Data Storytellers: Each month we'll be highlighting experts and advocates to share their data stories, knowledge, and insights into the future of data and analytics. Subscribe to the RSS feed to get the latest updates.

Analytics today faces many challenges and opportunities, making the journey toward insights murky. We at Cisco have spent twelve months seeking ways to address analyst productivity by focusing on the initial stages of data wrangling, data discovery and investigative analysis. This process occurs prior to random forest, algorithms, regression, modeling, and other analytics approaches.

Why, you may ask, are we focusing at the early stage of the analytics journey? Well just like you, our own experiences have proven analytics is only as good as the inputs, and we have inherited approaches that are not optimal.

So, let's go back to the future with data analytics by embracing where the journey starts and finding better ways of doing data discovery and investigative analysis.

1. Focus on discovery initially, not predictions

All stakeholders seek predictions to make better decisions, especially on complex problems we have yet to tackle. Cisco's emerging capability uses an unsupervised machine learning approach to rapidly discover columnar relationships ranked on a five star level of connectedness. No upfront coding, modeling, or querying is required. It provides an objective view of what variables are connected, which in turn, improves quality of data, quality of algorithms you will create, and eventually will improve predictions later in the analysis flow. 

2. Enhance data quality

The output of your analysis is only as good as the data you input. Address the cause, not the symptom of poor analysis. Oftentimes the cause is poor data quality, for example, misspellings, duplicates, blanks, inconsistencies, gaps, and more. Preparing data often consumes valuable analysts' time, and many tools cannot handle massive data sets. Cisco Data Preparation delivers enterprise-grade self-service data integration and preparation capabilities for non-technical business users.

3. Treat tools as inputs, not as outputs

No tool can promise to provide you the final answer and solve all of your challenges. Tools are meant to augment analysts and data scientists and serve as an input to your analysis. The process to getting to the output is iterative and evolving with your collaborative and collective business context.

4. Focus on unleashing the legions of data scientists and spreadsheet jockeys

The question I hear most from Fortune 500 companies, public sector agencies, and commercial businesses is "how do I find more data scientists to hire?" Though there is a lack of data scientists or analytics experts, you do have a deep bench of business analysts who understand the key business problems their executives need to solve. Rather than requiring business analysts to be dependent on data scientists, unleash them with tools for data discovery and investigation.

It's time to break from the accustomed approaches and realize going back the future, going back to the data, has the potential to dramatically enhance the time, output, and impact of predictive analytics.