23
Automated Classification of Demonstrated Interests at Scale & Depth

Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

Automated Classificationof Demonstrated Interests

at Scale & Depth

Page 2: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

WHO IS eCONTEXT?● Digital Behavior

Classification● Click-stream & Search

Query Classification● Social Media Analysis● Video & Audio Classification

● Audience Insights● Media Planning & Buying● Personalized Content

Recommendations● Targeted Advertising

THE WORLD’S LARGEST TEXT CLASSIFICATION ENGINE

Page 3: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

How Insightful are Your Consumer Insights?

Page 4: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

IDEAL CONSUMER ANALYSIS

● Deep understanding of the authentic interests, affinities, values driving consumer activities

● Granular & structured segmentation of users● Standardizing answers across a large pool of consumers

○ Drawing trends and identifying commonalities

Page 5: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

Active vs Passive Data Collection

Needs, Shortcomings & Opportunities

Page 6: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

ACTIVE DATA COLLECTION: SURVEYS, INTERVIEWS, DIRECT CONVERSATIONS

STRENGTHS● Specific and granular responses● Standardization

○ Possibility for quantitative analysis

WEAKNESSES● Slow & difficult to scale● Introduces an Uncertainty

Principle● Possible to introduce an

Observer Effect or other bias○ Social Desirability Bias○ Demand Characteristics○ The Hawethorne Effect

Page 7: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

Declared vs Demonstrated Interests

Page 8: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

“Most of our personalization right now is based on what [users] actually watch, and not what they

say they like. Because you can give five stars to An Inconvenient Truth because it's changing the

world, but you might watch Paul Blart: Mall Cop 2 three times in a few years. What you actually

want and what you say want are very different

Todd Yellin, VP Product Innovation, Netflix, May 2015

Page 9: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

SCALABLE PASSIVE DATA COLLECTION & ANALYSIS VIA

HI-RES TAXONOMIC CLASSIFICATION

DEEP, GRANULAR AFFINITY & INTEREST DATAAND

CONSISTENCY ACROSS CONSUMER RESPONSES

Page 10: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

WHAT DO I WIN?

DEPTH, BREADTH AND REPEATABILITY

CROSS-CHANNEL DIRECT COMPARISON

SHORTENED TIME-TO-INSIGHT

MINIMIZED RESEARCHER/ EXPERIMENTER IMPACT

CURRENT DATA ASSETS WORK HARDER

Page 11: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

Closed Network Use Case:Mobile Device User Panel (UK)

Page 12: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

CLASSIFICATION OF USER BROWSING INTERESTS

● Agency with full activity data from users’ mobile devices

● Agency could ask quantitative questions about device usage, but qualitative questions were limited

● Agency needed deep topical understanding of browsing data at individual page level

SITUATION

Page 13: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

CLASSIFICATION OF USER BROWSING INTERESTS

● As users browse the web, agency passes complete URL data to eContext’s API endpoint for website classification

● eContext classifies the URL, replicating the experience the user had at that time

SOLUTION

Page 14: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

CLASSIFICATION OF USER BROWSING INTERESTS

● Agency can sell deep, qualitative insights with no additional participant interaction required

● Panel can be leveraged to tell any type of story○ Path-to-Purchase○ Media Planning

● Opportunity for on-prem version of the website classification service for providers with privacy requirements

OUTCOME

○ Campaign Effectiveness○ Demographic Segmentation

Page 15: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

Closed Network Use Case:User Desktop Search &

Browsing Panel (US)

Page 16: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

USER SEGMENTATION VIA SEARCH QUERIES

● Examined search queries from members to recruit them into studies

● Segmentation was a manual process, using basic boolean search and hand-coding

● Company needed to automate and scale the classification

SITUATION

Page 17: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

USER SEGMENTATION VIA SEARCH QUERIES

● Agency passed all panelist search queries to the eContext API endpoint for search classification

● Tens of thousands of keywords classified to highly granular topics in seconds, not weeks

● Any queries passed to human analysts for additional segmentation had much lower noise

SOLUTION

Page 18: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

USER SEGMENTATION VIA SEARCH QUERIES

● Company drastically lowered person-power costs while increasing coding accuracy

● Identified a greater number of panelists, in more specific clusters

● Subsequent surveys could be normalized if declared behavior conflicted with demonstrated behavior

OUTCOME

Page 19: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

Open Network Use Case:DIY Consumers on Twitter

Page 20: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

MODERN MEDIA PLAN FOR HOME IMPROVEMENT BRAND

● Agency for Home Improvement brand needed to connect on “shared meaning” with their key customers

● Primary KPI was relevant connections for higher CLTV

● Agency wanted to automate a manual-coding process

● Agency wanted to aggregate insights across channel

SITUATION

Page 21: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

MODERN MEDIA PLAN FOR HOME IMPROVEMENT BRAND

● eContext analyzed thousands of social media posts by brand followers & competitive followers in only a few hours

● eContext built a standard social topic index

● eContext combined socially-shared content by Home & Garden publishers

SOLUTION

Average Topic Frequency

Page 22: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

MODERN MEDIA PLAN FOR HOME IMPROVEMENT BRAND

● Distinguished brand followers compared to competitive followers

● Cleanly identified where brand values and consumer values/interests aligned

● Delivered a map for future content creation & clear path to audience connection

OUTCOME

Brand Followers Competitive Followers

Easy Woodworking

Projects

Home Improvement Supplies

Car

pent

ry

Tool

Cab

inet

s

Cus

tom

Fur

nitu

re

Ant

ique

Tab

les

Bat

hroo

m T

ile

Doo

rs &

Win

dow

s

Sin

k In

stal

latio

n

Pla

ntat

ion

Shu

tters

Page 23: Automated Classification of Demonstrated Interests …insightinnovation.org/wp-content/uploads/2016/07/PDF/ian.pdfInconvenient Truth because it's changing the world, but you might

Thank YouIan McCarty

VP of Operations [email protected]

www.econtext.com @econtext 312-477-7300180 N. Michigan Ave. Ste. 1400, Chicago, IL 60601