+ All Categories
Home > Technology > Pinar Donmez - Kabbage at the Chief Analytics Officer Forum, West Coast

Pinar Donmez - Kabbage at the Chief Analytics Officer Forum, West Coast

Date post: 16-Jan-2017
Category:
Upload: corinium-global
View: 54 times
Download: 0 times
Share this document with a friend
35
The Impact of Big Data and non- traditional data sources in online lending Chief Analytics Officer Forum West Coast May 26, 2016 San Francisco, CA By Pinar Donmez Chief Data Scientist at Kabbage, Inc.
Transcript

The Impact of Big Data and non-traditional data sources in online

lending Chief Analytics Officer Forum West Coast

May 26, 2016 San Francisco, CA

By Pinar Donmez

Chief Data Scientist at Kabbage, Inc.

Agenda

Purpose and Objectives

Turning Big Data into Big Value

Value of Big Data in Understanding Customers

Kabbage Data Centric Approach to Lending

Use of Big Data Analytics to Manage Future Customers

Q & A

Goal:

i ) Deliver a message explaining the value and importance of big data technologies with supportive examples

ii) Discuss how big data is leveraged at Kabbage

Agenda

Purpose and Objectives

Turning Big Data into Big Value

Value of Big Data in Understanding Customers

Kabbage Data Centric Approach to Lending

Use of Big Data Analytics to Manage Future Customers

Q & A

Turning big data into actionable insights will be the key innovative drive for the foreseeable future

Big data spending will grow to $50billion by 2019, with 23% annual growth rate

Turning big data into actionable insights will be the key innovative drive for the foreseeable future

Big data spending will grow to $50billion by 2019, with 23% annual growth rate

Exponential data growth

• 90% of the world’s data was created in past 2 years

• By 2020

• 10 times more mobile data

• 19 time more unstructured data

• 50 times more product data

Turning big data into actionable insights will be the key innovative drive for the foreseeable future

Big data spending will grow to $50billion by 2019, with 23% annual growth rate

Exponential data growth

• 90% of the world’s data was created in past 2 years

• By 2020

• 10 times more mobile data

• 19 time more unstructured data

• 50 times more product data

Ability to answer ‘What’, ‘Why’, ‘When’, ‘How’ questions

In 2013, only 13% of companies had extensive predictive analytics. By 2016, 70% of most profitable companies will manage their business by real-time predictive analytics capabilities.

Turning big data into actionable insights will be the key innovative drive for the foreseeable future

Big data spending will grow to $50billion by 2019, with 23% annual growth rate

Exponential data growth

• 90% of the world’s data was created in past 2 years

• By 2020

• 10 times more mobile data

• 19 time more unstructured data

• 50 times more product data

Ability to answer ‘What’, ‘Why’, ‘When’, ‘How’ questions

Analytics of things

• IoT will produce more data than ever

• Tremendous opportunities to understand, connect, serve and learn from all possible interactions

Big Data turn into actions even when least expected

In Japan, dairy farmers used wearable and big data technologies for optimizing successful insemination

Artificial insemination success rates today are 70% with a pregnancy rate of 40%

If you successfully detect when a cow is in the heat, then pregnancy rates rises up to ~70%

Question: How can you detect when estrus begins?

Big Data turn into actions even when least expected

In Japan, dairy farmers used wearable and big data technologies for optimizing successful insemination

Artificial insemination success rates today are 70% with a pregnancy rate of 40%

If you successfully detect when a cow is in the heat, then pregnancy rates rises up to ~70%

Question: How can you detect when estrus begins?

Researchers found cows start walking around furiously when they go into heat. They monitor the steps a cow is taking. When the level spikes, the system knows estrus has begun.

Fujitsu solved this problem by Gyujo: sexiest system in the cloud

A set of smart, connected devices send alerts to the farmer when estrus begins

Agenda

Purpose and Objectives

Turning Big Data into Big Value

Value of Big Data in Understanding Customers

Kabbage Data Centric Approach to Lending

Use of Big Data Analytics to Manage Future Customers

Q & A

Netflix understands what its subscribers enjoy watching through big data analysis

Netflix captures

what they watch

when they watch

where they watch

what they search

what rating they give

They also track

plot, genre, character development on movies/shows

color, tone, scenery, other visuals, volume levels

geo-location

where they pause, fast-forward, rewind

The use of big data at Netflix results in hit shows with original content that hooks us all, bringing millions more new subscriptions and revenue sources to the company.

