It’s hard to believe, but now that 2011 is almost over it’s time to look ahead. The technology pundits are starting to publish their 2012 predictions, and it’s not surprising to see topics like analytics, cloud, big data, mobile, social networking, virtualization, open source on these lists. Instead of creating another list of predictions, I thought I’d comment on analytic trends from an IT perspective. So here is my top 10 list of analytics considerations for IT.
1. Cloud: Think about how cloud technology should be used as part of your analytics infrastructure – Many organizations have started to use salesforce.com or other cloud based operational applications. Some organizations have started to use infrastructure or platform options from Amazon and other cloud providers. Others are starting to use cloud based principles for how they design their internal systems and have plans for full-fledged private cloud deployments. We are seeing a tremendous level of interest in our cloud-based analytic offerings, but we are also seeing a considerable level of confusion. Although salesforce.com has done a great job in driving interest in the cloud, it has also created a level of expectation that is not applicable to all application types. Implementing a basic CRM solution that is devoid of customization is simple and can be accomplished in a short period of time. But once you customize salesforce to meet your needs, integrate it with your other on-premise applications, and start to think about your needs outside of sales such as marketing, customer support, executives, etc., the magic of the cloud starts to dissipate. This is not to say that the cloud is not compelling, it is, it’s just that you need to think about the needs of the users, the type of application, etc., to ensure that you achieve success. When it comes to analytics in the cloud, unless you are looking for basic reporting, don’t think salesforce – if someone is selling you that notion, an alarm should be going off! For real analytics, you have to consider the analytics lifecycle, it’s simply not the same as the typical application development lifecycle that consists of design, dev, test, deployment. The analytics lifecycle consists of an iterative analytics sandbox that requires flexibility and intense data cleansing and integration. Once the analyst has developed and tested the analytical models, the models need to be operationalized – the required data needs to be provided in a sustainable, trusted fashion, models need to be monitored and the analytics need to be embedded in operational systems to derive maximum value. This certainly doesn’t mean that analytics can’t be cloud enabled, we have a significant number of customers that are leveraging the SAS cloud, it’s just that IT and the business has to design the correct approach based on their requirements.
2. Enterprise Architecture Approach: Leverage an enterprise architecture approach to your analytics infrastructure – In many cases, the analytics environment is thought of as a separate data warehouse initiative that is not designed with the same degree of rigor that is applied to operational applications. IT’s role may be limited to managing the infrastructure and providing data support, an approach that typically leads to problems down the road. IT should strive to apply the same level of enterprise architecture discipline to the analytics infrastructure. IT needs to understand the potential role and relevance of all architecture options that are open to them: in-database processing, in-memory, grid processing, etc. IT should ensure that the proper architecture choices are made when it comes to designing the database storage approach for multiple analytics projects in order to limit data warehouse and datamart sprawl. Lastly, IT should also be involved in designing an architecture that supports the entire analytics lifecycle – data exploration / preparation, analytics sandbox support, and operationalizing the production environment.
3. Big Data: Start a pilot project leveraging big data possibly with unstructured text and Hadoop – We’ve said quite a bit about big data, so from a planning perspective I’ll just point you to several previous posts, and suggest that 2012 should be the year that you map out your plans for big data. Start small while planning for the future – consider working with your business counterparts to take on an unstructured data project, possibly leveraging Hadoop as your data store for unstructured data. Or consider leveraging Hadoop as part of your ETL process – use it as a vehicle for raw data, then leverage the distributed processing to extract relevant data into your data warehouse for additional analysis.
4. Information Management: Start taking a strategic approach to data with information management – As typical with other technology disciplines, the data management market is undergoing constant change. Organizations leverage a variety of technologies to manage their data; in addition to traditional ETL technologies, technologies including data quality, MDM and data governance are now leveraged to ensure that data is optimized for operational and analytical use cases. Organizations need to move beyond the traditional approach that tends to be reactive and silo-based to a managed, even predictive approach that values information as a strategic asset. There are many things to consider in terms of implementing a comprehensive information management approach, but here are several areas that SAS will be driving in 2012:
- Information Governance: A collaborative approach involving business and IT to information management starts with a complete solution for data and analytics governance, including data stewardship, business glossary and reference data management support and is backed by a best practice approach that includes the option to implement a formal Information Management Center of Excellence.
- Analytics Advantage: SAS is a strong proponent of leveraging analytics to drive any business process, so why would the process of transforming data into actionable information be any different? SAS leverages analytics in a variety of ways that results in a unique information management approach that delivers superior results. At the front end of the data acquisition process SAS’ unique stream-score-and-store design pattern combines semi-supervised analytics up front as data is being access to identify, classify, filter and tag data of interest. This is especially important in Big Data environments since it is not always feasible to store the data prior to analyzing or processing the data.
- Integrated Decisions: SAS provides the ability to embed rich information and analytic services directly within operational applications bringing rich analytics results directly to the point of decision. SAS provides a closed loop information continuum that cycles the resultant behavior that is driven by the information and analytical services back into the information and decision lifecycle. This improves an organizations ability to make decisions based on the latest information that leads to improved business results
5. Open Source: Avoid the hype, think about your overall requirements and experience when considering open source – Having worked at both open source and private source software vendors, the only thing I have to say about the politics of open source is that people that think that open source proponents are driven by altruism should spend time with open source vendors as they are driven by the same business reality as private source vendors. That being said, there is a place for both business models and IT has to make decisions based on a combination of product maturity, functionality, risk, reliability, service, trained resources, cost, etc. Take Hadoop for example, it’s currently better suited to organizations that are willing and have the resources to utilize a relatively immature technology. It has limited tooling, it’s overall technology stack lacks definition, talent is difficult to find, it’s better suited for batch processing, and the Hadoop marketplace is evolving, with rapid change of vendors, products and standards. This isn’t to say that you shouldn’t use Hadoop, SAS has an aggressive Hadoop roadmap and is engaged in many Hadoop projects, it’s just that IT really has to assess open source technologies for proper fit, support, etc.
