Expert Commentary

Learn the Essential Practices of Highly Effective Analytics Teams

To be truly effective, analytics teams must adopt best practices at every level.

Di Aarti Gupta and Paul Markowitz

First published on novembre 16, 2016
Tempo di lettura min.


Analytics teams can help promote growth for any organization, by putting customer and operational data to fruitful use. To be truly effective, however, these teams need to adopt best practices at every level, from their approach to basic data hygiene to strategic issues such as team structure and model governance. Otherwise, the analytics team risks having poor relationships with the company’s senior leaders; the two groups will talk past each other and fail to work together toward a common goal. It’s useful to group best practices in three areas: Analytics teams need to have the right data, the right methodology and the right team collaboration model. Each area is essential to success, and while we have seen many of our clients get the first two areas right, they often fall short on the third, leading to suboptimal results. Let’s review each of these three areas. Read More Advanced Analytics Expert Commentary Success with advanced analytics requires both technical know-how and a thoughtful approach. In this series, Bain's experts offer practical advice on some of the most common data issues. The right data. Any analysis is only as good as the quality of the data going in, so proper data hygiene is essential. A few simple rules serve to guide the organization here: Assemble a comprehensive data dictionary and documentation to map and understand the process flows. Document the assumptions and techniques used to build a master data set. Ensure that the organization has the right variables available. Data must align with key business performance metrics, and should allow executives to answer pressing business questions. Keep data current, by measuring and refreshing at appropriate times. The right methodology. Complex or sophisticated statistical modeling is not always required to address a particular business need. A strong analytics team will develop an extensive toolbox of methods, and will know which tool to wield for the specific job, how to apply it correctly, and when to stop tinkering. A few guiding principles: Build a data architecture that can be easily scaled up. Because one size does not fit all, revisit model architecture based on changing patterns in the data. Build the model with appropriate accuracy and precision for the desired business outcome. Don’t overinvest in precision if it’s not required to make a business decision, and be willing to risk an occasional false positive or negative in order to run more models and address more issues. Emphasize testing and validation to assess the models and track performance. The right collaboration model. An effective team will have an overarching set of protocols and norms for ways of working. This includes a structure for how the team receives and prioritizes requests, allocates work among team members, communicates progress and sets expectations, and presents results to the business. Guidelines for successful team governance include these: Collaborate with business stakeholders to determine objectives, and with the IT department to ensure seamless integration. Build forums for model decision rights, so that different departments come to trust one another. Make the model process transparent to all, as a way of increasing acceptance and adoption. Document everything from data dictionaries to model summaries. That will make it easier to answer stakeholder questions and explain model outcomes, build trust and hand off the model to other team members as workloads shift. Communicate the limitations of model architecture—but without undermining confidence in the results achieved. If you believe the findings, make sure your business audience does too. These guidelines do not in themselves constitute success, of course. Analytics teams will still need to excel in other dimensions. For example, they can raise their level of accountability through a detailed work plan that adjusts as priorities change or new requests emerge. They should socialize the model structure with others in the organization. And they can actively explain the effect on business outcomes of model changes, in order to avoid ad hoc requests for information. Yet with a solid foundation in data, methodology and collaboration, analytics teams will advance the cause of strong data-driven decision making. Aarti Gupta is an expert and Paul Markowitz is a principal in Bain & Company’s Advanced Analytics practice. They are based, respectively, in San Francisco and Boston.

Analytics teams can help promote growth for any organization, by putting customer and operational data to fruitful use. To be truly effective, however, these teams need to adopt best practices at every level, from their approach to basic data hygiene to strategic issues such as team structure and model governance. Otherwise, the analytics team risks having poor relationships with the company’s senior leaders; the two groups will talk past each other and fail to work together toward a common goal.

It’s useful to group best practices in three areas: Analytics teams need to have the right data, the right methodology and the right team collaboration model. Each area is essential to success, and while we have seen many of our clients get the first two areas right, they often fall short on the third, leading to suboptimal results. Let’s review each of these three areas.

Advanced Analytics Expert Commentary

Success with advanced analytics requires both technical know-how and a thoughtful approach. In this series, Bain's experts offer practical advice on some of the most common data issues.

The right data. Any analysis is only as good as the quality of the data going in, so proper data hygiene is essential. A few simple rules serve to guide the organization here:

Assemble a comprehensive data dictionary and documentation to map and understand the process flows.
Document the assumptions and techniques used to build a master data set.
Ensure that the organization has the right variables available. Data must align with key business performance metrics, and should allow executives to answer pressing business questions.
Keep data current, by measuring and refreshing at appropriate times.

The right methodology. Complex or sophisticated statistical modeling is not always required to address a particular business need. A strong analytics team will develop an extensive toolbox of methods, and will know which tool to wield for the specific job, how to apply it correctly, and when to stop tinkering. A few guiding principles:

Build a data architecture that can be easily scaled up.
Because one size does not fit all, revisit model architecture based on changing patterns in the data.
Build the model with appropriate accuracy and precision for the desired business outcome. Don’t overinvest in precision if it’s not required to make a business decision, and be willing to risk an occasional false positive or negative in order to run more models and address more issues.
Emphasize testing and validation to assess the models and track performance.

The right collaboration model. An effective team will have an overarching set of protocols and norms for ways of working. This includes a structure for how the team receives and prioritizes requests, allocates work among team members, communicates progress and sets expectations, and presents results to the business. Guidelines for successful team governance include these:

Collaborate with business stakeholders to determine objectives, and with the IT department to ensure seamless integration.
Build forums for model decision rights, so that different departments come to trust one another.
Make the model process transparent to all, as a way of increasing acceptance and adoption.
Document everything from data dictionaries to model summaries. That will make it easier to answer stakeholder questions and explain model outcomes, build trust and hand off the model to other team members as workloads shift.
Communicate the limitations of model architecture—but without undermining confidence in the results achieved. If you believe the findings, make sure your business audience does too.

These guidelines do not in themselves constitute success, of course. Analytics teams will still need to excel in other dimensions. For example, they can raise their level of accountability through a detailed work plan that adjusts as priorities change or new requests emerge. They should socialize the model structure with others in the organization. And they can actively explain the effect on business outcomes of model changes, in order to avoid ad hoc requests for information. Yet with a solid foundation in data, methodology and collaboration, analytics teams will advance the cause of strong data-driven decision making.

Aarti Gupta is an expert and Paul Markowitz is a principal in Bain & Company’s Advanced Analytics practice. They are based, respectively, in San Francisco and Boston.