Netflix knew subscribers watch movies starring Kevin Spacey and directed by David Fincher More importantly, they knew the same people loved the 1990 BBC original “House of Cards”

Netflix knew subscribers love dark comedies, plots revolving around prison/crime, and a likeable female lead

Texas Hospital predicts the 30-day readmission risk of heart failure patients

Texas Hospital predicts the 30-day readmission risk of heart failure patients

Analytics software has helped the hospital to cut its 30-day readmission rate for nearly half, from 23% to 12%

Agenda

Purpose and Objectives

Turning Big Data into Big Value

Value of Big Data in Understanding Customers

Kabbage Data Centric Approach to Lending

Use of Big Data Analytics to Manage Future Customers

Q & A

Traditional lending is becoming obsolete

Time-consuming

Hard to work with banks

Sub-prime population

Take weeks to get access to cash

Kabbage difference lies in the utilization of in a variety of atypical data sources

Credit history

Detailed sales transactions Payment processing data

Checking account activity

Social data

3 K’s of Kabbage Lending Kapacity: Does the business have enough capacity to handle the loan?

▪ Sales growth ▪ Transaction volume ▪ Account Balance ▪ Revenue vs

expenditures

▪ Past delinquencies ▪ Public records ▪ Bankruptcies ▪ Reviews ▪ Likes ▪ Ratings

Kharacter: Does the business show reliable characteristics?

▪ Store age ▪ Repeat sales ▪ Buyer retention ▪ Reputation

Konsistency: Does the business consistently perform well?

Main Concerns Focus Kabbage’s situation

▪ 12 node Hadoop Cluster ▪ Xlarge instances ▪ Nightly importers ▪ Daily jobs analyzing/scoring customers

Size ETL pipeline

Hadoop

▪ Deeply nested information hierarchies ▪ Dozens of file formats ▪ Inconsistencies and gaps in the data

Challenges Data Types

Scale Complexity

Acceleration & Growth

Kabbage Data Architecture

▪ ~5M new data files every day ▪ Terabytes of raw transactional data ▪ Very complex, rich, (un)structured data sources

Data Mining

& Machine

Learning

Risk & Underwriting

How can we identify and manage risk?

Examples

- Credit risk modeling

- Fraud Detection

- Business sustainability forecast

- Seller-Buyer relationship

Collections Can we have a more optimal Collections strategy?

- Marketing analytics

- Pre-approval models

- Response models

- Collections model

- Payment reversal prediction

- Collection optimization

We solve problems that will ultimately help us better understand SMBs and serve a fingerprint product to them.

Problems we tackle at Kabbage

Marketing How can we spend wisely?

Social data is a particularly interesting area for us to use in risk and underwriting

4

5

6

3

2

1

Straightforward social clues Ratings

Followers

Likes

Reviews Comments Posts

Sentiment analysis

Sophisticated text mining

Content extraction

Collective Knowledge

Ensemble different pieces of information

Search for a compelling story

7

8

▪ Intuitive yet valuable correlations

– More reviews are correlated with more sustained businesses

– More reviews + high ratings correlate well with credit worthiness

– More likes, more followers, etc. show better growing trajectory

▪ Informative insights from text

– Text mining: “…wrote $20K in bad checks…”

– Sentiment: “…stay away, total scam…”

– Sentiment: “…poor customer service, terrible communication…”

Examples of Key Findings

Agenda

Purpose and Objectives

Turning Big Data into Big Value

Value of Big Data in Understanding Customers

Kabbage Data Centric Approach to Lending

Use of Big Data Analytics to Manage Future Customers

Q & A

Financial institutions use big data to understand customer relationships

Banks monitor customers’ journey through

websites, call centers, branches, …

use the data to predict churn or purchase of a financial service

Connected Sales channels

Various sales channels communicate with each other so that a customer who starts an application online but does not complete it, could get a follow-up offer in the mail, or an email to set up an appointment

Targeted offers

Cash-back deals based on where customer has made payments in the past

Financial firms can leverage big data to predict and avoid customer churn and do better staffing

Four elements are important for developing advanced analytical capabilities: People, Data, Intent, Tools

Lack of relevant data technologies is the biggest reason why organizations cannot leverage big data analytics. Data storage and format problems follow.

Where financial orgs should invest $$ in big data for big gains? Create customized, consistent customer experience Social media plays a big role in this goal “…Jon and Cheryl’s FB page reveals they just had a baby. Company sends them life insurance offers…”

A unified, single data view of the customer Combine all data into one place, make it accessible to a large audience “…Mortgage dept. just closed a deal with Mr. Smith. Insurance department is unaware, lost cross-sell opportunities…”

Data insight flows to the right people at the right time The information flows in real-time “…Jane is currently in the branch and looking for a bank product. She gets a valuable offer from the teller right there and then…”

Enhance customer relationships through shared values “…Oliver only purchases organic produce at the grocery store. His credit card company sends him a letter showing their commitment to a sustainable environment…”


Recommended