The things we do with business analytics technology right now do not represent the ultimate purpose for these technologies. It is all very immature, and so it may be useful to some to understand where this current fascination with business analytics is headed. But first of all let me give some context for the term ‘business analytics’. Most businesses have spent the last half a century streamlining and automating their transactional activity and business processes – it’s taken a long time, and the process is ongoing. Business analytics is concerned with a different issue – streamlining and automating the decision processes within businesses. As such the relevant technologies include all forms of analytics including business intelligence, visual analytics, predictive analytics, machine learning, decision support technologies, and anything else that may emerge to support decision making processes. These technologies should not only support greater decision efficiency, but are also used to improve the quality of decision making. Decision automation is already a reality in many businesses and includes recommendations to customers, fraud detection, predictive sales to identify the best prospects, and of course the provision of information to employees so they can also make better decisions.
The actual technology used will change quite rapidly; this is an immature market. And so we see some organizations already considering Spark as an alternative to Hadoop – and widespread use of Hadoop is less than a decade old. As such this is not a technology discussion – technology fads and fashions are notoriously fickle things and sit below any strategic considerations.
A maturity model needs to start at the beginning, and for most of us this is the well established tradition of regular reporting. In reality this is not really business analytics, but simply statements of fact. Paginated reports are in most respects simply tables of numbers stating facts that relate to historical performance. The recent explosion of visual analytics technologies, where these same facts are displayed in more easily digested formats, takes us nearer to the idea of analytics since pictures do convey meaning that is hidden in rows of numbers. In the diagram below we start with data discovery and exploration. The two enclosing rectangles divide business analytics into R&D and Production. R&D might not be the term some would choose, but the activities contained in this box are not production oriented, but take place prior to any applications that might assist with or automate decisioning processes.
The notion of production analytics is a simple one. Any process that is regular, predictable and repeatable is a prime candidate for automation – either partially or completely. Businesses exist because they do automate much of what they do, and hence drive down transaction costs. Decisions are not different from transactions or other processes. If they can be automated the business will save money, and may even generate additional income – as we shall see shortly.
Data Discovery and Exploration
It goes without saying that the starting point of any analytical activity is that of gaining an understanding of the data available for analysis. This might be data from within the business or external sources – it really doesn’t matter. What does matter is that someone is tasked with the job of understanding which features in the data are important and how the data might be transformed to create powerful new features. There are also issues around data quality and preparation that will need to be addressed. But one thing is for sure – this is a background activity that is necessary for any further activities to be successful, and does not usually deliver anything that immediately modifies the decisioning processes within the business.
Self Service Analytics
It hardly needs to be stated that this is currently the hot topic in business analytics. The notion that anyone can poke around in their data with easy-to-use visual tools is obviously appealing, but in some important ways it is also flawed. There is without doubt a need for some people to slice and dice data on a regular basis, although the notion that broad swathes of people might spend a large proportion of their time doing this is just absurd – people have other things to do. And so self-service analytics also comes under the R&D category. No one should be manually repeating the same analytical process week after week. If it can be automated then it is no longer the province of self-service analytics, but should sit in the production analytics domain.
Predictive Model Creation
Predictive analytics employ technologies capable of identifying patterns in historical data, which in some cases can be used to predict future business outcomes. The classic example is loan approval. If you apply for a loan it is almost certain that the application will be processed using a predictive model of some sort. As with self service analytics, data exploration and discovery is a necessary prerequisite. Unlike traditional business intelligence applications however, predictive models say something about the future, whereas business intelligence is purely concerned with a look in the rear-view mirror.
So far we have been concerned with what has been categorized as R&D activity – understanding our data and bespoke analysis. Production analytics is where we are headed since this is how the business will realize significant productivity gains and new revenue streams.
Broad Distribution Embedded Analytics
It is well known that context switching destroys productivity. If employees need to switch between the applications they use day after day, and some type of analytics environment, then we can wave goodbye to any notion of productivity.For this reason it is essential that analysis is embedded into production applications. So call center operatives might have a graphic showing the behavior of customers with a profile similar to the ones currently being dealt with. The opportunities are endless. Obviously we only want to embed analytics that have proven to have value in a high transaction environment – otherwise its ad-hoc self service analytics. Over the next few years we will see businesses moving from the flawed notion of widespread self-service analytics to embedded production analytics. As a final comment it should be noted that many suppliers will use the term ‘embedded self-service analytics’ – it’s just marketing.
Decisioning Applications Within the Business
While embedded analytics refers to the provision of useful visual artifacts within production applications, decisioning applications start us on the path toward decision automation. We’ve already mentioned the example of loan approval. Instead of a bank employee spending considerable amounts of time processing an application, loan approval analytics make the decision almost immediately and, the decisions are typically much more accurate. There is a step-change in sophistication here however, and it is problematical. Building predictive models is a highly skilled activity and many medium sized businesses will just not have the necessary skills. And so we see the emergence of solution providers. If you need to deploy decisioning applications to reduce customer churn then a large number of solutions are available. Bespoke applications may require the help of an external agent.
Decisioning Applications as Revenue Generators
As we climb the business analytics value chain so the thinking of business management needs to become more creative. Some of the most forward thinking businesses are already using their considerable data assets to create revenue generating decisioning applications. A simple example might be a large distribution network members of which, would pay to see information that helps with their own efforts. A supplier to this network would be completely justified in providing analytics that could not be got any other way. Even more creative thinking might see the opportunity to pull in data from a wide variety of external sources to add even more value to the analysis.
Federated Decisioning Applications
One of the lessons of ‘big data’ is that data volumes are usually nowhere near as important as data diversity. A federation of businesses operating in related markets can produce analysis that is orders of magnitude more powerful and meaningful than any single business. Consider the travel and hospitality industries. Airlines, hotel chains, car hire firms and other related businesses can create a much more compelling analysis of customer behavior than any single supplier. Again, this is already happening and in some cases revenue generating applications are being built.
The higher up this we move in the business analytics maturity model so the more numerous will be innovative new businesses exploiting the need for information. It’s all a long way from the notion that business analytics is an ad-hoc activity dedicated to the description of historical performance or diagnosing anomalies. Of course these things are still important, but this isn’t where analytics will determine the fortune or otherwise of many businesses.
Suppliers on the Maturity Model
Below is not an exhaustive list of suppliers, but simply gives notable examples of suppliers for each level in the business analytics maturity model:
Datameer – data exploration and discovery for big data.
Tableau – a popular visual analytics platform.
SAS – advanced analytics platform and predictive analytics toolset.
Logi Analytics – mature, well respected embedded analytics platform.
FICO – decisioning applications and model building platform.
GoodData – very advanced approach to analytics as a revenue generating opportunity, including federated analytics.