Show me the data!
Organisations, in general, have access to huge swathes of data. And getting the ‘new oil’ is not necessarily the problem. The difficulty is making use of that data; whether that’s direct customer or IoT device data. How can I best use this precious asset and put it to work?
The most important thing any organisation can do, is come up with an effective and holistic data management strategy — prioritise it!
First, you need to understand what data you have; second you need to be able to analyse that data; and third, you need to be able to apply intelligence on top of the data “to either drive other types of experiences, or computations or workflows,” says Simha Sadasiva, co-founder and CEO at Ushur.
To achieve this comprehensive data management strategy, a partner is needed. This ally should enable enterprises to not just look at the data that they have, but also drive automation, through artificial intelligence and machine learning techniques, to analyse and complete missing data; “either by interacting with different constituents or by showing them insights into how proximate the customer is based on the data they currently have in their back office,” continues Sadasiva.
Sources of data
Traditional sources of data are those that exist in enterprises, and have done so for a while. This can include data management or SQL databases, which can be structured*, or unstructured** document databases, such as “things like Mongo,” explains Sadasiva. This the data that exists in the back of the enterprise.
But, in today’s connected environment, there are data sources coming in from constituents that interact with enterprises.
“Think about end users, think about agents, think about business partners, even employees to some extent. They are contributing to data source in the form of photos, images and videos etcetera,” says Sadasiva.
These newer sources of data — such as a customer/employee receipt — need the artificial intelligence techniques to be made sense of. For example, using optical character recognition to look at a receipt, organisations can automatically extract the information and bring it back into a data warehouse at the backend of the enterprise. “That requires tremendous amounts of capabilities and also the ability to securely transfer that information to the back office,” says Sadasiva.
The data journey: It’s only the beginning for digital transformation — Big Data LDN
The right kind of data
From an automation and machine intelligence perspective, the deep learning techniques that are used can analyse all these types of data that enterprises currently possess — human sources or historical ones. Organisations need to combine this underlying data set with supervised and unsupervised learning.
Ushur, for example, has created tools that take advantage of the data that exists in the core enterprise and acts on that supervised and unsupervised learning by leveraging the data set enterprises already have.
There are many different types of data that can be analysed, from an even greater number of sources. But, there’s a problem that occurs, although perhaps not as relevant for the backend of enterprise.
Feeding the wrong type of data, or biased data into systems can lead to negative results that harm a business or institution. You don’t have to look further than Amazon’s sexist AI recruitment tool that the tech giant ditched last year. Or, that in 2016 it emerged that US risk assessment algorithms — used by courtrooms throughout the country to decide the fates and freedoms of those on trial – are racially biased, frequently sentencing Caucasians more leniently than African Americans despite no difference in the type of crime committed.
AI researcher Professor Joanna Bryson, said at the time: “If the underlying data reflects stereotypes, or if you train AI from human culture, you will find bias.”
Data and AI recruitment: A risky business?
The way round this is by snuffing out stereotypes and bias, and making sure that the data reflects this.
“It’s all about how much data that you have as an enterprise, what can you do with that data and how can you make that actionable” — Sadasiva
Unstructured** vs structured* data
Structured data typically referring to databases, such as SQL. This structured information that can be organised, either based on customer information or specific types of information about a specific business problem.
In contrast, unstructured data can be a “big blob” of textual information. “It could be problem statement that a customer has described, through an email or pdf document, for example,” says Sadasiva. “Inside that can be nuggets of structured information, such as customer name, phone number, claim number, policy number or credit card number.”
The key to harnessing unstructured data? Visualisation
This type of information is referred to as semi-structured data, which can be buried in unstructured information. Having the ability to extract this semi-structured information from unstructured data requires fairly state-of-the-art ‘artificial intelligence’ techniques.
Extracting this semi-structured data is a growing trend for organisations that take their data seriously
A business case for data
The best way to show the business case for good data management is with an example: the email overload.
Large enterprises receive tens of thousands of emails from customers. “To manually go through those for segregating, triaging and sending that email out to the right department and right person before going back to the customer is ridiculous,” says Sadasiva. “It’s a first level problem that takes a significant amount of manual work.
Solving this problem requires the ability to extract and separate out the semi-structured data present in the unstructured information of an email. “A broad set of data tools can be used to extract this customer information and automatically you can triage that out to the right person and take action on that text. That’s a simple example of how you can apply automation on top of data that an enterprise already has,” continues Sadasiva.
“Most enterprises have millions of bytes of unstructured information, in the form of emails, problem statements, articles or information sources that they can tap into.
“This is legacy information that they already have. And by applying data science to it, enteprises can make it useful for training computer models, whereby they can actually reduce the amount of manual work for the future.”
Making use of your customer data: a new way of thinking
It’s all about micro-engagement, according to Sadasiva, which is short snippets of interaction that are back and forth between consumers and enterprises. When businesses start to look at every interaction with a consumer or a customer as a micro-engagement, it leads them to rethink the entire customer journey and segment it between prospecting, on-boarding the customer, supporting the customer, up-selling, cross-selling and maintaining that relationship with the end user.
There are various interactions of engagement that can be had during the customer journey. By applying data science to interact and collect that information from the customer, real insights can be gained.
“This is somewhat of a new age, a new way of assessing how proximate a certain brand is to their customer, and how proximate the customer feels towards the brand,” says Sadasiva.
Applying a combination of artificial intelligence, data analytics and machine learning (etcetera) on this micro-engagement level to consumer data can transform customer engagement.
