Bancography | Branch Planning, Marketing Research, Brand Strategy, Products & Profitabilty

The Empty Promise Of Big Data

Aite Senior Analyst Ron Shevlin knows a thing or two about data analytics. He’s crunched numbers in the banking industry for 20 years, and he says financial institutions should purge the term “big data” from management discussions. Here Shevlin outlines three reasons financial marketers should dodge the “big data” trap.

By Ron Shevlin, Senior Analyst with Aite Group

Claims of the potential impact of “big data” are everywhere. One technology vendor writes:

“The value in big data is in being able to aggregate new information sources into a unified platform to gain new insights and knowledge. With [our big data technology], financial services firms can: reduce risk by acquiring real-time visibility into transactions and complying with current and future regulations; cut costs by simplifying IT infrastructure and quickly building big data applications; and increase revenue by spotting trends before competitors and developing new products that meet customer demands.”

Reduce risks, cut costs, simplify IT and increase revenue? These claims are typical among many of big data’s proponents. What’s missing from most (if not all) of these claims is any logical or empirical proof of results. The promise of big data is nothing more than that — just a promise.

There are three really good reasons why smart financial services executives should avoid getting caught up in this “big data” hype:

  1. The real drivers of marketing effectiveness
  2. The real challenges of customer data integration
  3. The real cause of the data scientist shortage

1. The Real Drivers of Marketing Effectiveness

Yes, it’s true that analyzing data can produce insights that can help produce improved marketing results. And I’ll concede that incorporating new types of data, and in more timely ways, could improve the quality of insights.

But… the analysis process — call it “Sensing,” the ability to identify consumer needs and intentions based on their behaviors and actions — at best only partly determines marketing effectiveness. Another determinant of marketing success is “Responding” — delivering appropriate advice, guidance, and offers at the right time and in the right channels.

Financial services firms can analyze data — big data, new data, whatever data — until they’re blue in the face to generate insights about customer needs and market trends. But without the ability to reach customers and prospects with appropriate messages, marketing ROI will not improve.

In other words, identifying needs in real-time makes absolutely no difference without the ability to get corresponding messages to the appropriate customers in real-time. If financial institutions are going to “increase revenue by spotting trends before competitors” (as promised by so many big data proponents), then they need real-time methods of contact through mobile devices and social media. And yet how many financial institutions use snail mail as their only official mode of communication?

But you don’t hear the big data proponents talking about this, do you?

Getty Images | Content Marketing

2. The Real Challenges of Customer Data Integration

Similar to the “Sense” and “Respond” concept in the previous section, there are two mechanisms involved in data analysis: putting data in, and getting data out. The former is (relatively) easy, the latter is really hard.

One of the top 10 US banks went through a “customer data integration” effort a number of years ago, integrating data from a number of business lines in an attempt to create a 360-degree view of customers. The bank achieved that objective by building a technology infrastructure pulling data from across the organization, but in doing, but they lacked the ability for users to easily access the data.

As a result, marketers in the various lines of business were forced to submit data requests to IT for campaign sizing and segmentation. These requests often take weeks (if not months) to complete, hampering the bank’s ability to rapidly respond to changing conditions and customer demand in the market.

The real challenge to customer data is not incorporating new types of data, but in making what is captured useful. Who cares if you have petabtyes of data if no one can access it and make any sense from it?

Maybe that’s where you think “data scientists” will come in. Don’t hold your breath.

3. The Real Cause of the Data Scientist Shortage

There is a common misconception among many big data proponents that a new organizational role — so called “Data Scientists” — will emerge alongside the boom in big data analytics. What is a data scientist? According to eConsultancy, a data scientist…

1) May be involved in the design and development of systems that collect and process large amounts of data using tools like Hadoop and programming languages like R

2) Need to have a deep understanding of statistics and probability

3) Are capable of designing and testing predictive models

4) Provide the greatest value by answering the questions “Where are we likely going?” and “What would we need to do to go somewhere else?”

5) Will realistically need to acquire a high level of domain expertise

That sounds a lot like the database marketing analytics professionals that exist in many financial institutions today. eConsultancy tries to distinguish between data scientists and analytics experts by saying that analytics experts:

1) Analyze smaller or more specific data sets typically collected by third-parties

2) Primarily use existing services and apps for data visualizations

3) Do not require a formal scientific background

4) Answer questions like “Where have we been?” and “Where are we today?”

5) Should have some domain expertise

The analytics folks I’ve worked with easily meet all the criteria in the definition of a “data scientist” (with the possible exception of having expertise in Hadoop). They’re hardly limited to small(er) or specific data sets, or data collected from third-party tools, or using “existing” services. Most of the analytics experts I’ve worked with have a PhD in statistics, which should meet most people’s definition of a “formal scientific background.” And the analytics experts I’ve worked are generally focused on the question “What should we do?” not “Where have we been?”

