One of the biggest challenges that every data-driven organization faces is managing the various analyses the teams develop. If your analysts are scrambling to complete all the work leadership is requesting, or you are having trouble getting consistent and accurate results from the team, there may be problems with your analytic governance model. This post discusses seven (7) mistakes organizations make with their analytic governance models, and ways to address them.
What is Analytic Governance?
If the term analytic governance is new to you, then let me take a moment to explain the concept. The concept of data governance has been prominent in business for about 20 years. Data governance refers to the policies and procedures an organization creates to ensure its data are high quality, are accessible to those who need it when they need it and are kept secure from unauthorized breaches.
Analytic governance is like data governance, but focuses on managing analytic activities with respect to the following components:
- Data owners
- Subject matter experts
- Routine vs one-off analyses
- Alignment of skills with analytic demands
A good analytic governance model helps ensure that your organization is efficiently identifying and completing analyses with consistent results that are shareable across units. Analytic governance also helps prevent duplicative work for your analytic team, reducing overall workload and increasing the quality of results.
With a working definition of analytic governance in place, let’s turn to common challenges organizations face with implementation, and ways to address them. You will notice in this discussion that several of these issues overlap with data governance models, an indication of their overall importance.
Too Many Kingdoms
The first challenge is one that lands squarely in both the data and analytic governance arenas. Teams that collect and maintain organizational data may silo those resources within the owning team and prevent direct access to other groups in the organization. Worse yet, data owners may treat the data they hold as part of their kingdom, only allowing analyses they deem worthwhile to proceed.
To prevent data kingdoms from forming and becoming institutionalized in the organization, revisit your data governance model to ensure there are mechanisms for sharing data between groups. One approach includes developing data documentation to share across the organization, educating users on the contents and caveats of each data source. Creating standardized data request forms for specific data sources ensures that when different users make similar requests, they are receiving consistent data in a searchable and trackable format.
Another approach is to develop cross-functional teams for the development of new analytic projects. Set explicit expectations for the team for collaboration and that data is to be shared freely between team members. The primary purpose of the cross-functional team is to identify and address specific data requirements and issues. By staffing the team with members from different data-owning units, you also reap the benefit that data-specific knowledge can be shared across units.
Too Many Generals
Just as problems arise when there are too many kingdoms, the same can happen if there are too many generals. Analytic capacity is a finite resource in every organization. Business owners and leaders will always have their pet projects and interests they want to pursue. You need to have a way to track the overall load on analytic capacity so that the demands coming from various sources do not overload the teams.
When business owners place too many demands on analytic teams, the likelihood of the following problems increases:
- Analytic errors
- Poorly designed analyses
- Incomplete analytic validation
- Ignoring established protocols in favor of quick results
- Staff stress, burnout, and turnover
You need to coordinate across your business owners to set analytic priorities and track analytic workloads. Adjust when necessary to accommodate new projects and special needs. Pay specific attention to the procedures for handling ad hoc analyses so that business owners do not unintentionally flood your analytic teams with one-off analyses they believe are quick to complete or simple.
Too Much Overlap
Different business owners in an organization often have overlapping concerns. As a result, multiple requests may be submitted to analytic teams for similar projects, often producing different results. The duplicative nature of the projects and variation in results creates conflicting messages across teams that can confuse decision-making.
As a part of a good data governance model, you should collaborate with your teams to develop an agreed upon set of metrics for analytic projects on various subjects. Additionally, when there are multiple analytic teams, the plan should identify which teams will serve as the source of truth for calculating each metric.
Finally, you’ll want to vet the methods used for each metric and routine analytic process to identify the most robust approaches. Do not let seniority or “that’s the way we’ve always done it” drive poor methods to the front.
Every Project Starts from Scratch
In contrast, you don’t need to reinvent the wheel every time a data question comes up. Identify the critical key performance indicators (KPIs) you always want available for analysis. Develop processes and a schedule with your analytic teams to ensure these metrics are always available in a timely manner for inclusion in analytic projects.
You can also build a database containing values for KPIs tracked over time so you can trace quality and process improvement efforts over time. During the design phase of the database, don’t put everything and the kitchen sink into this effort. While it’s tempting to add lots of data elements and features into the design of your database, the longer the list is the less likely everything is to be completed on point and on time. Your mantra needs to be, “Just because you can doesn’t mean you should.” Take your time, set priorities, and let the project evolve incrementally.
Every Project is A Fire Drill
Until you have a mature data process in place, most analyses will require non-trivial lead time to generate results. Too often, non-analytic staff assume that data analysis is a quick process where the results you want are just a button push away. This leads to increases in the number of analytic requests that leadership needs ASAP, right now, or worse…yesterday. In the worst circumstances, business owners begin to institutionalize the need for rapid results, and everything becomes a fire drill to complete.
While changes in business environments and circumstances can always lead to the need for rapid analytic decision-making, it is management’s responsibility to keep this to a minimum. Constantly asking for immediate results has the same effect of overloading your analytic team, leading to stress, burnout, and turnover. As the old saying goes, “A failure to plan on your part, does not constitute an emergency on my part.” The same holds true for your analytic team.
Inconsistent Data Collection Systems
Setting up your data collection systems properly will have a substantial impact on your analytic team and projects. When staff collect the information and enter it into your data system in unusual ways, analyses become more difficult at best, and impossible at worst.
For example, having all your data fields set up as free text entry allows users to enter information in any format. Imagine trying to track customer referrals and your staff enter the referral source as a company name, an individual’s name, and a customer ID number all in the same field. The resulting data will require that your analysts spend additional time to clean up the inconsistencies and make the information useful.
By setting up your data systems to capture information consistently, you will create a cleaner database that will be more readily available for analysis. If you don’t have a system with features designed for more consistent data entry, then set standards for your team on how data should be captured in the system. Develop documentation on which fields to use, in which format to enter the data, and methods for verifying all entries are correct. Work with your team to focus on accuracy and consistency, and your database will become much more useful for your analytic projects.
Imbalanced Analytic Teams
Regardless of whether your organization is just starting to stand up an analytic team, or has an established analytics group, aligning the skill sets of your team with the work they will be doing is critical for successful analytic governance. Depending on your analytic processes and goals, you may need more senior data scientists, or you may need more junior analysts to manage databases and pull reports. Review your ongoing analytic projects, and the ones you have planned for the coming year to determine the skills that will be needed.
As you review your team balance, keep in mind that your staff will need to have interesting enough projects and enough variety to stay interested in their roles and develop new skills. If you have higher end data scientist staff with a lot of complex projects to develop, they will need junior staff to assist with some of the more mundane and repetitive data management tasks.
Similarly, your junior analysts need to have work that is within their capabilities, but also have more senior staff to learn from, and have opportunities to develop new skills over time. If you don’t have a senior analyst on staff, then providing opportunities for training your junior analyst can build both capacity and loyalty from within.
Every team has turnover. If your analytic team is imbalanced for the job they need to do, however, this will lead to higher levels of frustration and turnover. The downstream result for your bottom line is greater costs in recruitment and training. By considering your analytic goals, and aligning your team accordingly, you can build a balanced and sustainable infrastructure.
Conclusion
To make the most of the data you have available, you need to couple good data governance with good analytic governance. Setting up an accessible, integrated, and secure data ecosystem is most useful when you have also created a solid strategic decision-making structure. Addressing the analytic governance issues discussed in this post will help you build a more accountable, streamlined, and internally consistent analytics process. The benefits of your improved analytic governance model will be more consistent information for decision-making, greater reliability in analytic results, and a more efficient and happier analytic team.
