My goal today is to inspire you to make a strong business case for applying big data in your enterprise, a key part of which is taking big data beyond analytics.
Intro
My goal today is to inspire you to make a strong business case for applying big data in your enterprise, a key part of which is taking big data beyond analytics
Hadoop has been working its way through the hype cycle,
Early barriers to adoption include availability and cost of Hadoop engineers, technology stability, enterprise support (security, resilience, etc), ecosystem, lack of tooling
Many of these challenges have been addressed (largely thanks to the likes of Cloudera pushing enterprise grade features)
- now on the slope of enlightenment
- starting to deliver on its promises
- leading organizations are fully committing to Hadoop
However, many organizations still struggle to take Hadoop out of that pet science project status, moving from ‘interesting’ to ‘valuable’
But part of the challenge is creating a viable business case, one that clearly articulates a return on the range of investments that are required to adopt a new technology within the organization
This is underpinned by research from the likes of Gartner, one of the real sticking points is linking capability, the art of the possible, back to real business need and advantage within actual business processes.
Big data investments in 2013 continue to rise from 2012, with 64% of organizations investing or planning to invest in big data technology
This presentation will show you a few examples of how some of the customers are working with Aptitude Software are adopting Hadoop for operational business applications
to become truly data-driven enterprises, where there have been clear business cases to support the introduction of Hadoop
An investment bank
A business requirement not normally associated with big data, to understand cost to serve and true customer profitability
When you look at business structure, the broad range of direct and indirect costs, number of customer, etc. you can quickly see how the data challenge builds up
So this brings together data integration, process management, business user control and high volume data processing
Significant investment in data warehouse appliance – many, many racks, many, many millions $
Infrastructure heavily utilised (…overloaded).
But they were challenged with balancing system resources – between end user reporting and analytics, and processing new data coming in.
They wanted to do more …. What-if models, scenario modelling, etc.
They had to weigh up their options - continue to invest heavily in the appliance platform, or think differently
So they looked to Hadoop as a platform for offloading a range of batch processing jobs
The results still loaded into their DW appliance
So these requirements were no longer just IT driven. Business need and demand created a real viable business case for Hadoop
This slide from Cloudera maps the evolution as they see it. Far right being the panacea of big data, where we are bringing together data, analytics and business applications
Challenge on the vision to greater value from enterprise- and industry-specific objectives
Ensure there’s a tested and relevant incremental growth path to advanced analytics
Sustain the journey to scale with an enterprise data hub by letting the savings from operational efficiency projects pay for the downstream information advantage projects
Enable at every step of adoption and deployment with training and professional services aligned to objectives, building a center of excellence that spans the organization
Here is another great example of Hadoop being used for operational business applications:
Spotify who have been very open and public about their use of Hadoop to capture, process and calculate all of the royalties owed. From transactionally recording each track that has been streamed in Hadoop, to marrying up streaming data with contractual data and calculating the royalties owed to record labels, music societies, etc. and reporting accordingly.
As you can see this is clearly not analytics (well there’s a lot of that as well to improve customer experience and service features), but very transactional in nature, resilient – beyond five 9s, accountable - financial statement of record
They tell a great story of their journey, which was not without challenge, but clearly proves that Hadoop is ready for the most demanding operational business applications.
All 690 nodes and 28 PT of it !!
A great story, but one from a (not quite start-up) but very young company. No legacy, no integration, therefore no complexity, no mess, etc. - greenfield !!
The case is not always that easy to make
But how can you make the same business case within your enterprise when you have a complex landscape, and different attitudes towards new technologies, etc….
How do we take this approach and make it workable in a typical enterprise, one with multiple ERP systems, billing systems, CRM systems, a data warehouse or two
You don’t have the benefits of a greenfield, Hadoop is not going to be self contained and in splendid isolation.
The majority of business centric, big data applications are going to require integration to one or more of your existing systems, access to various operational data points that can’t afford to have the latency of load into Hadoop.
They’ll need to plug into existing or modified business processes and interface directly with business users.
But of course they have to take advantage of big data platforms and seamlessly integrate with the likes of Hadoop as they would any other technology.
So let’s take a look at a final example where all of these requirements come together:
Data integration
Real-time access – both to Hadoop and operational systems, such as ERP
Data processing
Analytics
Business user control and management
So let’s imagine a B2C business, one that has the opportunity to dynamically adjust product pricing based on a range of factors. We’re all aware of this with airlines where flight prices are based on day, time, demand and proximity to departure. It’s also done by stealth in retail through store loyalty cards – not dynamic, but with vouchers you are getting an individual price for a product based on a range of factors. But to drive the competitive edge, companies are looking to become more dynamic with their approach, responding in near-real time to a range of factors and conditions. If you’ve noticed, some supermarkets are busy installing digital price tags on their shelves, gearing up for it already.
One company we have been working with sells tickets to concerts and events around the world (essentially a ticket agent). They wanted to dynamically adjust the price of tickets based on a range of factors, including trends, demand, remaining available seats, etc. But done in a way that had an eye on cost of sale and profitability for a given set of tickets.
There is a requirement for
Data integration, ERP, CRM for order history, product info,
Access to broad data sets, ideally already in Hadoop - e.g. customer and product profitability
Pre-processing, e.g. social media to derive trends and demand. E.g. Concert ticket prices fluctuating based on Twitter trends
UI interaction – to define pricing models and to analyse pricing
Real time access, e.g. through services, to request specific prices
Real time access to latest data, e.g. latest customer master record
Down stream integration, e.g. to CRM or ERP
The breadth and volume of data that can be stored enables pricing engines to go way beyond the traditional approaches and start to leverage a wealth of information to price more competitively and profitably
Fortunately Hadoop and its ecosystem is at the point of maturity where this is now a reality
Aptitude Software is proud to be a partner of Cloudera, selected because of their commitment to making Hadoop suitable for the enterprise.
We differentiate by taking big data beyond analytics and embedding across the enterprise, integrated into business applications and in the hands of the business users that will derive the greatest benefit
Our goal is to provide a platform that enables business and IT teams to deploy data centric applications quickly and collaboratively
Where business rules, business process, data and application integration can all be defined graphically without the need to understand the underlying technologies involved.
And finally one that can take advantage of whatever processing horsepower you have at your disposal, in-memory, in-database and of course in-Hadoop.