+41 44 500 21 28

2015-01-janus_1The Quantum Matchbox

Matching unrelated digital records to resolve customer identities

[Download PDF]

Millions of customers or millions of strangers?
A global transportation company, with millions of customers, accumulates a huge amount of data – but not enough information. Online bookings, newsletter subscriptions, loyalty programme, ticketing system: customers routinely use a range of different channels, resulting in a deluge of unrelated digital records and a collection of databases. In fact, so many that the customers’ identities have become elusive. Which is of course a problem when you want to offer the best possible service to them.

How can the company understand which of the millions of existing disconnected profiles, collected over the years, belong to the same person?

 

When the math seems to work against you

Customers can, and do, book tickets with or without loyalty programme information, providing minimal personal details, or even through intermediaries. They may have separate profiles and subscriptions for the online booking page, the newsletter and the loyalty card. And of course, over the years people move, have different needs, adopt new behaviours, get new loyalty cards and change phone numbers. All these disparate profiles, which may include name, birth date, address, id card number, phone number, email address and more, must be compared to each other to find real identities of real people.

However, when comparing a million records, there are 500 billion possible combinations. On top of this, names can be spelled in many ways, phone numbers come in different formats, people have multiple addresses, not to mention use different alphabets such as Cyrillic or Chinese. So not only are there too many comparisons to perform with standard methods – they are also riddled with uncertainty.

 

Quantum’s Matchbox: clever algorithms, serious performanceQuantum Matchbox - clusters 

The solution Quantum designed and implemented to solve this problem is called the Matchbox. Quantum’s data scientists created record-matching criteria based on carefully designed tests, in order to reduce the sheer computational volume while preserving quality. This smart filtering, coupled with a highly parallel environment, guarantees the computing performance required by the task.

A set of re-usable applications cleans and enriches the input data – giving the Matchbox great robustness in dealing with records of varying quality. Standardisation of email and world-wide postal addresses, phone numbers, names; transliteration of non-Latin alphabets; assignment of gender; conversion of unstructured data into structured records: these are some examples of mechanisms that allow the Matchbox to digest nearly any input.

We understand that different companies use different technologies. This is why Quantum took great care in creating a solution that is compatible with a wide range of IT environments. The cloud-based version does not need any dedicated on-site IT environment at all, thus avoiding any strain on scarce company resources.

 

From digital records to just people

Match-clusters-v03

The transition from having millions of unrelated digital records to knowing the people behind them allowed the client to understand its customer base: who has bought which products, where the best customers are, the level of service they experience, how behaviours evolve and individual preferences change. Having a full picture of their customers means being able to offer them better and more relevant services. It also helps in detecting fraudulent behaviour and in designing products. In short – it’s key in business.

 

About Quantum 

Quantum is a data science and analytics company, located at Technopark in Zurich. We help clients to identify their most valuable customers, products, or services; determine potential risks; discover hidden potential in their markets; pinpoint and eliminate bottlenecks and inefficiencies; and provide other insights to steer their business. We do this by combining business experience and knowledge with the application, implementation and teaching of scientific methods of data analysis, data management, reporting and modern visualisation to turn data into information.

 

Discover what data science can do for you
Learn more about how modern data science can help you and your business or contact us at info@qbis.ch.