In the first post of the Data as an Asset blog series, Quisitive looked at why data is often a challenge for even the most proficient companies, and how its data science team, led by Chief Data Scientist James Roberts and Data Science Strategist Shannon Ragland, developed strategies to start viewing and using data as an asset to solve business problems.

For any data initiative, a business undertakes, what is first needed is a solid foundation on which to move forward. This ensures that the end product provides a high-quality, predictive outcome that is both actionable and implementable. Quisitive’s Six Keys to a Successful Data Initiative provides this established base by ensuring that the proper team, technology and approach are in place, allowing clients to identify and tackle a well-defined business problem.

James and Shannon walk us through the six key steps below:

1.       Start with a Question

The most successful projects we find are the ones that start with a question, idea or issue, rather than simply starting with raw data and hoping to uncover something previously unknown in its contents. Among our clients who have a dedicated analytics team, we often see analysts simply mining data for the sake of mining data. Data mining became popular over the past decade, but overall, it has proven to be inefficient in comparison with a more targeted approach. For instance, business leaders often report that data mining does not produce enough relevant and useful analysis to inform their decision-making, or they become inundated with so many reports that the time is not available to determine the most useful information from among them. Overall, this is the result of business intelligence teams trying to prove they are doing something, rather than focusing effort on questions that produce the most meaning and impact—and consequently—that lead to action. A well-defined question provides the framework for business transformation.

2.       Develop a Business Case

Using the question developed, you want to have a specific goal in mind that advances a business interest so that there is justification for pursuing its answer. For instance, you may have a lot of data from a particular area of your company or customer base, but if there’s little-to-no tangible business impact or return on investment to be had from the analysis, it’s likely not a question worth pursuing. This is where an experienced data science team comes in handy.

At the beginning stage of a project, Quisitive helps businesses to identify the areas of opportunity where their best business case can be made. This is done by performing a gap analysis on data that is available (versus data that is desired), which provides for a more successful outcome that allows for the subsequent development of a roadmap, which indicates which projects to tackle and in what order. Business cases should be created for each question on the roadmap in order to rank the data initiatives with the greatest potential for impact. It is also important to consider the most feasible questions or “lowest-hanging fruit,” since these inquiries—while potentially not generating the most impact on business—may provide the return on investment (i.e., working capital) required to tackle the larger and more significant questions.  Once your primary case is identified, criteria for success can be specified to measure the impact achieved.

3.       Have a Targeted Approach

Once your question is solidified and business case built, you want to approach your data collection process from a hypothesis-testing approach to be most efficient and successful. Businesses often make the mistake of performing exploratory analyses, which involve essentially wandering around the data and testing without discrimination. When you develop a hypothesis, your scope is narrowed. Additionally, it is helpful to bring in subject matter experts who assist in constructing and engineering the most important features of your data. Subject matter experts may also speed up the time required for data cleaning and data transformation. Overall, it is often at this step in a data process—where a team will have chosen to simply explore data in a non-targeted manner—that ultimately leads to failure or a collective feeling of unease with the project. However, when the approach is well-defined, you are more likely to uncover useful information and indicators of success along the way that let you know you are on the right track.

4.       Gain Stakeholder Buy-in

Stakeholder buy-in, which typically comes from within your company and can include other departments and leadership, is critical to the success of any data initiative. Since you are likely conducting your project to take some measure of actionable steps, you need investment in the work, not only to support the process, but also to ensure commitment to implementing changes when it is time to take action.  By ensuring that all necessary members of your team see the value in the initiative—which is supported by a well-developed business case—you have a much lower likelihood of project failure resulting from unanticipated stakeholder barriers.

5.       Understand and Clean the Data

This is the step of the data collection and analysis process that is always underestimated, especially by those not coming from a formal research background. Typically, about 80% of the time spent on a data initiative project takes place here. While it’s the most time-consuming step, it’s also the most important, as when data is not properly cleaned to catch errors early in the process, the analyses is not only slowed down, but the entire initiative now rests on a faulty foundation.  You don’t want the rug pulled out from under the data initiative for lack of results, especially when the culprit for an error that led to this outcome remains unknown. For instance, if during a data transfer, a critical variable needed for analysis has an inordinate amount of missing data or out-of-range data, the results of your analyses will certainly provide incorrect or misleading information. Not surprisingly, this will lead to frustration and may lead to failure of the project. Ultimately, a thoroughly cleaned dataset is well worth the investment, and if done right, insights from the data cleanup will inform the feature engineering steps that greatly impact the project’s success.

6.       Develop a Diverse Team

One of Quisitive’s key tenets is that data is not just a technology project. To have a successful data initiative, you need to pull in the right people and skill sets, including team members with technology expertise, research and statistics backgrounds, and domain knowledge. This not only ensures involvement of both business and technical expertise in a project, but also provides multiple points of input that produce a greater likelihood of catching errors. All of this talent may be found in-house, but often times, personnel capacity, time constraints and expertise require businesses to look to a partner such as Quisitive to fill in the gaps.

In future posts of the Data as an Asset blog series, James and Shannon examine common trends in data initiative projects that can help businesses be more successful in their data outcomes, including common business goals companies are attempting to achieve through data science, understanding why data projects fail, managing expectations in a data process, and common errors in data analysis. Quisitive is committed to helping your company manage and understand its data so as to seek transformative solutions to business challenges.

Connect with James and Shannon on LinkedIn