The point here is that the people who are most likely to help your organization analyze big data already exist within many large financial institutions, and within their marketing services vendors. They are folks with advanced degrees in statistics, and have significant experience in market research and statistical modeling.

And the reason why there aren’t already hordes of data analytics people out there today? Because statistics is really hard, and most people don’t have any interest or desire in learning advanced statistical techniques. Crunching numbers just isn’t sexy.

The data scientist shortage isn’t going to be addressed if people don’t want to acquire the basic skill sets required for the job.

Getty Images | Content Marketing

The Big Data Management Fad

Management fads come and go. In the 30 years I’ve been working, I’ve witnessed at least five: Client/Server, Reengineering, Knowledge Management, Digital Business, and Social Media.

A common trait among all management fads is that they’re poorly (or loosely) defined. Management fads become fads, in part, specifically because they’re so loosely defined; it allows for multiple interpretations, giving many folks the impression that what they’re doing falls under the banner of the new fad. As a consequence, at the peak of any particular management fad, I see at least half of a firm’s new initiatives for a given year described in terms of that fad.

2013 is the year of the big data fad. Is big data totally useless? No. Like all the other fads before it, there are elements of value. In financial services, there are plenty of opportunities to make smarter business decisions by using new and different types of data.

But it will take years — yearrrssss! — for companies to develop and integrate “big data” competencies. The claims of big data ROI bandied around are unattainable in all but the most exceptional of cases.

The antidote for any fad’s hype is to focus on business problems and business value. Instead of asking “What can big data do for us?,” financial services executives should focus on asking “What are the most important business issues we need to address?” and “What can we do to alleviate those issues and capitalize on opportunities?”

Smart execs should ban abstract fads like “big data” from all management conversations, and force executives to talk in terms of concrete problems and specific technology solutions.


Ron ShevlinRon Shevlin is a senior analyst at Aite Group, a Boston-based research and advisory firm serving the financial services industry. You can pick up a copy of Ron's latest book, Snarketing 2.0, at Amazon.com. The book is a balance of keen, entertaining observations loaded with educational and practical advice that doesn’t pull any punches. Ron is a regular contributor here at The Financial Brand. You can read more from Ron on his blog, or follow Ron on Twitter.

Search For More: Data Analytics, ,

All content © 2014 by The Financial Brand and may not be reproduced by any means without permission.

CUNA | Youth Week 2014

Comments

  1. I agree and disagree with Ron on his take around ‘big data’. First and foremost, we can’t confuse perfection with direction. When you speak of not having the capabilities to reach out in ‘real time’ to take advantage of opportunities, I agree that banks are not well positioned to deliver text messages or communicate via social media. But, these institutions do have call centers and valid email addresses on roughly half of their base to enable rapid response communication. There is also technology available to target households digitally on a household basis if a buying opportunity exists. Finally, ‘snail mail’, while not the most effective channel for event-trigger opportunities, still generates a stronger ROI than programs using traditional modeling. Not perfection, but better than results today.

    So, while ‘big data’ is a big aspirational term, banks can do a much better job of focusing on collecting the better insights to deliver real time messages. And, in the meantime, they can use data (big, small or otherwise) much more effectively than is currently done. For instance, the data is available to determine which households in a customer base are shopping for a lending product daily. Yet very few organizations use this insight to acquire new loans or retain current loan relationships from their customers. This can provide a steady annuity of opportunities that most banks miss. The same is true for ‘quick hit’ programs like new movers acquisition programs and balance trigger programs that can lead to cross-sells.

    So, as opposed to thinking that ‘big data’ is a fad, I prefer to think that the term may help banks focus on common sense ideas they continually miss as they chase the next ‘shiny object’. If this buzz term gets banks to collect more email addresses, capture more cell phone numbers, utilize social media more effectively and implement programs based on logical triggers, so be it.

    Maybe with this foundation, these institutions will be in a better position to move to the next level, better integrating (and using) the current silos of data they have at their disposal and building the first wave of proactive marketing initiatives that can sense and respond, moving closer to SoLoMo initiatives.

  2. What a great, controversial blog post! I look forward to the comments.

    The critical aspect of “Big Data” that no one talks about is the “OK, now you know this–so what are you going to DO about it?” If you can’t take action in an appropriate time frame, what good is knowing anything in real time? Shevlin’s anecdote about snail mail response to real-time Big Data is spot-on.

