Even though the field of predictive analytics still has maturing to do, it is on the verge of going mainstream, which means businesses should start thinking about ways they can benefit from data mining techniques, said Fern Halper, research director at TDWI.
Speaking as part of a TDWI webinar titled "Predictive Analytics for Accelerating Business Advantage," Halper said the pace at which organizations are adopting predictive analytics technology is increasing. At the same time, the results of a recently completed TDWI
"Respondents did not cite forward-looking things like competitive differentiator, which is actually how a lot of successful companies look at analytics," she said. "So that suggests that while predictive analytics is going mainstream, it's still very early."
Read more on developing data mining techniques
See why so many people are excited about uplift modeling
Read how hospitals are using predictive modeling to solve readmissions
Learn what it takes to get your data mining project right
There are some ways businesses can begin moving their analytics initiatives forward. Halper said that data mining techniques need to be run by the right people using the most appropriate technology. This may mean different things to different organizations, but there are some general guidelines to follow.
- Don't expect results overnight. Halper said a lot of the work involved in making analytics projects successful is driven by relationships. Business users need to trust IT developers to deliver the services they need. This trust doesn't develop immediately.
- Work in steps. Businesses that try juggling too many pieces right off the bat will inevitably let some things fall. Halper said organizations should start implementing predictive analytics tools and processes slowly. This helps ensure the technology is on solid footing and workers have time to adjust to any changes.
- Think about the skills you need in-house. Vendors have been busy producing predictive systems that practically run themselves. Halper said this technology may be a good starting point for businesses that have limited analytics need, but they should understand you can only go so far with simpler options. Organizations should think about whether they will need a statistician or data scientist in-house.
- Think about organization. There is no perfect model for organizing analytics projects. Some companies form small teams to lead projects, while others develop centers of excellence that bring in team members from different departments. Some are even hiring chief analytics officers to lead initiatives from the top down. Whatever the best fit for your business is, just make sure the organizational structure is clear at the outset, Halper said.
- Manage your models. Halper said many businesses that are first getting going with predictive analytics find it relatively easy to develop their first data model. However, once the organization becomes more mature and starts using many models, it's easy for them to get tangled. She recommended businesses work with their vendor on developing a system for managing various models.
- Think about different kinds of data. Most businesses that are engaged in predictive analytics base their models primarily on structured and demographic data. However, there is a growing pool of unstructured data, such as social media posts and other types of online content, that can improve models. Analytics users should try to see if they can get their arms around different data types.
- Consider new technologies. Emergent technology can be exactly what businesses need to leverage new data types, Halper said. New analytics platforms, such as Hadoop, are making it easier than ever to analyze unstructured data.
- Get your business intelligence foundation laid out first. Halper said businesses need to be able to walk before they can fly. A solid BI infrastructure often supports organizations as they try to develop more intensive data mining techniques. Many of the lessons learned from standing up a BI infrastructure will be relevant to developing predictive analytics capabilities. Some of the technology, such as a standard data warehouse, will also help.
This was first published in January 2014