Current Trends Technologies in Enterprise Analytics
Kamran Ashraf, vice president and head of analytics of Visa Europe, discusses why analytics has become today's modern IT buzzword and why such a system must fit the end user employee need and not just IT.
[Editor's note: This article was originally published by Enterprise Management 360° and is reprinted by permission.]
EM360°: There are many different forms of analytics in today's market and it seems such a convoluted sector with so many different tools and concepts around today. Do you agree?
Kamran Ashraf: Historically, we have had some very big organizations such as SAS and SPSS where they have provided great infrastructure and powerful tools. However, these are like your industrial tools and so you need to be skilled in using them — like learning a special language, such as SQL (structured query language), to actually write queries in code. Over the last five to 10 years, across the consumer world and business world, you've had [many] new enterprises which are making technology available for the masses so people can go and engage with technology and data to get the answers they want rapidly. What needs to be understood by IT managers is which niche part of the value chain is this supplier or technology saying it has a Unique Selling Point' in delivering and then figuring out how they can be plugged into a coherent data architecture.
You have had the enjoyment of testing out analytic functions in your career from the likes of predictive enterprise, customer experience, advanced, and big data analytic solutions. So which functions fit which organizational enterprise?
I'm in the lucky position working in Visa Europe because my role isn't confined to a single market or a single sector; I'm allowed to work across many different markets consisting of around 90 projects a year, so I can see the ins and outs of many organizations and High Street financial services companies, but also retailers as well. The advice I would give anybody if they're starting this journey or looking to understand where to go next is that they should be working with their IT partners to improve their business intelligence (BI) maturity model. This framework will help IT and the business agree where the business is today (the as-is situation), where it needs to be, and the gaps that need to be filled in terms of data, infrastructure, tools, and skills.
The more specific answer is there are three sets of functionality that most organizations need. One is some reporting functionality in which most people have rebranded as "dashboards." The ability to create dashboards is important because this gives you a fundamental overview, the same as if you were driving a car.
The second thing is to understand the customer better in terms of their historical behavior and predicted future. This is where you need descriptive analytics like segmentations, propensity models and so on.
The last things organizations need are big data solutions, especially [those with] the ability to take in external data sets. For example, taking a feed from external data would enrich the understanding of shopper behavior when there is a lack of understanding about demographics or attitudes to get a fuller picture of the customer.
There are some voices in the industry that are saying systems should adapt to real time communications. Do you view real time as a core business value an organization must obtain?
I think many people are skeptical about real time because it was advanced and tried a few years back but a number of companies found that near real time was still reliable enough for them.
However, for most online or mobile companies real time analytics is vital and there are a couple of fundamentals around trying to test the things that you're trying to do in real time. For example, you need to be able to perform A/B tests. For instance, if customers are coming to my clothing website because he/she just seen the latest catwalk in Milan, we need to be able to host different offers that compete against each other (e.g., a price discount or a "complete look" offer) and the response rates help us determine a champion offer that we roll out to wider segments.
The second thing that has to happen is you need to be able to make sure these offers are happening in real time using the customers' preferred channel. For example, as you surf the net, you don't want to see an e-mail in your e-mail account a day or two days after the shopping experience trying to do is over and done with.
You mentioned in July 2013's "Big Data for Financial Services" event run by IQPC that Visa moved over some 200 million customer records in the space of four months. Can you tell us about that?
I am part of a consulting team working on projects for banks and retailers in order to help them improve the service to their customers. In one project, a product team wanted to improve their understanding of customers but we found that customer data actually didn't exist in a single place: It was not available for the business to analyze a customer in a complete way.
You have a risk team, fraud teams, and other teams working in separate silos. You have account information and channel information about branches or online somewhere else, too.
We agreed to execute a data migration project to put these disparate records in a secure environment within the organization where it could be better accessed by more teams and anonymized into aggregates, business metrics, and key performance indicators (KPIs) that don't need to be re-invented by each analyst using data. It's really important to have agreements on data dictionaries and to see a full historical view as variable definitions. Too narrow an analysis window with too few data points would be a shaky foundation to make decisions on!
How do you think customers would then perceive what Visa has undertaken in terms of its approaches in security and compliance?
Well, hopefully positively -- and I say that as a consumer and not as a Visa representative. When Visa first started its journey in terms of trying to understand its data and its assets, the first project that we started was actually all around governance because we want to make sure that under no circumstances will data reside in a place where it can be accessed and authorized in a way that we do not want it to be used, from a regulatory compliance and a responsible corporation point of view.