    What’s most frustrating to me, though is how poorly companies contend with data integration. Having worked for marketing information companies (including Nielsen and Harte-Hanks), I’ve witnessed clients of all industries struggle to bring together data from different, incompatible systems. Central data repositories were the mid-2000s fad–and they still haven’t solved the problem of needing data that live in different places. It doesn’t have to be this way–but I’ve never seen a client make a concerted effort to solve this problem.

    Full disclosure: I’m the product marketing director for JackBe, a real-time business intelligence company. JackBe builds software that allows data from any system to be pulled into interactive dashboards–in real time. And it’s designed for business users to use and control, once the basic data connections are made and the security governance set up.

    I really believe that, for companies that surmount the “Where’s the data?” issue, Big Data is going to pay big dividends by freeing up resources to tackle the “What are we going to do about what we know?” dilemma.

  3. Shevlin, I love you but this article might have been more accurately titled:
    “You Aren’t Ready For Big Data Because You Value Hype Over Strategic Planning, Move Too Slowly, and Stereotype Technology”

    ==1. The Real Drivers of Marketing Effectiveness==

    Translation: Information will not help you if you don’t act on it.

    My take: Duh.

    My response: Of course you need the infrastructure, plan, and talent to be ready to act on the data. So, like any new project, have a plan and be prepared. Jeremy Howard’s piece on the Drivetrain Approach to building data products will help you get your mind right: http://strata.oreilly.com/2012/03/drivetrain-approach-data-products.html

    ==2. The Real Challenges of Customer Data Integration==

    Translation: Financial institutions are too much of a bureaucratic mess to make use of big data.

    My take: This has nothing to do with whether big data is a fad or not and everything to do with financial institutions standing in the way of their own progress.

    My response: This argument is like saying driving isn’t useful because you can’t be bothered to learn to drive. LEARN TO DRIVE.

    ==3. The Real Cause of the Data Scientist Shortage==

    Translation: Statistics aren’t perceived as sexy.

    My take: People fear or ignore what they don’t understand.

    My response: Robot magazine writers, the Google self-driving car, and Moneyball…data science is coming into its own sexiness. As there’s more opportunity for curious minds to dip their toe in the water of data science through competition sites like Kaggle and open education like Coursera’s Machine Learning course, i believe this will be an increasingly accessible field. It only takes a dip to begin to see the level of creativity and intuition alongside hard data anlysis wrapped up in data science.

    . . .

    Like any hyped technology (see: the last decade of bullshit social media conversations), big data won’t be useful to you until you come with a strategic objective in hand and the have talent and infrastructure to act on it.

    Don’t confuse industry/organizational ineptitude with a fad.

  4. We get back to the definition of ‘big data’, but if we define big data as the combination of both internal and external insight that could include transactional and/or behavioral data, there are institutions that are combining internal customer data (including purchasing information) combined with external credit bureau trigger insight, combined with compiled home valuation data to drive loan acquisition and retention programs. In addition, there are banks that are currently using ACH as well as check writing insight to build cross-sell initiatives.

    Just last week, I was with a mid-tier bank that had combined all of their internal data silos and have built an offer portal using a Teradata Relationship Manager. This offer portal uses account, balance, transaction as well as external insights. Finally, Bank of America is using purchase insight combined with geolocational information to provide offers with their rewards program.

    While I agree that there are very limited examples of ‘big wins’ with ‘big data’ there are some. The key remains to use what is available first and then expand as the potential value dictates.

  5. Besides fraud protection and various other security-related, non-revenue generating concepts, I’m still waiting to hear one really good idea about how financial marketers can put “big data” to work. Yes, financial institutions sit on mountains of data. The question is (and always has been): How can this data be used to build the bottom line (e.g., more loans)? Right now, proponents of big data are largely pitching an abstraction, a conceptual theory with vague potential.

  6. Brent:

    Big Data proponents claim huge ROI and promise significant results from investments in Big Data, stemming from the collection of new types of data and/or the analysis of data. Sadly, they offer no empirical proof.

    Determining which data elements are more important than others, figuring out how to use the analysis of data to develop offers and messages and sequence messages, and how integrating that data with existing data stores may be a big DUH to you, but since none of that is ever mentioned by Big Data proponents, I am compelled to point it out.

    Self-driving cars and robots are great, but have little to do with marketing-related claims regarding Big Data. Are self-driving cars and robots examples of Big Data? I can claim that the examples you list are NOT examples of Big Data. Since there is no generally accepted definition, you have no grounds to dispute my claim.

    Fidelity Investments tweeted the following:

    @Fidelity: Big data can unlock new insights for companies, governments and organizations.

