In this blog series, James Brenza has been talking about implementing big data and analytics programs using a composite case study to illustrate the process. Each week James focuses on one of the seven steps giving specific examples to help illustrate how the tools can be used in a very practical manner. This is the sixth in the series that corresponds with the seven stage implementation model. More information on that robust model is available in the Innovative Leaders Workbook for Implementing Analytics Programs by Maureen Metcalf and James Brenza (scheduled for release in September 2014).
Follow a methodology: The execution phase of an analytics initiative has some similarities to other efforts, but is also very unique. The similarity is the extensive use of systems and processes; the use of databases and servers is quite common. However, some of the types of databases or servers will be unique to analytic efforts. Due to the volume of data or use of statistical analysis, new infrastructure may be required. It is critical that the infrastructure is available and validated prior to attempting any model development. It is possible and reasonable to deploy the infrastructure while the teams assess data and build models, but concurrent development of all levels of the solution multiplies risk. Establishing core infrastructure design and the data for initial analysis contains the risk as small models are built incrementally. Additionally, the data discovery methods may be unique to many team members. The team can significantly reduce barriers by embracing an industry standard (e.g., CRISP-DM), leveraging published documents on its use and demonstrating how they align with it.
For the preparation of data for analysis, the leader should be very clear on the progress of establishing, exploring, understanding, cleansing and integrating the data. This is especially critical if new data sources are being used or if it’s the first time a well-established data source has been used for this type of initiative. Even though data may be used for business transactions, the team can’t assume the data is sufficiently standardized for analysis. The leader must ensure this clarity through collaboration with business analysts and data scientists to develop a graphical depiction of each data source going through validation and normalization processes.
Let the data guide you: Since the team should be following an iterative execution methodology, they must be prepared to demonstrate flexibility and adaptability as the results of the data analysis guide their tasks. As hypotheses are constructed, the team should embrace the possibility of proving themselves wrong. If the team focuses on data that will only prove and support their hypotheses, they are creating a potential failure. Bad models don’t age well—so the team has to attempt to break their own assumptions. They’ll also need to hold back portions of the data for model validation and training. This standard process will allow them to increase confidence in the models.
Another example of this could be when two (or more) data sources that are being mined have historically supported different operational purposes. At first glance, data elements may appear to align across the systems. As the analysis evolves, data issues arise in which apparent logical matches are broken. With deeper analysis, the team may discover that similarly-named attributes serve unique purposes. In isolation and context, each source is 100 percent accurate. By being open to opportunities, the team may simplify the challenges being solved, or identify additional solutions based on a single model.
Embrace continuous change: At present, the pace of technology and technique innovation is faster in analytics than in nearly any other technology field. The industry is continuously adding new capabilities that will greatly simplify solutions. Additionally, many business intelligence and visual tool vendors have enabled community development for new features faster than any single vendor can invent them. While building an initial infrastructure is required, it is critical to consider that it must have the capability to evolve quickly. Embracing this continuous state of evolution is key to maintaining team productivity as well as user satisfaction.
Keep communications flowing: Throughout the execution, the leader should constantly refine the plan and communications to transparently discuss the progress of the iterations and the stride toward the final outcome. This can be accomplished by documenting planned versus actual progress for each sprint. By ensuring each sprint completes some demonstrable progress, the leader will be equipped to maintain sponsor confidence. As each progress demonstration is completed, the leader should review the stakeholder management plan to ensure that sponsors and stakeholders are maintaining or increasing their engagement. If support is waning, the stakeholder management plan must be revisited to reactivate support.
As the models evolve, the team can assess the system and process changes necessary to implement them. At the same time, they should review the measures that will ensure use of the models. An ideal mechanism to support that is to simultaneously define and communicate the rewards associated with the model utilization. Communicating the rewards will help reinforce the new operational methods and behaviors as well as support the retirement of the outdated processes.
Focus on team health: Due to the iterative nature of the process and continuous data discovery, the team will go through many cycles of elation and challenges. The leader should focus on team resiliency to ensure team members remain committed and cohesive. It’s important to recognize that challenges in the data or models are not a reflection of the team’s ability, rather they are an artifact of reality. By focusing on celebrating successes and alleviating team stress, the leader can help the team maintain its momentum. To help reinforce that consistency, the leader should regularly engage the sponsors to provide additional support to preserve and cultivate team morale.
How is leading a big data/analytics initiative different than other projects? So let’s take a moment to sum up what’s unique about data and analytic initiatives.
- The constant discovery of data and model strength requires vigilant and transparent updates to the sponsors. They should receive regular status reports that reinforce the notion that the initiative is not a traditional system implementation, but a process of discovery.
- The solution infrastructure and tool use may evolve throughout solution development.
- Referring to the initiative as a journey is an appropriate phrase.
- The leader should be prepared to provide frequent updates on data discovery and model evolution which may include frequent bad news (which if it disproves outdated assumptions is actually good news).
- The constant flexibility to link model effectiveness to a business outcome is very unique to these initiatives.
As the initiative progresses, the team must be prepared to incrementally implement improvements as they become available. In the next section, we’ll discuss techniques to embed the transformation.
If you are interested in reading more by James, you may want to read: Evaluating Big Data Projects – Success and Failure Using an Integral Lens, Integral Leadership Review August – November 2013. You can also listen to the NPR interview that accompanies this paper including a dialogue between James Brenza, Maureen Metcalf, and the host Doug Dangler.
We also invite you to join the LinkedIn group Innovative Leadership for Analytics Programs on LinkedIn curated by James.
To become a more innovative leader, please consider our online leader development program. For additional tools, we recommend taking leadership assessments, using the Innovative Leadership Fieldbook and Innovative Leaders Guide to Transforming Organizations, and adding coaching to our online innovative leadership program. We also offer several workshops to help you build these skills.
If you are interested in receiving James’ seven part blog series or our other articles by email, please sign up in the box on the right labeled Get Email Updates From Us.
Photo credit: www.flickr.com Marlus B.