We have a companywide governance structure and enterprise data committee that looks at the way data is used and [at the] controls [used] to ensure the use of data complies with regulation, with competition law, with data protection, and with people's personal data being protected to such standards as PCI DSS.
What do you assess then are the most significant challenges you face being in the forefront of the analytics space?
I think there are four challenges -- and I'll use this acronym: VIVA.
The first term is virtualization and the choice to keep data in the data warehouse. Virtualization technology needs to be added because it allows users to access the data directly without having to put it through a traditional BI process (ETL -- extract, transform, and load) into a data warehouse and data mart, which could take months to organize and govern.
The second term is the question around integration. This is not just about integrating the data; this is also about integrating teams. Originally, you may have had only one central team doing analytics or a BI team that used to do all the reporting on all the dashboards. That is probably not going to work going forward, so you need to have an integrated team approach whereby you're empowering the end user. Users no longer want to 'outsource' the analytics to different teams and then not understand what's coming back from the data; they want to learn as much from the data as possible to improve their strategies, plans and actions.
The next V is visualization and dashboards, which are becoming increasingly important. Tableau Software, for example, has had success in their adoption as the level of data organizations carry now is so vast that business people higher up in organizations want to see a simple dashboard that views key performance indicators in one or two pages. If people want to analyze further, they will instead use different technologies.
The last challenge is about applications. Having business applications that use the power of the data to tackle business problems is the best way to "monetize data" -- to reduce churn, improve acquisition, increase cross selling, etc. Ultimately, you need to get improved value propositions and service out the door to opted-in customers to give them a more personalized service.
Your points on applications really interlink with other concepts in the infrastructure management space. Do you need applications that will run across the cloud?
I still do think there is hesitancy in people wanting to put things in the cloud. Just to give some industry-level context, data has been in a physical environment for many years in servers, warehouses, and so on, and there have been data breaches. Yet the standard people are putting on migrating to cloud technology is unless you can prove there is never going to be breach, we're never going to put it in the cloud -- which is too high a standard to achieve. The practical approach is "Does a cloud based approach improve security, access, cost, etc." On [numerous] factors, cloud-based solutions significantly outperform traditional approaches, so I think it's just a matter of time before people really migrate to the cloud big time and that will be another step change for analytics.
Do you have any final thoughts that you would like to resonate across enterprise decision makers within their realm of analytics?
There are a couple of things really. The first thing is that there are a number of engagement techniques that BI teams deploy which actually "turn off" the business. Business users are told, "Here's your tool and here's your technology; if you want anything else it's got to fall into our BI road map," which is a long haul plan of how we're going to plan out delivering requirements."
BI road maps are effective in terms of understanding the business requirements and the technology stack that's needed to address the need, but it isn't always implemented or kept up to date. There are so many new players emerging every six months, if you have a five-year plan, it's going to be out of date within nine months, guaranteed. A BI road map should be used as an ongoing, refreshed engagement tool and not rigid like "If it's not on the road map, it's not going to happen" because flexibility and agility are proven needs for businesses.
The next technique, even if you're deploying enterprisewide technology, is to be clear about your goal when integrating analytics. There are some tools that will only go to 5 percent of your total employee base. You should not try and over-engineer enterprise wide solutions that attempt to fit 100 percent of employees but really are only designed for 5 percent of the employees. You need to be more segmented with the approach to deploying enterprise technology. There are some technologies that should be segmented for data scientists, some for super users and some for casual users. An industry analyst, Wayne Eckerson, has written about how BI systems have really been designed for the lowest common denominator for people who are the lowest power users, and not for the one's who get the most insight -- the super users.
The point is not to get hooked up on the word "enterprise." Let's understand it as about serving people within the enterprise. It's not just about having a generic enterprise view. One size doesn't fit all and I think that's where the BI industry has somewhat found themselves a little bit trapped as they are asking users to play golf with a single club!
That's really the philosophy I push when I talk to CIOs, and that's what I recommend people think about.
Kamran Ashraf is vice president and head of analytics and information services at Visa Europe. He is a cards, retail banking, and retailer payments practitioner, having worked for 16 years across regulatory, finance, marketing, consulting, and analytics divisions. He has led over 100 projects at top European banks and retailers including RBSG, Yapi Kredi, Carta Si, DKB, Hilton and Tesco. He is also responsible for the commercialization of data through the delivery of data-analysis-intelligence solutions across 36 European markets. Kamran's projects have won industry awards in 'CVP Development' and 'BI Business Transformation' categories competing against the likes of Tesco, Telefonica and Disney. He has a BSc (Hons) in Economics from Queen Mary & Westfield, University of London, and additional qualifications with the Chartered Institute of Bankers and Chartered Institute of Securities and Investment.