Embedding and Operationalizing Analytics for Action: Three Important Takeaways
Close to 50 percent of TDWI survey respondents said they embedded models or algorithms into business processes in some way. What can we learn from these respondents?
- By Fern Halper
- February 4, 2016
I recently completed TDWI’s latest Best Practices Report, Operationalizing and Embedding Analytics for Business Success. Operationalizing and embedding analytics is about integrating actionable insights into systems and business processes used to make decisions. These systems might be automated or provide manual, actionable insights. Analytics is currently being embedded into dashboards, applications, devices, systems, and databases. Examples run from simple to complex and organizations are at different stages of operational deployment. Newer examples of operational analytics include support for logistics, customer call centers, fraud detection, and recommendation engines to name just a few.
I want to focus here on the notion of embedding models and algorithms into business systems and processes because this is becoming increasingly important in organizations. For instance, a model might operate behind the scenes of a call center suggesting next actions to the call center agents, or a model might operate as part of a website, making recommendations. Some organizations are using models and algorithms for predictive maintenance to determine when a part needs repair or when a piece of equipment needs to be shut down. Others are using many embedded models that might make small decisions all across the organization.
Close to 50 percent of survey respondents said they embedded models or algorithms into business processes in some way. What can we learn from these respondents?
Organizations need to plan to deploy models. Only one in five respondents stated that it took days or weeks to deploy a model into operational use. About 23 percent said it took one to two months. Twenty-eight percent said that it took three to five months! Part of the problem is that IT often has to recode analytics for production. This can be the case even if the vendor provides an output language such as PMML. Sometimes organizations require that the models be recoded (it’s a company policy because of the systems already in place). Other times, the organization did not plan for the actual deployment of a model. Successful companies do plan ahead and have staff in place to actually deploy a model. This can be helpful even if the models need to be recoded.
Companies need to manage their models. It is imperative to be able to keep track of models to make sure they don’t get stale or become inaccurate. Yet, only 30 percent of respondents are utilizing some sort of model registry to share models with IT/development teams. The majority use shared folders. Others email models to developers. There are clearly issues with these kinds of solutions. Model management and documentation are needed. Modern decision management software allows organizations to register, deploy, monitor, and reuse models that might be instantiated into a business process. When you only have a few models you want to operationalize, it might be acceptable to store them in a directory and ask IT or your development team to recode them. However, this is not scalable or practical in the long term.
Make sure you have organizational buy-in. In the survey, 40 percent of the respondents stated that lack of trust in the data or results was a top challenge for operationalizing and embedding analytics. If the people involved in making embedded analytics part of their business process haven’t bought into it, it is likely that the implementation won’t succeed. Respondents cited concerns about displacing people and power shifts that staff fear might occur if the organization becomes more analytics-to-action driven. Organizational buy-in via socialization of the concepts is crucial to success. Effective organizations seek executive sponsorship, develop a proof of concept to illustrate the value of the technology, and make a point of getting everyone on board. They communicate by highlighting accomplishments and pushing the innovation message, and continue to evangelize. It can take work, but it is worth it.
These are only a few of the findings in this report. You can download a complimentary copy of the Best Practices Report here.
Fern Halper, Ph.D., is well known in the analytics community, having published hundreds of articles, research reports, speeches, webinars, and more on data mining and information technology over the past 20 years. Halper is also co-author of several “Dummies” books on cloud computing, hybrid cloud, and big data. She is the director of TDWI Research for advanced analytics, focusing on predictive analytics, social media analysis, text analytics, cloud computing, and “big data” analytics approaches. She has been a partner at industry analyst firm Hurwitz & Associates and a lead analyst for Bell Labs. Her Ph.D. is from Texas A&M University. You can reach her at firstname.lastname@example.org, on Twitter @fhalper, and on LinkedIn at linkedin.com/in/fbhalper.