2016

Robert Moore CEO, RJMetrics

Page 10: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

Why We’re Here

Page 11: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Storage in the Early 1980s

Page 12: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Practical Implications

Data StorageType SpaceRequiredDateandTime TIMESTAMP 4bytesTransac7onAmount DECIMAL(10,2) 4bytesItemID SMALLINT 2bytesCustomerID INTEGER 4bytesLoca7on SMALLINT 2bytesTotal 16bytes=32¢

Page 13: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Practical Implications

Data StorageType SpaceRequiredDateandTime TIMESTAMP 4bytesTransac7onAmount DECIMAL(10,2) 4bytesItemID SMALLINT 2bytesCustomerID INTEGER 4bytesLoca7on SMALLINT 2bytesTotal 16bytes=32¢

50¢

Page 14: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Data Analysis Was Irresponsible For Most Companies

Not long ago: •  1 byte: $0.02 •  1 kb: $20 •  1 mb: $20,000 •  1 gb: $20,000,000

Page 15: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Things Have Changed

Page 16: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

What A Time To Be Alive! That’s 62.5 million cups of coffee! Not to mention inflation (avg 2016 price)

Page 17: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Changing Times: It’s Not Just Storage

Page 18: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Page 19: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Page 20: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Page 21: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Page 22: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Page 23: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

What This Means for Business

Behavioral Data

Attribution Data

Transaction Data

Customer Data

Why?

Where?

What/When?

Who?

Time

Page 24: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

What This Means for You

It is easier than ever to use data to understand the behavior and habits of your customers.

Here’s how.

Page 25: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

Statistical Groundwork Part I: Correlation and Causation

Page 26: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Correlation and Causation Just because two things consistently move in the same direction, that doesn’t mean that one causes the other.

Page 27: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Correlation and Causation

A B Nick Cage Films Drowning Deaths

Page 28: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Explanation 1: Direct Causation

A B Nick Cage Films Drowning Deaths

Page 29: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Explanation 2: Indirect Causation

A B Nick Cage Films Drowning Deaths

Page 30: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Explanation 3: Lurking Variable

A B Nick Cage Films Drowning Deaths

C

Page 31: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Explanation 4: Cyclical

A B Nick Cage Films Drowning Deaths

Page 32: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Explanation 5: Unrelated

A B Nick Cage Films Drowning Deaths

X

Page 33: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Focus on Real Improvements, Not Correlations

An iterative process based on data allows you to find real results, even in the face of false positives

Page 34: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

Statistical Groundwork Part II: Statistical Significance

Page 35: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Statistical Significance: Go Big

Experiment A Old Conversion Rate: 1.0% New Conversion Rate: 1.1% Trials Required: ?

Page 36: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Statistical Significance: Go Big

Experiment A Old Conversion Rate: 1.0% New Conversion Rate: 1.1% Trials Required: 159,650

Page 37: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Statistical Significance: Go Big

Experiment A Old Conversion Rate: 1.0% New Conversion Rate: 1.1% Trials Required: 159,650

Experiment B Old Conversion Rate: 1.0% New Conversion Rate: 2.0% Trials Required: ?

Page 38: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Statistical Significance: Go Big

Experiment A Old Conversion Rate: 1.0% New Conversion Rate: 1.1% Trials Required: 159,650

Experiment B Old Conversion Rate: 1.0% New Conversion Rate: 2.0% Trials Required: 2,226

Lesson: If you’re small, go big or die waiting Get the numbers at www.testsignificance.com

Page 39: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Side Benefit of Going Big

Page 40: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

Techniques Part I: Golden Motion Analysis

Page 41: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Our Golden Motion Experiment What am I trying to learn? RJMetrics has a free trial. What behaviors (“golden motions”) increase the likelihood of user conversion? How will this change my behavior? If we identify actions that are linked to strong increases in conversion, we can change our onboarding process to that more users will take those actions. How will we measure success? Increase in free trial conversion rate.

Page 42: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Step One: Instrumentation

Just a few years ago, this was extremely hard. Today, it is trivially easy!

What might we want to track: •  Logins •  Chart Creation •  Chart Editing •  Adding Users

•  Sharing Dashboards •  Using APIs •  Connecting Multiple Data Sources •  Time Spent In Tool

Page 43: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Step Two: Consolidation Get that event data right alongside everything else you know about your customer or prospect. Again, third party tools have made this very easy!

(my only plug, I swear!)

Page 44: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Step Two: Consolidation

Page 45: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Step Two: Consolidation UserID Didtheyconvert? ChartsCreated UsersCreated DataConnected …

1 No 0 1 1 …

2 Yes 10 3 1 …

3 No 4 5 1 …

4 No 2 4 2 …

5 No 0 3 3 …

6 No 0 4 2 …

7 Yes 14 7 1 …

8 No 0 5 1 …

… … … … … …

Page 46: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Step Three: Run Tests Example: Does connecting more data sources increase conversion rate? Data: Result: No statistically significant difference. (two-tailed p = 0.71)

Data Sources Conversion Rate Sample Size

1 30% 90

More than 1 34% 110

Page 47: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Step Three: Run Tests Example: Does creating more charts increase conversion rate? (this was WAY down on my list of guesses)

Page 48: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Step Three: Run Tests Example: Does creating more charts increase conversion rate? Data: Result: YES! Statistically significant difference. (two-tailed p < 0.01)

Charts Conversion Rate Sample Size

0 30% 70

1 or More 62% 130

Page 49: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Step Three: Run Tests Example: Does creating more charts increase conversion rate? Data:

Charts Conversion Rate Sample Size Significant?

0 30% 70 Yes

1 48% 55 Yes

2 66% 35 Yes

3 70% 20 No

4 or more 71% 20 No

Page 50: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Step Four: Change Behavior Objective: Make sure every client creates at least two charts on their own. Changes: •  Introduce self-guided tour of report builder •  Give new clients “homework” after launch calls •  Beef up documentation on report builder •  Give our Account Management team tools to track this metric

Page 51: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Step Four: Measure Again Nothing is promised! Your work isn’t done.

Don’t get Nicholas Cage’d.

Page 52: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Step Four: Measure Again

0%

10%

20%

30%

40%

50%

60%

70%

1/1/13 2/1/13 3/1/13 4/1/13 5/1/13 6/1/13 7/1/13 8/1/13 9/1/13 10/1/1311/1/1312/1/13

ConversionRate

Change implemented here

BOOM!

Page 53: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

Techniques Part II: Cluster Analysis

Page 54: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Cluster Analyis We can start with that same event data from the last example

Page 55: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Cluster Analysis Decide on attributes to cluster on: Maybe user profiling is more complex than just “viewers” and “creators.” And where do you draw the line?

Page 56: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

Cluster Analysis Algorithms (in this case the k-means algorithm) group our users into groups that are most statistically similar to one another based on the behaviors we’ve observed

Page 57: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit
Page 58: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

We found the TVs!

Page 59: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

We found the Execs!

Page 60: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

What’s Left? Opportunity!

We know

where to focus.

Page 61: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

rjmetrics.com

The Opportunities Are Enormous

Give data a seat at the table!

It won’t answer everything, but it’s an important voice when you’re studying

user habits.

Page 62: Detecting and Building Habits: A Data-Driven Approach - Robert Moore, CEO, RJ Metrics - 2016 Habit Summit

@robertjmoore

[email protected]

@robertjmoore

rjmetrics.com

Q&A