Representative End User Clients
Representative Automation Clients
Representative Software Clients
That, as I believe Shakespeare almost said once, is the question. I've written before that, at the highest level, ARC thinks about analytics in four categories: Describe, Discover, Predict and Prescribe. That's fine, it gives us a nice cozy way to bucket business intelligence, operational intelligence, and analytics tools and solutions. But, that's as far as it goes. As a model, it doesn't help technology buyers figure out what is the appropriate type of solution for any particular situation. To work through that process calls for something different.
Fundamentally, it means thinking about the decisions that people need to make, and how they get made. Meshing that thinking with a bit of thinking about data and technology can help. There's a lot to consider, but thinking about five core questions will get you a long way in my experience:
It's perilously easy at the moment to get swept away on a tide of big data mumbo-jumbo. The acronyms are coming thick and fast - it seems every day introduces a shiny new Hadoop-related project or start-up. Eventually yes, discussing the technology is important, and choices like on-premise vs SaaS vs cloud also kick in.
But fundamentally, successful analytics are very simple: Get the right information, to the right people, at the right time. If you can't get that right, nothing else matters. The 5 questions above are a handy way to think about that before you even think about technology. Because solution choices should be driven by the management and operational decisions people need to make, not by technology.
Fundamentally, it means thinking about the decisions that people need to make, and how they get made. Meshing that thinking with a bit of thinking about data and technology can help. There's a lot to consider, but thinking about five core questions will get you a long way in my experience:
- How soon after events occur does the data need to be used to make decisions? Often referred to as the decision window, data could be analyzed as immediate as real-time, or decisions might be based on data that is years old. There's a spectrum here, and many decisions will be made on a decision window that lies somewhere between these two extremes. This question will help you understand the urgency and nature of the flow from raw data through to actionable information
- What's the scope of the decision? Or, if you prefer, how "big" is the decision. Does it just have immediate, local impact, or is it more wide ranging than that - perhaps defining strategy for years to come. The scope and impact of many decisions will lie between these two extremes. This question will help you think about who is needed to make a decision, how often decisions are likely to be made, and how precise and considered those decisions will be.
- What type of processing do you need the software to (easily) perform on the data? The word "easily" is key here - you can do practically anything with any software if you have enough time, budget and patience, but that's not the point...Does the person using the tool to make decisions require just basic summation and aggregation of data, or something more sophisticated - like linear regression or statistical analysis? This question focuses on the complexity of the data, and the complexity of the decisions that need to be made.
- What data is needed to inform the decision? This could be answered at a high level, or in much more detail, depending on the selection in question. At a high level, we might include SCADA data, data about physical or chemical properties, or data about business transactions. At a lower level, if we were interested in analyzing pump performance for example, the data might include many pressure readings, temperature measurements - and so on. Important to note, at this point we're not interested in where the data is - whether it's in a relational database, a spreadsheet, or something else....That affects the implementation, but it doesn't really affect the decision that needs to get made.
- How does action get taken? Is action automatically triggered by the data, or is it presented to a user (or operator) as information in some way? That presentation might be static, with no possibility for interaction. Or, it might require a more dynamic presentation of information that allows the user to explore and understand cause and effect. Understanding how someone will interface with data and information colors many of the other aspects of the solution.
It's perilously easy at the moment to get swept away on a tide of big data mumbo-jumbo. The acronyms are coming thick and fast - it seems every day introduces a shiny new Hadoop-related project or start-up. Eventually yes, discussing the technology is important, and choices like on-premise vs SaaS vs cloud also kick in.
But fundamentally, successful analytics are very simple: Get the right information, to the right people, at the right time. If you can't get that right, nothing else matters. The 5 questions above are a handy way to think about that before you even think about technology. Because solution choices should be driven by the management and operational decisions people need to make, not by technology.