Data products: definition, design & lifecycle management
A data productdata productA data asset managed like a product, with an owner, defined users, guaranteed quality, and measurable business value.View full definition → is not a dashboard. It's not a report. It's not a raw dataset.
A data product is a curated, managed, and governed dataset that is treated with the same rigor as a software product: it has an owner, an SLA, documentation, versioning, and quality guarantees.
This distinction matters because it fundamentally changes the economics of data. Raw datasets are internal assets that disappear when the engineer who built them leaves. Data products are organizational assets that persist, scale, and accumulate value.
What makes a data product
A data product has five defining characteristics:
1. Clear ownership, A named individual or team is accountable for this data product's quality, availability, and evolution. Not "the data team" generically.
2. Defined consumers, The data product knows who uses it and why. Consumer requirements drive the product roadmap, not internal convenience.
3. Quality SLA, Committed freshness (data updated within X minutes), completeness (Y% non-null for critical fields), and validity (Z% of records pass business rules).
4. Documentation and discoverability, Consumers can find the product, understand its schemaschemaA schema is the formal blueprint that defines how data is structured, named, typed, and related within a database, file, or message.View full definition →, and trust its definitions without asking the owner.
5. Versioning and stability guarantees, Breaking changes follow a deprecation process. Consumers aren't surprised by schema changes.
Data Products: From Theory to Practice
Knowledge check
1. According to the lesson, what fundamentally distinguishes a data product from a raw dataset?
2. Why does treating data as a product rather than a raw dataset matter economically?
3. How does the Data Product Manager (DPM) role differ from a BI analyst and a data engineer?
4. Select ALL characteristics that define a data product according to the lesson.
Select all the correct answers.
5. Select ALL statements that correctly reflect the principle of 'defined consumers' and good data product practice.
Select all the correct answers.
The data product manager role
Building data products at scale requires a new role: the Data Product Manager (DPM). This role sits at the intersection of domain expertise, technical understanding, and product management discipline.
A DPM for the customer domain would own: the customer 360 data product, the churn prediction feature store, and the customer events stream. They work with data engineers to build and maintain these products, with domain stakeholders to understand requirements, and with governance teams to ensure compliance.
This is distinct from a traditional BIBITechnologies and processes that turn raw data into actionable insights via reporting, dashboards and analysis, so teams can decide based on facts rather than intuition.View full definition → analyst (who consumes data products) and a data engineer (who builds the infrastructure). The DPM owns the product lifecycle.
Airbnb pioneered this model. Their "data product managers" own specific data domains and are accountable for the quality of data products those domains produce. The result: clearer accountability, faster delivery, and higher quality.
Data product design patterns
The Domain Event Stream, A real-time stream of everything that happens in a domain. Example: all checkout events with standard schema. Consumers build their specific views from this stream.
The Aggregate Entity, A curated view of a core business entity. Example: the "customer 360" that aggregates behavioral, transactional, and demographic data about each customer. High value, high maintenance.
The Feature Dataset, Precomputed ML features served to models in real-time or batch. Example: "customer purchase probability features" computed daily, served to the recommendation model. Managed by the ML platform team.
The Metric Dataset, Standardized, agreed-upon business metrics. Revenue, DAU, conversion rateconversion rateThe percentage of visitors or prospects who complete a desired action (purchase, sign-up, contact form), calculated as conversions divided by total opportunities.View full definition →, defined once, computed consistently, used everywhere. This eliminates the "why do the finance and marketing dashboards show different revenue numbers?" problem.
Measuring data product quality
Data product quality is not subjective. Define it with measurable SLAs:
- Freshness: Data is updated within X minutes/hours of source event
- Completeness: < Y% null values for required fields
- Accuracy: Business rules validated by automated tests
- Uptime: Product available Z% of the time
- Schema stability: No breaking changes without 14-day notice
Publish these SLAs. Track them. Alert when violated. Report them to data consumers. This is product management applied to data.
Quiz Questions
- Quelle est la principale différence entre un data product et un dataset brut ?
A) Le data product est stocké dans un système différent
B) Le data product a un propriétaire, un SLASLAA formal commitment defining the service level a provider guarantees to a customer, with measurable targets and consequences if they are missed.View full definition → de qualité, de la documentation et un versioning, traité comme un produit logiciel
C) Le data product est uniquement pour les données en temps réel
D) Le data product est plus facile à maintenir
Réponse: B
- Quel est le rôle du Data Product Manager (DPM) ?
A) Il remplace le data engineer
B) Il consomme les data products pour créererThe ratio of interactions (likes, comments, shares) to reach for a given piece of content, used to gauge how well audiences respond relative to how many people saw it.View full definition → des dashboards
C) Il possède le cycle de vie du data product, entre expertise domaine, technique et management produit
D) Il gère la sécurité des données
Réponse: C
- Quelle métrique de qualité mesure si les données sont mises à jour suffisamment rapidement ?
A) Completeness
B) Accuracy
C) Freshness
D) Uptime
Réponse: C
What to do, from this lesson
These actions are compiled in the role's Playbook.
- Manage critical datasets as data products with owners, SLAs, versioning
Related articles
Recent articles from the blog that build on this lesson.
- DataIf agents are the new primary consumer of your data, is your infrastructure built for the wrong audience?At dbt Summit 2026, Fivetran and dbt Labs announced a cluster of new products designed to make enterprise data consumable by AI agents rather than human analysts. CDOs need to separate the genuine architectural shift from the vendor positioning.
- DataHow JPMorgan Chase built data contracts across 50+ domainsJPMorgan Chase's data mesh initiative forced the bank to confront a problem most large organizations prefer to defer: who actually owns a data product, and what obligations come with that ownership? Their approach to data contracts offers a detailed, replicable model for CDOs managing complex, federated data environments.
- DataFeature stores and the ML data supply chain: a CDO's execution playbookMost ML projects stall not because of model quality but because data preparation is reinvented from scratch every time. This playbook gives CDOs a concrete sequence for building a feature store and treating ML data as a managed supply chain.