DATA SCIENCE MAY 2018
@SiaPartners
01 OUR DATASCIENCE APPROACH 03 GLOBAL OVERVIEW OF OUR BOT FACTORY 05 CAPTURE – Insurance Pricing 07 CAPTURE – RegWatch 09 VISUALIZE – Weather & Climate 11 INTERACT – R-Bot 13 INTERACT – Recruitment Chatbot 15 ANALYZE – Fraud Detection 17 ANALYZE – Image Recognition 18 PRICE – Motor Insurance Pricing 19 PREDICT – Parking Availability 21 SIA PARTNERS
Our datascience approach 50 data scientists Datascience expertise based on Sector Knowledge • Banking 8 M€ turnover 3 teams worldwide +50 % annual growth • Insurance • Energy & Utilities • Transport, Industry & Retail An understanding of the whole data value chain • Data Acquisition • Data Processing 20+ clients trust us • Data Modeling • Data Visualization A quick glance at our Bot Factory, please visit d ata s c i e n ce- l a b .s i a - p a r tn er s .co m
We enable business activities with AI based use cases… MARKETING ANALYTICS SENTIMENT ANALYSIS • Segmentation / Profiling • Perception Study • Targeting / Client Life-value • Satisfaction Investigation • Geo-Marketing Implementation • Electronic Reputation FORECAST & ASSET MANAGEMENT BUSINESS PERFORMANCE • Consumption Forecast • Process Performance Measurement • Predictive Maintenance • Entity Operational Monitoring • ST and MT optimization • Analysis & visualization of time series FRAUD DETECTION REVENUE MANAGEMENT & PRICING • Uncommon behaviors • Measuring Demand Elasticity • Fraud Detection • Revenue Management • Data Quality • Dynamic Pricing …thanks to an internal and operational Platform (PaaS) Developer Platform Monitoring Dashboards (python, R scripts) Operational Platform API Catalog
Suppliers Car Pricing Price Inducers Financial Press Review Fraud Detection Recruitment Chatbot Home Insurance Benchmark Regulatory Chatbot Price database Tariff Optimization Weather & Climate Data Capture E-Reputation Tourism Electric Vehicle Impact CAPTURE Text Categorization VISUALIZE INTERACT
Mortgage Rates Benchmark Translator MOOC Search Algorithm Parking Availability Deep Learning Image Recognition RegWatch Risk Area for Package journey Hybrid Forecast ANALYZE PRICE PREDICT
INSURANCE PRICING Include market data in your pricing model Market prices are a leading piece of information in a competitive environment Systematic use of the Internet to compare offers, coupled with the possibility to henceforth terminate policies at any time, may generate a competitive shock in the individual insurance market. Knowledge and consideration of competitors’ prices may become necessary to not lose ground to market players holding such information. Sia Partners has developed an innovative data capture technology to collect Internet prices and a high expertise in including this data into the pricing process. Sia Partners provides its clients with its technology in order to assist them with their pricing strategy within this new competitive environment. CAPTURE An innovative technology designed for a competitive market PROSPECTS • omparing prices on the internet is already a common practice that will become sysC tematic. • ree cancellation may translate into substantially higher cancellation rates, up to a F potential doubling. OUR DATA CAPTURE SOLUTION 1. Collecting: prices are extracted through our data capture technology from online quoting websites and price-comparators (car, home, health, loan insurance,…). 2. Storage: the collected information (prices, guarantees, offer, options) are stored and historicized within an internal database. 3. Analysis: the pricing model and the business strategy are decrypted (pricing structure, variables impact & interactions, zoning, vehicle clustering, exclusions, positioning…). 4. Application: models and algorithms allowing the consideration of external data have been developed in order to include market data within the pricing model. Which applications ? Sia Partners provides its data capture technology and its technical expertise within packaged pricing offers. Pricing audit Pricing strategy • Technical benchmark against market models (pricing variables, interactions, impact levels) • Pricing and business strategy analysis • Modelling price elasticity • Including market prices (position, pricing gap,…) • Building a customer value algorithm Technical pricing Strategical follow-up • Building the pure premium model • Including market models (zoning, vehicular,…) • efining strategical segments and target positioning D • ollow-up of the strategy with respect to the market prices F
Data Capture – Use Cases Geographical segmentation Geography is a highly discriminant feature in car insurance. Taking geography into account can be performed through a zoning which assigns a pricing modulation coefficient to each zip code. A robust estimation on each of the 6 300 zip codes seems barely feasible in regards to the required data volume. Market approaches mainly head towards a residuals spatial smoothing or a demographic-based classification (density, agglomeration type,...). However, these approaches lie on strong and hardly verifiable assumptions. Data capture makes the zoning building reliable through external data integration with the market zonings. Data capture may thus be seen as an indirect means to access other players’ portfolio information. Technical zoning • Mapping the technical model residuals • Building the zoning with spatial smoothing (Gaussian kernel) Competitors’ zonings • Selection of market zonings (key players, similar guarantees,…) • Extracting and reconstructing the retained market zonings Benchmarking • Identifying the deviations between technical and market • Investigating the gaps on robust points Joint zoning Technique & Market Application of a credibility model based on the following principles: • ach zip code coefficient matches a weighting between technical and market zonings’ E coefficients. • he weighting depends on the robustness of the technical zoning estimated from the T company internal data. Price elasticity model The price impact on the customer’s behavior constitutes a critical piece of information for the gross premium positioning. The customer behavior is taken into account within the pricing process through a price elasticity model which aims to describe the subscription, cancellation and renewal dynamics. Sia Partners has developed a price elasticity model that includes the market prices from the data capture on one of its customers’ home insurance product. Variables relative to pricing positioning are displayed among the 10 most explanatory variables of the cancellation model. Moreover, the model analysis highlights an increasing importance of these variables over time. The customer behavior evolution and the cancellation facilitation lead us to conclude this trend could continue growing. The developed model can predict the impact of the pricing adjustments on the cancellation rate of every customer profile, including taking into account the competing offer in the market. The model has been used in order to aim for two specific profiles during the annual revaluation : • rofiles which are strongly sensitive P to the price for which any increase could translate into a substantial rise of cancellations. • rofiles which are poorly sensitive to P the price and then likely to go through a price increase without any substantial impact on cancellations. Data capture reveals its potential within the gross premium positioning through developing quantitative models taking into account the other market players’ impact.
Fleepit Digital © 2021