5 Rules for Successful Self-Service Analytics
Making the transition to self-service analytics can be fraught with challenges. These guidelines will help you start with a strong foundation.
- By Andrew Roman Wells, Kathy Williams Chiang
- March 21, 2017
Business leaders are seeking out the power of self-service analytics because it allows them to create and manage reporting and analytics tools without involving their IT organizations. However, as more business teams move to self-service analytics, these groups are struggling with how to best leverage and scale this new capability -- and to overcome new challenges that business teams have not faced in the past.
To successfully make the transition to self-service analytics, we recommend adopting these five rules:
Rule #1: Embrace good data quality practices; they lead to trusted data and faster analytics
Every house needs a foundation and data quality is the foundation for any analytics process. Knowing that your data can be trusted and meets the criteria of data quality standards allows the analyst to work on the analysis instead of spending time scrubbing data sets or worse, not being able to use certain data due to the lack of data validity.
There are six main components to data quality:
- Completeness of the data -- checking for any missing or partial data in the data set
- Consistency requires that the values of the data must be consistent throughout the data set
- Checking for data duplication refers to removing or correcting records in a database that are exact or partial duplicates of each other
- Conformity refers to how well the data adheres to standards
- Accuracy addresses the validity of the data and how many errors are in the data set
- Integrity of data refers to the accuracy and consistency of the data over its life cycle, ensuring that when the data moves from system to system it maintains a level of quality and standardization
Rule #2: Develop standards that act as guiderails to ensure team effectiveness
As a team begins this journey, it is easy to develop a lot of one-off metrics, data definitions, and reporting user interfaces (UI). We highly recommend developing a set of standards that the team adheres to. This will encourage reuse of best practices and avoid duplication.
For example, one analyst might create a metric called "average basket sale" that represents the average number of items purchased in the basket. At the same time, another analyst creates the same metric, but this version refers to the total dollar amount in the basket. The end user of the report now has two different definitions for the same metric, creating confusion and lack of trust.
To avoid these types of issues, create standards. Data definitions and lineage standards can be documented in a data dictionary that the team members use to define each data element or metric. It should include information about where each data element was sourced and any transformations that were applied. User interface standards refer to the look and feel of the reports the analysts develop. A standard look and feel will help users quickly interpret multiple reports rather than having to relearn the UI for each new report.
Rule #3: Create a report certification process to engender trust in the analytics
As more teams generate reports that are consumed by broader audiences, the standards and practices leveraged by the different teams will vary. One team may not use a source of trusted data; another team may not use the company's data dictionary to create their metrics. As reports proliferate, the report consumers will not know which ones they can trust. A certification process aligned to standards can help a report consumer know the level of scrutiny and rigor that was applied to the report.
Rule #4: Embed data science in your reporting solutions
Many business analytics teams employ a full-time or part-time data scientist. This person's skills allow the team members to generate deeper insights from the data to aid in making better business decisions. One gap that we often notice in these teams is the insights generated from the data scientist are usually a one-time study or only repeated on request. Many of these valuable insights can instead be leveraged by your team's analytics solutions, enabling consistently better decisions for a broader audience.
Rule #5: Develop security practices early to avoid data breaches
We often read about data breaches in the news when a company is hacked and their customers' credit card information is exposed to the world. Although most analytics do not involve personal consumer information, there is proprietary business information that you would not want your competition to discover.
Developing security standards now can help avoid mistakes in the future. The simplest security control is to password protect sensitive reports or analytical studies. This can often be accomplished via the reporting tool.
If you have a more comprehensive tool, you can get very specific with user security -- down to which data elements a particular user can access. You should encrypt your data, especially if it is in the cloud. When sending information via email, especially outside of your company, a simple encryption program can secure data assets against many hackers.
A Final Word
Making the transition to self-service analytics can be tricky and fraught with challenges. Implementing these five rules will create a strong foundation for your team to build on.
For Further Reading: