23
As central Ohio’s leading resource for information, we tell the stories that shape and change lives. BRAND CAMPAIGN, SWEEPS & DATA SCIENCE Friday, August 22, 2014

Media that Inspires: Roundtable 2014 Nikhil Hunshikatti

Embed Size (px)

DESCRIPTION

Nikhil Hunshikatti's presentation at the 2014 ROUNDTABLE preconference workshop on data-driven revenue diversification.

Citation preview

Page 1: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

As central Ohio’s leading resource

for information, we tell the stories

that shape and change lives.

BRAND CAMPAIGN, SWEEPS &

DATA SCIENCE

Friday, August 22, 2014

Page 2: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

Central Ohio is a strong

news market

Village Elders and Elite Empty Nesters

2006 – 26% of market

Traditionalist and Stubborn Seniors

2013 – 21% of market

Media Sophisticates

2006 – 13% of market

Media Sophisticates 2.0

2013 – 27% of market

Page 3: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

Understanding Our Market

Share Gives Us Insight Into

The Opportunity:

26% Combined Penetration

Among Key Segments

Traditionalists

34% (62,670 households)

Stubborn Seniors

16% (6,491 households)

Media Sophisticates 2.0

22% (63,501 households)

Page 4: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

Media that Inspires As central Ohio’s leading

resource for information,

we tell the stories that

shape and change lives.

Functional Benefits: Informed, Insightful, Guide to Life in Central Ohio and its Communities

Emotional Benefits: Empowerment, Serendipitous Experience, Guide to Everything that Matters, Trust

Page 5: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti
Page 6: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti
Page 7: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti
Page 8: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

The Columbus Dispatch - Sweeps Campaign The Overview

• During the month of March, The Columbus Dispatch launched an editorial “sweeps” campaign in conjunction with a subscription drive and a promotion tied with the Columbus Blue Jackets NHL hockey team.

• For the “sweeps” portion, our editorial team built a full month of planned news coverage for us to promote via in-paper ads (daily section front banners pushing to the next day’s featured story), as well as topical messaging on television, radio and electronic billboards.

• Dispatch reporters were also sought by local radio hosts for “on-air” appearances…as Experts to share the stories

Page 9: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti
Page 10: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

All stories were promoted in

The Dispatch the day prior to

running

The Columbus Dispatch - Sweeps Campaign The Editorial Highlights

Page 11: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

Other engagement measures that saw high scores include:

• Article being discussed with others (66%)

• Perception of the articles being more interesting (86%)

• Overall impression of articles (84%)

The Columbus Dispatch - Sweeps Campaign Results

19 sweeps stories tested

using a “split” RAM panel

27% lift in article reading (vs. comparable)

51% lift in article reading (overall)

Page 12: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

• 5,000 leads generated

• 5% Increase in Subscription Sales

• 7% increase in voluntary subscription starts

• Reduction in “Delta” between starts & stops (219 vs 529 weekly average)

• Improved single copy returns and sales (down only 5% YoY vs. 12% prior 2 months)

• 1 person $10,000 richer

The Cash On Ice Contest Results

Page 13: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

The Cash On Ice Contest Results

And, Editorial actually covered the promotion and ran the winner’s photo, video & caption!

Page 14: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

Coming Up…

• September Sweeps – Continued Editorial commitment

– Increased advertising commitment • $175,000 incremental ad buy from hhgregg

• Promotion ($30,000 hhgregg sweepstakes)

– Circulation Marketing goal of 10% lift in sales (vs. 5% in March)

• November/December & March Sweeps – Continue partnering with NHL franchise

– Automotive partner

Page 15: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

Data Science

Big Data Initiative

CLV & Applications

Churn Modeling

Page 16: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

Big Data Initiative

Tracker/Listener Customer

Lifetime Value Entity

Resolution Hadoop Stack

Digital Data Analytics

Page 17: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

Customer Lifetime Value (CLV) – Building Model

Once the source data is identified and available in a central location, the

CLV model can be implemented in a scientific way

• Construct a customer service history

– Start and end date

– Usage statistics (e.g., prices, payments, complaints, classified usage)

– Service changes

• Attach indicators to the service history

– Demographics (e.g., income, age, gender, children, education)

– Marketing-specific metrics (e.g., preprint revenue, ROP adv. value)

– Organization-specific metrics (e.g., payment methods, costs, credits)

• Develop a retention (survival) model

• Calculate “best fit” (regression) model correlating customer variables to

retention

• Calculate the CLV using the retention rates, revenue and cost metrics

How the model is built – and implemented – is crucial to making CLV scores

actionable and statistically significant

Page 18: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

Customer Lifetime Value (CLV) – Application

Enhancements in the works: 1. Web & mobile usage / content trends 2. Email/newsletter signups 3. ROP & Digital advertising attribution

Page 19: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

Customer Lifetime Value (CLV) – Application SOURCE: DIRECT

_DIRECT Group 1 (1-20) Group 2 (21-40) Group 3 (41-60) Group 4 (61-80) Group 5(81-100) All Subscribers

Avg Wkly Circulation Revenue $2.17 $3.11 $3.91 $5.77 $6.90 $2.98

Avg Wkly Preprint Revenue $0.90 $1.03 $1.11 $1.29 $1.35 $1.00

Avg Wkly Distribution Expense $0.63 $0.68 $0.71 $1.10 $1.04 $0.69

Avg Wkly Newsprint/ Ink Expense $0.43 $0.54 $0.60 $0.98 $1.05 $0.53

Avg Wkly Operating Margin $1.99 $2.89 $3.69 $4.95 $6.17 $2.74

Avg CLV $314.43 $563.16 $752.28 $1,026.36 $1,308.66 $507.68

Qty Active Subs 2203 1201 674 237 89 4404

Qty 7DAY subs 309 294 226 188 79 1096

Qty SUN subs 1527 603 304 16 0 2450

Qty WEDSUN subs 5 16 9 1 0 31

Qty WEEKEND subs 157 201 126 30 9 523

Qty OTHER subs 205 87 9 2 1 304

Sum

mar

y

SOURCE: INSERT

_INSERT Group 1 (1-20) Group 2 (21-40) Group 3 (41-60) Group 4 (61-80) Group 5(81-100) All Subscribers

Avg Wkly Circulation Revenue $2.44 $3.20 $4.06 $5.96 $6.93 $3.49

Avg Wkly Preprint Revenue $0.91 $1.00 $1.09 $1.26 $1.36 $1.02

Avg Wkly Distribution Expense $0.66 $0.68 $0.75 $1.13 $1.02 $0.74

Avg Wkly Newsprint/ Ink Expense $0.45 $0.53 $0.62 $1.03 $1.01 $0.59

Avg Wkly Operating Margin $2.22 $2.98 $3.75 $5.05 $6.25 $3.16

Avg CLV $322.92 $564.66 $750.37 $1,034.37 $1,315.05 $581.95

Avg Subscriber Tenure (days) 2481 4271 4304 4744 3857 3644

Qty Active Subs 1402 1226 613 388 144 3773

Qty 7DAY subs 172 264 212 331 118 1097

Qty SUN subs 785 575 231 8 0 1599

Qty WEDSUN subs 4 20 16 1 0 41

Qty WEEKEND subs 261 270 150 44 25 750

Qty OTHER subs 180 97 4 4 1 286

Sum

mar

y

Page 20: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

CIRCULATION SUBSCRIPTION TRANSACTION HISTORY WITH USER COMPLAINT DATA TO PREDICT DROPPED SUBSCRIPTIONS

Predicting Subscriber Churn using maximum entropy classification & topic analysis of user complaints

Page 21: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

• Various baseline predictive models were implemented and run, using only status

sequences, complaints counts, and service related features from complaints text as

the predictive feature

• Several augmented models were also built with previous history status sequence as

the baseline and complaints counts and textual features added on top by category

like complaints, routes, missed deliveries, service, WSJ, etc.

– The status sequences were built using filler statuses TRUE and FALSE in the normalization

step.

– Since these can show up in the label generation portion of the sequence these were

subsequently left out of the status sequence to make the predictor feature truly

independent

• 3 years data could be used ~250MB pre-processed (Training Set vs. Control group)

Churn Prediction – Multiple Iterations

Page 22: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

Results & Action items

Model % Accuracy Prediction Window

baseline+status 88.16 6

baseline+status+complaints 90.91 3

baseline+status+complaints 85.67 12

baseline+status+service 90.78 3

baseline+status+service 89.24 12

Action items:

Live testing predictions – month of August (maintaining status quo)

Adding variables to the mix

Institutionalize process

Page 23: Media that Inspires: Roundtable 2014   Nikhil Hunshikatti

Questions?

Nikhil Hunshikatti Director – Marketing & Research The Columbus Dispatch Tel: 614.461.5150 Email: [email protected]