Data products as an organizational strategy
The idea of productizing data is still relatively nascent, but it's transforming how data teams operate. If we go back to the basics, the reason to productize anything is to sell the same thing twice. We do it to make things more efficient. This applies powerfully to data as well.
When data teams productize their work, they build dataset(s) or infra once to be used multiple times. This shifts the mindset from creating one-off solutions toward developing reusable data products that can serve many stakeholders. The aim is to maximizing value and minimize effort.
Teams can organize around data products along a spectrum from consulting to platform approaches. Each organizational model offers different pros and cons in how data teams deliver value. There isn’t necessarily a one-size-fits-all model and it depends on a company’s stage. Here’s how I see these models working in practice:
Data team as consultants
Pros: Easy communication with teams and organization around head count (External teams ask for a head count for a pre-specified amount of time). High touch for users. Engineers have high domain knowledge and can make specialized datasets for complicated use cases.
Cons: Data team’s scope is entirely dependent on head count. More projects == hiring more engineers. Neither teaching a man to fish or making fishing easier.
Split scope between consulting and platform so it’s primarily self-service with some specialized data products
Pros: Build out engineering efficiencies; deliver high visibility specialty data products
Cons: Hard to balance. Not fair to engineers to be both good at domain speciality and platform infra
Only platform
Pros: Building out domain-agnostic data infrastructure and core data products. Let’s anyone “fish”. Data team not asking for more head count per dataset.
Cons: Needs executive buy in since the ROI isn’t immediate. Hard political transition and may not be the kind of engineering the team has been built on. Requires customer teams to hire domain-specific analysts to create speciality datasets.
