The Product Organization

 There are many ways to construct a data product organization, and each approach has pros and cons. (See figure 1.) 

Decentralized Model. A decentralized model, similar to the data mesh, pushes product creation and management to business domains, each of which will have their own DPP to store and publish data products. There is a central DPP that pulls data products from each domain DPP. This approach preserves local autonomy and avoids central bottlenecks but may underserve some domains. 

Centralized Model. A centralized model consolidates project management on a single team that publishes data models on behalf of domains to a central DPP. This improves efficiency and standardization but creates bottlenecks and reduces domain autonomy. 

Hybrid Model. A hybrid model takes the best of both decentralized and centralized models and eliminates most of the downsides. Authorized domains create their own data products and publish locally and to the enterprise DPP which is run by a central team that supports all domains that don’t have product development and management skills.

“Most organizations will end up with a hybrid model that delivers the best of both centralized and decentralized approaches with minimal downsides.”

Figure 1. Product Organization Models