    It’s this kind of BS that I’m railing against. Insights are not “locked up” waiting to be unlocked. Having more data or new types of data is no guarantee that new insights will be generated.

    BD consultants have an 18 month — and counting — window to sell their snake oil.

    Ron Shevlin

  7. Two great articles around the hype cycle and big data. I especially love the comment from CapGemini that says (as you did), let’s stop focusing on the technology and starting focusing on the outcomes http://bit.ly/UlTulo . Another great post from Gartner talking about big data falling into the ‘trough of disillusionment’ http://gtnr.it/10SIIbr. Talk about back peddling :-).

  8. Underwriting is a nice place to start.

    Zest Finance, founded by Google’s former CIO, uses alternative data sets to “extend credit to 25% more americans and increase repayment by 20%.” http://www.zestfinance.com/pdf/ZestCashHollerREV.pdf

    Wonga.com, not my favorite because it borders on predatory, used underwriting based on alternative data sets to issue over 100,000 loans, worth £20 million, earning about £15 million by in astronomical interest. http://www.wired.co.uk/magazine/archive/2011/06/features/wonga?page=all

    In 2010, MIT researchers used machine learning algorithms to predict credit risk and “Using conservative assumptions for the costs and benefit s of cutting credit lines based on machine-learning forecasts, we estimate the cost savings to range from 6% to 25% of total losses.” http://www.argentumlux.org/documents/CRisk_final.pdf

    We’re looking into it at Filene, and early early findings suggest we can use our hybrid scoring model to approve 1% more loans while reducing costs through default by 15%.

    . . .

    On the other examples:

    ////Narrative Science’s article-writing robots’ use of big data –

    “There is no shortage of data, in fact just about every company we talk to is drowning in data. As the volume of data continues to rise exponentially, companies need a better way to understand, use and monetize the data they already have. Narrative Science helps companies leverage their data by automatically creating easy-to-use and consistent narrative reporting”

    Their algorithms create original content for publishers like Forbes and the Big Ten Network.

    ////Google car using big data – http://www.cs.cmu.edu/~zkolter/pubs/levinson-iv2011.pdf

    “Improved perception and recognition algorithms now enable Junior to track and classify obstacles as cyclists, pedestrians, and vehicles; traffic lights are detected as well. A new planning system uses this incoming data to generate thousands of candidate trajectories per second, choosing the optimal path dynamically.”

  9. Narrative Science link: http://www.narrativescience.com/

  10. Jim is right: There is widespread disagreement about how “big data” is defined. I suspect “big data” is being used by industry analysts as a synonym for everything that’s new and interesting in the world of data analytics. It’s an interchangeable term, open to interpretation. Consultants in the data analytics space are all too happy to fuel hype in the “big data” discussion because it drags their company, their products into the forefront. Do they really care if banks and credit unions know what “big data” is and how to use it? No, not really… just as long as everyone is talking about data analytics.

    Jim cites examples of data analytics that are innovative and/or unusual. Brent cites examples where “alternative data” is used. But just because new and/or alternative data is being tapped, does that constitute “big data?” Where’s the scale? How is it “unstructured?” And where does “velocity” fit in?

    BofA’s reward scheme is a simple if/then search routine algorithm: If a customer bought from Brand-1 in the past and is within X miles of a Brand-1 or Brand-2 location, then present offer Y (from either Brand-1 or Brand-2). It’s a smart way to manage offers, but it seems like an example of geolocational marketing — not “big data.”

    And while combining data into a single system like Terradata’s “Relationship Manager” is certainly a smart thing for FIs to do, it strikes me as a classic CRM solution having little to do with “big data.” How/where are “vast seas of data” being analyzed?

    Wonga seems to have the model that most closely meets the academic, technical standards for defining “big data.”

    “[Wonga uses] a lot of social media and other tools on the internet you don’t even think about. That’s where the magic is. The crux of the algorithm is less about the individual pieces of data — your postcode, the colour of your car, how large your mortgage is — but how these pieces of information relate to one another. The data points are stacked against the other pieces of information gleaned from past Wonga clients. By 2009, Wonga had issued 100,000 loans. That’s 100,000 data sets contributing to an ever growing net of information, and each comprising 6,000-8,000 pieces of information about a borrower.”

    – William Shaw writing for Wired Magazine

    “It’s going to take a great deal of resources to provide a Big Data solution that will be proprietary, including technology such as search engines with in-memory analytics and a data collection and crunching capability. You need a big investment in enabling technology.”

    – Rodney Nelsestuen, Sr. Research Director, CEB Towergroup.

