GEO SYMPTOM SOLUTIONS Anurag Jain
Method of reach Content Categorization User Categorization based on site usage and declared information Scale for WebMD WebMD serves 2.6 billion pageviews a quarter That are consumed by million unique users a month Key challenge publishers face Relevant message can be given to users “after” they show interest in category. For many short-lived conditions, the message needs to reach users “before” they show clear interest. PUBLISHERS REACHING USERS
Identify “correct” geographic locations for area of interest Reach users if they belong to the geographic location of interest Adjust locations regularly to redirect reach as the target audience changes HYPOTHESIS
Data Source – Symptom Checker
Uses defined vocabulary MM USA visitors a year entering symptoms, spending on an average over 5 mins per visit Data consists of symptoms, qualifiers, and body locations entered by site visitors WebMD Symptom Checker Data Summary Sample Conditions (user shown multiple possible conditions) Influenza (Adult) Acute sinusitis Gastroenteritis Abscess Irritable bowel syndrome Sample Body locations and Qualifiers Pain or discomfort (Legs) - sharp or stabbing Difficulty sleeping – made worse by alcohol Cough – Worse at night Fever - over F (high) Weakness – made worse by heat Weakness – better with rest Sample Symptoms Pain or discomfort Headache Fever Numbness or tingling Dizziness
Seasonality of Flu (over 4 years) Population Proportion for ILI (Influenza like Illnesses) per week during Flu seasons
Variability of Symptoms
Western Cities 9 Geographical Variability Eastern Cities
Aggregate data by Geo Location DMA, Designated Market Area Apply Statistical model to: Mark DMAs with high/medium/low levels of activity Identify DMAs with clearly high level of activity with 95% confidence Pick top DMAs with highest level of activity Process
Symptom Checker data aggregated at State and County level for map Information Product – Geo Map (Dec 18, 2012)
Data aggregated at DMA level Compared with baseline to find areas under flu Make identified areas available for online and offline reach Commercial Product: Geo Symptom Targeting
Reaching users at WebMD based on inferred location Reaching users through Audience Extension Making data available for use by customer Supply chain management Better offline reach for users Commercial Product: Geo Symptom Targeting
14 WebMD/CDC Comparison: December Week 1 WebMD (Week ending 12/1/12) CDC (Week ending 12/1/12)
15 WebMD/CDC Comparison: December Week 3 WebMD (Week ending 12/15/12) CDC (Week ending 12/15/12)
WebMD/CDC Data Correlation
WebMD data ahead of CDC
Data highly correlated with gold standard of CDC Data available 1-2 weeks in advance of CDC as well as other Rx/Dx sources Being driven from declared symptoms, the data is more precise than search queries Targeting possible at custom group of Symptoms Advantage Over Similar Products
Geo Symptom Targeting works for conditions that are seasonal and that vary geographically Significant improvement to ROI Based on IRI study of the program for flu season Improved client interest for the product More than doubled committed revenue for flu season Multiple customers interested in product for next year The results are in…
Predictive models for better planning Other countries of interest Looking into UK, Canada and Australia Bio Surveillance - Identify unusual levels of activities and raise alerts Find information in other sources of data at WebMD Health Checks User Registration Quizzes What’s next for WebMD
Contact: Anurag Jain VP Engineering, WebMD Questions