Establish your analytic team’s organizational structure and team composition

By Pacific Northwest Data Analytics Leadership Board members including: Adam Grupp, Puneet Chopra, and Dave Albano with case study contributions from Jesse Lynch, Caroline Peani, and Mark Belvedere

Establishing your analytics team organizational structure is a crucial step in the process of driving analytics adoption within a company. Many organizations can suffer from poorly considered and designed analytics teams that restrict their effectiveness, often due to too much distance from the business and understanding of its needs[i]. When developing a  team structure, leaders can consider a spectrum between centralized and decentralized organizational patterns with hybrid models in between. Our analysis of best practices show that centralized data organizations can provide an important service curating data and scaling analytics projects. Decentralized and hybrid approaches can help ensure that analytics projects directly serve business needs as analysts are embedded in the business function they serve.

Product teams are a great example of where analysts can be embedded to improve business outcomes. While a traditional product development organization may include product managers and engineers. A multi-disciplinary team that includes data scientists can help ensure greater focus on how data, metrics, and measurement can inform decisions on product features and increase competitive advantage.  

In this article, we have outlined three best practice strategies that help analytics leaders establish an optimal organizational structure and team composition:

1. Utilize a central function to curate common data and provide targeted analytics

2. Apply a hybrid organizational model to maximize the effectiveness of analytics

3. Embed analytics skills into product teams

With these approaches, selected and customized for the unique needs of each company, analytic teams can maximize adoption of analytics and insights.

Utilize a central function to curate common data and provide targeted analytics 

Large companies have highly varied needs for analytics and insights and data preparation can take a significant share (45%+) of analysts and data scientists time[ii]. A centralized data analytic function can identify the most commonly used data sets and curate them for use in analytics leveraging activities such as integrating the data, developing schemas and data models, documenting metadata, reducing redundant data, and improving data quality. When common data sets are prepared and ready for analytics to be performed by internal customers the overall efficiency involved in producing actionable insights improves. The following case study illustrates the value of a centralized team.

A telecommunications company sought to democratize the use of data across the company, however they faced challenges with data accessibility, duplication, and trust. Jesse Lynch, Director of IT Information Management, created a tiger team of data engineers that generated curated data to reduce duplication of data preparation activities. Outside the tiger team, new users in the business and functional areas could request access to data and get it quickly. The front-end experience allowed users to explore data without SQL coding skills. This was combined with capacity for new, urgent analytics requests further meeting the insight needs of the business. Supply chain, marketing, finance, and credit teams benefited from more curated data and easily analyzed insights. To learn more, visit the case study at Jesse Lynch Telecommunications Case Study (pacnwdataanalytics.net).

Apply a hybrid organizational model to maximize the effectiveness of analytics

Centralized teams can provide valuable data analytics services, but their effectiveness often suffers from insufficient understanding of business context and needs. A McKinsey study found that 90% of companies with a breakaway level of success embed analytics into their business functions[iii]. To enable this, many companies create a hybrid analytics organizational model that is neither centralized nor decentralized. The hybrid model promotes the use of distributed analytics capacity embedded into the business units where the insight needs are best understood. The closeness to the business helps with relevance and adoption of the insights. A central team is still important for enterprise-wide analytics use cases and scaling of large projects. The following two case studies illustrate the value of a hybrid analytics team.

A large technology company sought to grow customer engagement and revenue across consumer products. Their centralized analytics team failed to provide the right value so Adam Grupp, Loyalty Program Director, applied a distributed model that embedded analytics practitioners within functional teams. This helped, but these smaller groups of practitioners couldn’t complete the bigger analytic initiatives required to scale the business and revenue. This led them to a hybrid approach of embedding analytical members within other teams while maintaining a central analytics group responsible for enterprise-wide experiments. This more flexible hybrid approach improved trust in analytics, enabled them to identify and scale their most successful marketing tactics, and ultimately achieve their profitability goals. To learn more, visit the case study at Adam Grupp Technology Case Study (pacnwdataanalytics.net).

The payments division of a large marketplace company sought to ensure that analytics better influenced product design, features, and performance. On the one hand, analytics team members are often too far removed from the business to understand what insights would be most helpful. On the other hand, when analysts are embedded in the business unit, they may have a better understanding of business needs but lack best practices in analytics. Mark Belvedere, Head of Global Data Science, deployed an organizational model with analytic resources that are embedded in the business yet have performance accountability back to his central analytics group. With this model, analytics team members worked closely with product teams enabling them to ensure relevance of analyses. The more effective organization design led to increased adoption of insights as their relevance, quality, and value increased. The business was able to grow the impact of their product and associated revenue through application of these winning insights. To learn more, visit the case study at Mark Belvedere Marketplace Case Study (pacnwdataanalytics.net)

Embed analytics skills into product teams

Many organizations sell products to customers, but designing products with the telemetry and analytics to gain insights on their performance is often an afterthought. Leading companies are promoting more data-driven design by embedding analytics practitioners on their product development teams. Data scientists on product teams, for example, can plan for future analytic use cases that can inform product feature improvements with data analysis. They can also determine metrics that help assess the extent to which features are working and then shift to other features more quickly if it becomes apparent they are not.

Caroline Peani, Director Data Science, found substantial benefit in embedding data scientists in product teams as they could work with product managers and engineering leads to inform design decisions throughout the product lifecycle. With the data scientists focused on establishing metrics and measurement early, the product team could use the metrics to understand whether an idea or feature was meeting business goals and shift to other approaches quickly when it was not. This helped to ensure accountability of product leaders through evaluation of success using a quantitative measurement of their ability to move the metrics. With these product success measurements, the overall business can integrate more data-driven decision-making in their product strategy, helping to achieve competitive advantage.  

In summary, combining optimal organizational design with multi-disciplinary team composition can result in major improvements to the usefulness of insights and their adoption.

[i] (PDF) A study on organizational design and operational planning of big data teams (researchgate.net)

[ii] Anaconda | State of Data Science 2020

[iii] Breaking away: The secrets to scaling analytics | McKinsey

Published by

Pacific NW Data Analytics Leadership Board

242 followers

Published • 2mo