    But what Wonga is doing is not just a “big data initiative.” They are completely retooling the entire lending model — starting from the ground up. Taking on a data analytics project of this scale is simply not something 95% of most financial institutions can even remotely consider. It is smarter and safer to wait for their core data processing provider (e.g., Fiserv) to update the lending module.

    As suggested previously, the real promise of big data seems to center around lending and risk management — not marketing. Underwriting, fraud, defaults, credit worthiness… these are all issues with little- to no impact on marketers.

    Will financial institutions find some value in “big data?” Perhaps. Will it be the marketing department that finds that value? Almost certainly not. Will it be the marketing department that uses the insights from “big data?” Maybe, but how so and in what ways is very unclear at this time.

  11. This conversation seems to be more about the semantics of the term than anything. Would it be better to use the term “data products” and toss “big data” aside? Let’s do it. Then focus on which types of data products are useful, which are hot air, and how we can build on the good stuff.

  12. There’s more than just semantics at stake here. “Big data” is being defined in very specific terms (volume, variety, velocity), it’s just that there aren’t that many concrete examples, at least in the financial industry (right now). Presently, it seems other creative uses for data are being amplified into “big data” case studies.

    The consequences are significant, because — if big data proponents are correct — financial institutions currently lack the resources to handle (1) the massive streams of new data, (2) coming at them from all these various sources, (3) at a ferocious speed. In a nutshell, the basic argument is that “petabytes are the new reality, and FIs simply aren’t prepared to handle it.” So there’s lots of banks and credit unions out there hearing the “big data” story who are scratching their heads wondering if they need to invest in new, super-fancy über-servers and hire a bunch of PhDs. There’s more involved here than simply looking at data analytics in a new way.

  13. This is not about semantics.

    Database marketers (who have been dealing with large data sets for years) have been approached, over the past 20 years, by countless number of vendors selling “data” that they promise will improve marketing performance.

    The first thing smart marketers do is run a test to see if a new type of data or data source really outperforms the existing data and, very importantly, if the benefit of acquiring the data outweighs the cost.

    Big Data proponents are making unsubstantiated claims that all this new data that’s being captured will improve marketing performance. Show me the proof.

  14. What many “big data” proponents really seem to be saying is that banking will increasingly become a business of algorithms.

  15. This is an interesting take on the current hype cycle for big data. It clearly highlights the fact that there are execution risks associated with any new technology acquisition. However, I disagree with the underlying premise that big data solutions offer no benefits over earlier technologies such as data mining and business intelligence.

    For the first point, drivers of marketing effectiveness, Ron basically states that whilst big data analytics enables companies to gain customer insights, unless you know how to use that to achieve marketing effectiveness, it’s worthless. That is an argument demonstrating that it is necessary, but not sufficient to use big analytics to understand and address customer needs.

    It is for this reason that big data tends to be sold not just as a product but also packaged with professional services to ensure that firms are guided in how to gain value from their investment. This model is true of most enterprise technology acquisitions that you can mention.

    On the challenges of data integration, big data technologies are specifically designed to address the issue of integration of data across storage silos. By accessing disparate internal data stores, combining them with a broad ranges of external and remote sources and consolidating them into a coherent, centralized repository, big data solutions form vast pools of information that typically could not be analyzed using traditional tools. Data is one of the few resources that has increasing returns to scale, becoming more valuable per bit as its volume grows.

    Concerning the causes of data scientist shortages, yes, Ron is correct that there are already people working in FSOs to understand trends from statistical analysis and data flows, but they tend to be working on datasets within structured data silos without the ability to analyse larger volumes of unstructured data of different types.

    Big data is a technology, not a job role. Arguing that an organization already has access to the necessary skills to benefit from the technology that this is a negative point for the promise of big data seems illogical. Surely, if a technology can enable existing staff to be more effective, that must be a good thing. It enables these employees to gain a broader understanding of the environment by consolidating disparate data stores becoming a target for analysis.

    Additionally, as the second wave of big data products come online, the objective is to reduce the requirements for data scientists and bring these analytics frameworks to the level of operational managers, rather than relying on the specific competencies of statisticians, by enhancing the simplicity of the visualization tools and improving usability.

    Having said all that, I agree on Ron’s final point. Big data is a management fad. At Tarmin, where we develop the GridBank Data Management Platform for big data, rather than getting swept up in the hype, we counsel our clients that, as with all technology solution acquisitions, they should ensure that they have clear objectives when considering the solution and that they know how, whether through internal competencies or professional services, they will implement big data to add value to their business.

  16. Thanks for taking the time to weigh in with your comments Joseph — very much appreciated.

Speak Your Mind

*

Next article recommended for you
what_is_big_data_icon