Rethinking Business Applications in the Analytics Landscape
Why Information Modeling is Critical to the Success of Analytics
Surveys conducted by Gartner and other research firms consistently place analytics, the science of analyzing business data to derive useful insights and make better decisions, as one of the top three priorities among business leaders and senior executives across all industries. A recently published study from Gartner predicts that business intelligence (BI) and analytics will remain a top focus of CIOs through 2017.
The explosion of information is causing leaders to scratch their heads as they strive to understand the impact all this data will have on their businesses -- and how to put their hands around such vast amount of data. The data is not only coming from traditional transactional systems but also from social media, mobile channels, search engines and other sources in structured and unstructured formats.
A digital business needs insight into data to drive better decisions. Any edge you can get over your competitors can put you in the driver’s seat. Analytics has the potential to do just that if it is done correctly.
Organizations implementing analytics as part of their business strategy should strive to understand the important developments happening in this exciting field. Analytics is slowly moving away from traditional business intelligence and data-warehouse frameworks. Traditional BI was only considered a real need after an application had been in production for years. Traditional BI systems were usually built separately from the transactional application and because of the underlying architecture it was difficult to do real-time data analysis.
Recent advances in the field of analytics (such as MapReduce, Hadoop, in-memory computing, Massively Parallel Processing (MPP),and NoSQL Databases including HBase and MongoDB) have enabled companies to take a radically different approach to data processing, analytics and applications. Ease of integration using web services and APIs have allowed vendors to embed analytics into business applications. With embedded analytics and advanced data storage and processing technologies, users can now look at real-time data. Organizations don’t require a vast amount of historical data before they can start analyzing it. With analytics accessible from within their business applications it is becoming much easier to slice and dice business data without relying on IT.
Analytics applications have also blurred the boundary between business analytics and the business process applications. It is now much more imperative that the road map for business analytics is complementary to the future plans for the business process applications.
I fundamentally agree with the assessment of the analyst community that analytics implementations should happen alongside the design and development of the business application for the best possible results. Having said that I am not implying that you cannot embed analytics into live applications or you won’t be able to get the desired results by doing that. You can, but you should be prepared to enhance or add to existing data models, or in some cases establish additional ones. You should also be able to clearly communicate why expanding the production data models is essential to build the right analytics solution even though that might come with additional cost. For example, your current design may not be capturing certain data points that are needed as part of the analytics solution. In that case you will need to consider adding those data elements to the model and most likely back-fill the data. That may also require altering some of the processes within your business application.
The most important aspect when creating an analytics solution is to make sure that it aligns well with your business goals and desired outcomes.
At MicroPact, we take a Data-First ™ approach to dynamic case management and business process management (BPM) application development. The data or information model is at the core of our application build process which in turn provides a well-grounded starting point for constructing flexible and agile analytics solutions.
Why does a Data-First approach align so well with modern BI/analytics requirements?
- The value provided by any analytics solution depends on the type of data captured. Data captured directly relates to the thoroughness of data model, which should have all possible data elements, corresponding data types and any dependencies identified early in the design process.
- A Data-First approach lets designers and managers see early-on if the data captured (as defined by the data model) is not only of the right type to meet the necessary business outcomes but also sufficient to perform data analysis.
- A Data-First approach looks at data in a holistic way – it helps you understand which data is most critical and what interdependencies are necessary for an effective analytics solution.
- System design based on the Data-First approach is also inherently flexible as you can always add more data objects to a system with relative ease.
- A Data-First approach allows end users to get engaged early on. Users can easily point out the data elements that they would most like to capture and analyze. This approach also results in a higher adoption rate of analytics among a given user base.
When designing new applications, it is prudent that your data/information modeling phase take your analytics requirements into account. Below are few simple use cases to illustrate why a good data model is essential to building an effective analytics solution.
- To analyze the fluctuation in the cost of processing a case over time, both cost and time should be captured appropriately in the data model.
- If there are fields where null or empty data can create issues with pattern analysis, such fields should not be allowed to accept null values. For example, if an event date is a must for each occurring event/activity and is required from analytics perspective, then the system must not allow users to save the event information without an event date.
- Another scenario will be where users want to review a snapshot of activities that occurred during a particular time period. For very specific activities, audit logs may not necessarily record in the format needed for analysis. For example, how often investigators get replaced over the duration of a particular case. To analyze such data in addition to storing information on the current investigator, historical data relative to each previous investigator (such as the date they started and when they were replaced) must also be captured.
- In some situations data from a separate system may be needed for certain type of data analysis. Provisions for such data import should be made as part of the data model design.
Taking the time up-front to model the correct and complete data structure goes a long way towards building an effective analytics solution – which in-turn can empower workers’ to achieve their (and your) business goals. For a deeper dive into the Data-First approach, download a complimentary copy of “How Data-First Development Overcomes Barriers to Creating Effective Applications”, a study by CITO Research. Also, for a more in-depth look at the future of BI/Analytics, you might want to review CTIO’s report Search-Based, Natural-Language Analytics: Because That’s How End-Users Think. And of course, if you are looking for a platform that can do all of the above, be sure to investigate entellitrak.
- Answering Your Questions about Partner Sales Enablement February 5, 2020
- Introducing Chris Flores, Vice President of Global Alliances January 15, 2020
- 6 Tips for Stress-Free Software Adoption December 11, 2019