6. Synergistic Analytics: Provide an analytics infrastructure that fosters synergy between analytic initiatives – In many cases, organizations leverage individual analytic disciplines such as forecasting, optimization, text analytics and data mining for a departmental specific business problem. As such, even though valuable, from an enterprise standpoint, the value is somewhat marginalized or not optimized. What we see is that departments, such as Marketing, Finance or Risk are using one technique to solve a problem but not multiple techniques. So, for instance, in a bank, the risk department will use econometric forecasting to manage the treasury portfolio and overall risk. But they are not proactively looking at text (emails and chat), to see where the bank could be exposed and tying risk associated with text to an overall corporate risk strategy. Or a retailer uses forecasting to predict what will sell, but they are not looking at on line sentiment and customer segmentation to optimize what product they should offer what customer and how to optimize distribution all as part of the same process. How does this impact IT? First of all, IT has to provide an analytics infrastructure that is capable of supporting multiple, enterprise class analytics initiatives and IT has to help provide integrated data that provides the basis for multiple, related analytics projects. And finally, IT must play an instrumental role in integrating the analytics results in operational systems.
7. Operational Analytics: Deliver the value of analytics directly to the point of decision – Organizations have to move away from treating analytics as a reactive, after the fact reporting and analysis mechanism. Analytics should be embedded directly within the operational applications – rich analytics should drive call center interactions, automatically serving up an appropriate script based on the customers propensity to respond or likelihood to churn. Analytics can be used in real-time to detect potential fraud, for example, for every single credit card transaction. IT needs to support an analytics infrastructure that provides the ability to web service enable any analytics or information capability – this provides the ability for operational applications, regardless of where they are running, to leverage / embed the analytics in the operational systems. This also provides the ability to re-use analytic or data preparation processes from different systems, applications or deployment approaches. This allows you to leverage the same data preparation job to load a data warehouse that is used to drive operational analytics as well as processing data that is streamed in real-time. Same with analytics, you could expose an analytical routine to a batch job or leverage it in real-time from a web-based or on-line application.
8. Approachable Analytics: Put processes and infrastructure in place to provide approachable analytics – Analytics are becoming much more pervasive. Reporting and BI have long been placed directly in the hands of end users and predictive analytics are now starting to be pushed out to end users vs. being limited to statisticians and analysts. This certainly doesn’t mean that the average end user has become a statistician, but it is driving the expectation that analytics are more approachable. End users are now expecting advanced forms of data and information visualization, they expect analytics embedded directly within operational systems, they expect that analytics or reporting will be available on any device at any time, and they expect that their analytics solutions will allow them to track additional context about their customers and prospects based on their ever expanding usage of social media. In 2012, IT should look to expand their analytics infrastructure and support to better enable end user support of analytics. Just as IT has moved to self-service enable basic reporting and business intelligence, IT should start thinking about the infrastructure necessary to enable approachable analytics.
9. IT/Business Collaboration: Don’t wait for the business, IT should drive collaboration in 2012 – It may seem like an old topic, but business and IT alignment keeps popping up – I attended one event this year only to be told that most organizations have achieved alignment, and now they need to focus on IT and business acceleration. On the other end of the spectrum, we have had IT and business focus groups where both constituents claim that the other side doesn’t get analytics. Regardless of where your organization is, it’s imperative that IT take a leadership role, IT needs to drive real collaboration in order to achieve satisfied business users and real business gain. It’s also important to note that with analytics, it’s not just business and IT, it’s business, IT and the data analyst or scientist. All of these constituents need to work together to deliver effective analytics. In 2012, consider implementing a center of excellence or competency center and focus on the role of data – data, or better yet, trusted information can be the linchpin that pulls these teams together.
10. Analytical Talent: Train your IT talent in the basics of data preparation for analytics – There have been a variety of recent studies that indicate that organizations are struggling with finding the analytics expertise they need. With the growth in demand for analytics, the talent gap is likely to get worse. How does this impact IT? Many IT organizations struggle with the basics of analytics. Since they don’t have a proper understanding of analytics, IT just manages the infrastructure and analytics remains a mysterious black box. This leads the users and the business analysts to do things on their own, leading to Shadow IT engagements. Although it isn’t necessary for your IT resources to become trained statisticians, IT must have a good understanding of the overall analytics lifecycle, and IT needs to understand the data requirements for analytics. With data, it’s not one size fits all, the format requirements for transactional data vs. OLAP vs. predictive analytics varies widely. In 2012, instead of devoting all of your training dollars and time to new IT tools and languages, why not take a small step and send some of your IT resources to analytics training that is focused on the proper understanding of data requirements. This will provide IT with the proper knowledge and context to support the demanding requirements of the statisticians and the end users, and will help break down the barriers between business and IT.
*NOTE: Originally published on SAS Voices.