WELCOME TELLING DATA STORIES TO STORYTELLERS DATA MATURATION AT NBC - - PowerPoint PPT Presentation

WELCOME

TELLING DATA STORIES TO STORYTELLERS

DATA MATURATION AT NBC NEWS DIGITAL Asher Feldman, Patrick White

3

YOUR STORYTELLERS FOR THIS BREAKOUT

Asher Feldman

Senior Manager, Analytics

NBC News Digital

Patrick White

Manager, Analytics

NBC News Digital

4

OUR DATA STORIES FOR ALL OBJECTIVES

5

GETTING NBC NEWS OFF THE GROUND LAUNCHING OFFERING DATA-FIRST DATA SCIENCE FOR BUSINESS STAKEHOLDERS

Media

HOW NEWS AND DATA MATURITY WERE BUILT

6

NEWS/MSNBC TODAY

NEWS DIGITAL

Trial and Quest Initiate Evaluate Test Our Strength Conquer, Then Expand Our Journey

GETTING OFF THE GROUND

HISTORY WAS OBSCURING A DATA RICHNESS

8

80

years of archives

278K

videos published

111K

articles written

110M

unique visitors per month

A full accounting…

The boring chapter

Establish a visual language…

World-building

A QUEST TO ACCESS UNDERSTANDING

9

Asher’s Inbox

(10 Workdays)

Shared Files Reports Intros

Key Aspects

(Inferring purpose)

Benchmarks Pacing Share of the pie KPIs

DIRECT MANIPULATION = DIRECT VIZ

10

Creating Medians for Like-Like Comparisons

• Data sets compared to previous pacing via

median analysis

• Replicated data flow by series of brand and

time frame (weekdays, Sundays, weekly, etc.) Scan and Pass Visualizations

• Though the connotation is negative, the
• utcome is positive
• Users don’t really have to think, they can just

compare Combine to Explain Longer Trends

• Training on certain metrics makes adapting

weekly or monthly reporting easier

• Marimekko charts are especially useful for

showing content relative to overall size

XX

A VISUAL LANGUAGE OF DOMO AT NBC NEWS

11

“Complete game-changer that will save me so much busy work.” – Editor, TODAY Parents “I love this!” – SVP, News Digital AND Director, HR (We have these emails)

Adaptation & Color

• When ‘share’ is key, rectangles over pies
• Utilize defined toolbelt of comparative metrics
• Side-by-side series across related metrics

Single Point of Truth

• Create one interactive dashboard instead of

15 filtered copies of the same card per metric

• Increase visibility by reducing share across like

teams Start Simple

• Adapt visual language as frequently as

possible for familiarity among users

X X X

LAUNCHING OFFERING DATA FIRST (NBC NEWS NOW)

LIVE NEWS WITHOUT THE HASSLE OF CABLE

13

NEWS DIGITAL +

+

Meet consumers where they are, not just TV 3p – 11p EST, soon to be 24/7 NBC News apps, XUMO, & Pluto TV with more to join

What a new idea needed…

Creating requirements from scratch

What we achieved (and didn’t)…

Executing on these ideas where we can

Idea Analysis Sign-off Launch Inputs Reporting

INTEGRATING NEW INPUTS INTO DATA / DESIGN

14

NEWS NOW NEWS DIGITAL STRATEGY METRICS LEARNINGS DUMMY DASH CADENCE NEW TACTICS WEEKLY SYNC USABLE METRICS

RESULT: BUILDING UNDERSTANDING VIA DOMO

15

Communication

• Communicating success through the

dashboard, not around it

• Incentivizing DOMO usage through historical

comparison Visualization

• Clear breakdowns of consumption patterns to

understand the audience

• Sticking to established visual language

Actualization vs. Norms

• Content analysis to see what works and what

doesn’t

• Even more importantly, understanding when

works

USING DATA SCIENCE FOR BUSINESS STAKEHOLDERS

Theory: Everything. Is. Broken.

Especially Google

Problem:

SOLVING A TRUE BUSINESS PROBLEM: PANIC

17

11M

UVs

5.5M

starts

7.2%

convert

The Right Question:

How weird is weird?

‘MANUAL’ DATA SCIENCE AS STRESS RELIEF

18

Data science by the hour…

Channeling the earliest data scientists

Understanding among friends…

An illustrated journey

Hour Sun. Mon. Tue. Wed. Thu. Fri. Sat. 5:00 AM

7% 11% 6% 5% 7% 6% 8%

10:00 AM

6% 9% 4% 7% 10% 4% 7%

11:00 AM

6% 10% 5% 10% 17% 4% 6%

12:00 PM

7% 12% 5% 9% 14% 5% 5%

1:00 PM

7% 14% 5% 7% 15% 6% 6%

2:00 PM

6% 20% 4% 8% 10% 5% 6%

8:00 PM

6% 7% 7% 21% 26% 8% 10%

9:00 PM

6% 7% 5% 31% 43% 4% 8%

10:00 PM

5% 5% 8% 25% 33% 12% 7%

• Avg. Deviation

from Hour-Day ‘Normal’

Understanding Sources of Volatility Examining Source of Volatility

0% +50%

• 50%
• 25%

+25%

Establishing Thresholds by Volatile Source Asking the Right Questions

• f Volatility

How weird is weird?

68% 95% 99%

How much weirdness are we seeing?

UNDERSTANDABLE RIGOR AND VIZ IN DOMO

19

“You’d be so proud of me, I used it!” – Assistant Managing Editor, News

Source 1 Source 2 Source 3 Source 4 Source 1 Source 2 Source 3 Source 4 Source 1 Source 2 Source 3 Source 4

Set-up

• Recontextualize data findings in column charts
• Explanation for clarity on each visit
• Color set is consistent across dashboard

Hourly Tracking & Alerts

• Bar charts track hours below threshold day-of
• Alerts trigger necessity for investigation
• Replication of ‘manual’ data science

Long-Term Analysis

• 10-day tracking for regular more casual users
• Identifies ”non-alertable” trend
• Allows for constantly updating comparisons

WHAT CAN YOU ACCOMPLISH?

20

TELL DATA STORIES AT OLD COMPANIES BUILD NEW STUFF WITH DATA LEADING SEED DATA SCIENCE INTO DATA CONVOS

THANK YOU