May 201619 applications that are fraudulent. We can do analysis on grant application data to identify possible frauds and take action on them. But if we stop there, the work done will not be able to detect fraud in future grant applications. We need to build the analytics into the grant application processing system, possibly by visually marking out suspicious applications to grant application-processing officers, so that they can investigate those cases further. Taking it one step further, we can even consider automatically approving or rejecting some grant applications, and only surface the more complex cases to processing officers for decision.As such, to operationalize analytics, we need to ask the question "How can this analytics effort be operationalised?" for every analytics effort. To answer this question, we need to know which process the analytics effort will change, and how the changed process might look like. Moreover, we should examine how often the process gets executed. If it is a yearly process, e.g. work plan review, then the potential for impact might be less. But if it is a daily process, it is likely to be of greater impact. For example, we wanted to perform analytics on the daily reports filed by our officers, to draw out key trends and new developments. However, we were not clear how to operationalise. If we were to perform the analytics, when will the management look at the findings, and how will they look at it? Do we need to table the findings regularly to management meetings? Eventually, we decided to stick to an existing monthly reporting process. This is still work in progress, but I believe that this will change the way the organisation reads those reports.What are some ways to operationalise analytics? One key way is through en-hancements to IT applica-tions, as IT applications execute organisations' pro-cesses. Going back to the grant processing example, the grant management application was targeted for enhancement. However, there are some who felt this approach is not feasible as their application vendors do not have analytics know-how, and thus are unable to make such enhancements. We overcome that through a simple design: loosely link up the IT application with the analytics by using data as the interface. Everyday, the IT application generates out data need by the analytics model, and this data is transferred over to a location accessible by the analytics model. The analytics model then reads in that data, runs it through the model and finally outputs its recommenda-tions. Subsequently, the output is transferred back to the IT application, which ingests that data and displays it ac-cordingly (e.g. putting a red flag to grant applications which are likely to be fraudulent).Change management is a key consideration in operationalising analytics, as processes are changed and people naturally resist change. In the grant-processing example, processing officers questioned the basis for recommendations made by analytics, as to them it was a black box they did not understand. Change management in itself is a big topic and I should not attempt to go deeper into it here.I believe the key to moving analytics beyond POCs and PowerPoints is to operationalise analytics, as operationlisation changes organisations' processes, and processes is the conduit for analytics' impact. Probably the only thing that is more powerful than changing processes is changing culture, since culture decides what organisations do and also how they do them. A good followup question to chew on would be "How can analytics change culture?". Change management is a key consideration in operationalising analytics, as processes are changing and people naturally resist changeMing Fai
< Page 9 | Page 11 >