pythagoras was wrong - insight innovation
TRANSCRIPT
Pythagoras was wrongDan Foreman
OKRA
All things consist of three . . . establish the triangle and the problem is two-thirds solved.
~Pythagoras
OKRA
Representation Learning Matching
Salesforce targeting
Client: Top ten biopharma companyAutomation task: Sales targetingBusiness context: Salesforce effectiveness
Sales targeting
Data: Prescription data + Activity data limited to only 6 months (3 used for learning, 3 for blind testing)Accuracy: 85% improved performance – conversion and churnProspect: High accuracy for multiple outcomes by adding more data sources
AUC = 0.86
t
True
pos
itive
rate
False positive rate
• .90-1 = excellent (A)• .80-.90 = good (B)• .70-.80 = fair (C)• .60-.70 = poor (D)• .50-.60 = fail (F)
Client: Top ten biopharma companyAutomation task: Sales targeting and messaging Business context: Sales force
GeographyHCP
Rep
Drug
8
Operates in real time Combines multiple data sources (activity,clinical,
claims, social, sales)
Creates individual eco-profiles (e.g. patient, drug,
practice, provider, geography)
Learns and understands health ecological patterns (e.g clinical, geographical
& behavioural patterns)
Spots anomalies and detects the exact moment of changes for each profile
Automates evidence generation for healthcare
and Pharma
Monitors all the profiles across all the data sources
provided
Before After
Manual rep activity planning
One-off salesforce studies, expensive and inaccurate
Time consuming studies and monthly planning
Repeated Monthly effort across
No evidence based planning
Automatic targeting and messaging,integration with CRM
Continuous and up-to-date insights
Reduced the planning time andeliminated salesforce studies
Evidence based salesforce planning
An accuracy up to 95% when moredata is used
Market research
Client: Global leader life scienceAutomation task: Social media listeningBusiness context: Launch campaign
Sales targeting
•94% detection rate of campaign related conversations and adverse event reporting
•74% reduction of time spend and manual effort by the vendor’s social listening team
•Automatic segmentation of content produced by: KOL, Patients, Caregivers and stakeholders.
•Real-time report generation instead of monthly report.
Client: Global leader life science sales and marketingAutomation task: Social listeningBusiness context: Market research
GeographyKOLs
Conversation
Source
Patient
Brand
0
0
1
1
1
1
0 0.25 0.5 0.75 1 1.25
Receiver Operating Curve for automatic social media listening
11
Operates in real time Combines multiple data sources (activity,clinical,
claims, social, sales)
Creates individual eco-profiles (e.g. patient, drug,
practice, provider, geography)
Learns and understands health ecological patterns (e.g clinical, geographical
& behavioural patterns)
Spots anomalies and detects the exact moment of changes for each profile
Automates evidence generation for healthcare
and Pharma
Monitors all the profiles across all the data sources
provided
Before After
Intensive manual social listening
High cost teams
Delays due to the manual process
Repeated effort across campaigns
Inaccurate due to human error
Automating social listening processwith a detection rate of 94%
Reducing the time spend by 74%
Reducing the volume of conversationto be manually reviewed by 76%
Allows for real-time instead ofmonthly insights
Filtering and segmentation accuracyof 94%
Other use cases
✦ Brand perception
monitoring✦ Brand attributes and
barriers tracking ✦ Parallel markets
comparison
✦ Customer insights ✦ Sentiment analyses
✦ Predicting clinical
outcome ✦ Disease parameter
ranking✦ Patient group
segmentation
✦ Patient with a disease detection
✦ Patient disease progression analysis
✦ Uplift modelling
✦ Business
intelligence ✦ Sales activity
optimisation✦ Target list
prediction
✦ Churn prediction✦ Customer bonus
automation✦ Multichannel
activity
optimisation
✦ Transaction fraud
detection and prevention
✦ Credit decision making support
✦ Social scoring
✦ Multi-source fraud detection and
prevention
OKRA real world insight
OKRA Sales & marketing support
OKRA market research OKRA Financial Services
Biometrics
13
Sync’ed to Apps
Everyday tasksMedicationCaloriesActivity levels…
Inevitable“By 2030 ML is replacing 90% of the operational tasks in an organisation.
Gartner research
Raphael AmorimAstronaut
Jeremy Howard Chief Scientist at Kaggle
Don’t take our word for it
“Thereisnoorganisationthatshouldn’tbethinkingaboutleveragingmachineintelligence,becauseeitheryoudo– inwhichcaseyou’ll
probablysurpassthecompetition– orsomebodyelsewill.”
“Yesterday we obeyed kings and bent our necks before emperors. But today we kneel only to truth, follow only beauty, and obey only love.”
― Kahlil Gibran, The Vision: Reflections on the Way of the Soul
Pythagoras was wrongDan Foreman
OKRA