Getting Started with Data Analytics Use Cases
We've learned a few lessons when it comes to the process a business can take to maximize success with data science.
More than ever, businesses across every industry are looking to incorporate data analytics to help drive value throughout their corporation. What’s changed is the nature of the questions we're receiving from customers. In a nutshell, we’ve seen our primary engagement pivot from collecting data to now supporting data analysis and curation.
Effective data science use cases
WWT’s Business and Analytics Advisors (BAA) team provides consulting services in analytics, AI and enterprise architecture engagements. Based on our experience delivering data science work with our clients, the BAA team has found some commonalities in the process a business can take to maximize success—we have also learned from challenges, road bumps and lessons along the way.
The key elements of designing and developing successful data science use cases are:
- evaluation of the data required;
- selection of the analytical processes and data science techniques needed to enable the use case; and
- hierarchy of the type of outcomes we expect to achieve from the use case, from understanding how the output would look like to expected impact on business KPIs of each successful use case implementation.
Using the above as a structure, let’s investigate each one in further detail.
Systematically collecting high-quality data provides the confidence for decision making.
An analytics use case is as effective as the data collected and used to enable it.
This begs the question, what are the characteristics of “good” data? What are the key criteria that can help us distinguish good vs bad quality data?
Characteristics of “good” data for analysis
- Key Data Characteristics: Frequency of refresh, granularity and completeness suit the desired purpose.
- High Quality: Data is populated with minimal “gaps” in elements that are important for analysis.
- Well Governed: Level of trust is built that the data worked with is accurate and contains what analyst thinks it contains.
- Sources of Truth: A “single” source of truth is not needed for a given piece of information, but a single source for each piece of information and context is needed.
- Easily Accessible: Well-documented inventory of datasets can be easily used by the people who need them.
- Data Linking:Data sets can be "joined" together as needed based on key data element.
As companies grow in their data maturity, the challenges have migrated from collecting large quantities of data to curating good quality data that is useful for impactful data analysis. Here are the most notable challenges with data strategy that we have observed:
The purpose behind collecting high-quality data is to eventually convert raw data into meaningful, actionable insights. There are varying complexities of use cases that businesses can choose to move forward with. The decision of choosing which type of use case a business should employ is dependent on the availability of resources, degree of data maturity and the stage that a client is at at the time of engagement.
Businesses can extract increasingly powerful insight from raw data as they become more data driven.
We remind clients that the output of a use case can vary, from something as simple as a report on a spreadsheet, to an application requiring significant software development.
Data analytics output can influence tactical and strategic decisions.
Once a use case has been conceptualized, the question of prioritization arises. Many organizations prefer to focus on high-impact, low-cost initiatives. The following outline several areas of complexity that affect the development of each use case.
Evaluating each use case's ease of implementation
But not all use cases make for optimal big data analytics. This is where WWT often provides customers with expertise as an independent advisor on data science, strategy and data governance.
What makes a use case sub-optimal for data analytics?
We learned that it isn’t always necessary or efficient to pursue the data science route to an engagement. The key is to not force complex analysis under the assumption that it will always lead to higher gains. Here are some examples from client engagements where we came across sub-optimal uses for analytics:
- Limited impact on KPIs: An engagement with a large fast-food business could have taken an AI predictive approach to forecast customer demand—the gain was small enough that it became more practical to build a small set of business rules that ended up delivering a similar outcome more efficiently.
- Quality of data is poor: Working with a large mining company interested in using IoT sensors to predict truck failure, a significant data challenge was differentiating between sensor information that indicated machine failure and poor sensor data collection on a perfectly well-operating machine.
- Complex modeling approach not worth analytical gain: While encouraging, the enthusiasm from businesses to invest in data analytics can sometimes lead to forced analytics. It is important for companies to fully flesh out the gains from smaller use case successes before diving into complex and costly analytics initiatives that may result in minimal business impact.
From concept and design to implementation
Now that we’ve designed a use case, evaluated its complexities and decided that it’s suitable and effective to be taken on as a data science initiative, the natural next step is implementation. Shown below is the process we find most clients engaging with when designing and implementing use cases.
The analytics process can vary based on the degree of complexity of the use case.
For businesses that are on the cusp of transforming their data for meaningful analysis, the key is to follow a logical process. The above method has been generalized from WWT’s interaction with a variety of clients across industries, and we have noticed greater chances of success when following the outlined process.
The BAA team continues to work with organizations to create successful data governance and analytics strategies. Analytics use cases add value to customers in healthcare, retail, higher education, manufacturing and other industries where large amounts of valuable data are collected.