Data Product Canvas — A practical framework for building high-performance data products

Data Product Canvas Framework
The Data Product Framework : Data Product Canvas — DPC


Do you know why most data products fail? It is because they are not always aligned with the real objectives of the companies.

What is the Data Product Canvas?

It is a framework for the development of data products, based on a Canvas model, which follows the principles of the Agile/ Lean methodology. Its main objective is to serve as a practical tool for generating a data product roadmap, aligning in a single document the complete view of all those involved about the real purpose of the project.

What is it for?

The idea of ​​the DPC is to ensure a practical and objective planning of the data product based on a common understanding between the technical and business areas. In this way, it is possible to map the entire project from end to end, thus allowing a single view of what was planned and what will be executed. In addition, it also helps in creating actions and monitoring the main metrics to follow the solution that will be developed throughout its life cycle.

Data Product Canvas — 10 blocks divided by 3 domains
The Data Product Canvas is divided into 10 blocks and separated by 3 domain areas.
  1. Problem definition;
  2. The solution that will be adopted;
  3. Data mapping;

Why use?

The success of a data-driven culture depends on the definition and implementation of strategies , not technologies. That’s why it’s important to make it clear that data products are a business domain problem, not a technology one.

How to use?

1. Start with the problem

How to define the problem by using DPC
How to define the problem.
  • What ‘s the problem?
  • Why is it a problem?
  • Whose problem is it?
  • What ‘s the problem?
  • Why is it a problem?
  • Whose problem is it?

2. Try to identify the solution that will be adopted

DPC — identifing the solution
How to identify the solution.
  • What kind of solution will be adopted? (Ex.: Analytics, Machine Learning , AI, etc.).
  • What will be the solution? For example:
    If the adopted solution is Machine Learning, we must take into account that: for each problem, we have different approaches. For each approach, several algorithms. And for each algorithm, several parameterizations. That is, there is not and never will be a “best algorithm” for a given problem. But in any case, having the mapping of what you want will provide guidance for the development of the project.
  • What is expected of the solution? What would the outputs of the product be?
    Eg: A report with the final product of an analysis? A specific prediction about a data type?
    Oh, and don’t forget, at the end of each of these questions, always ask: why, why, why???

3. Map all Data

DPC — mapping the data
How to map all that you need about the data.
  • What is the source of the data, ie what is the source? (Ex.: Is it on a system? Is it a set of files? Does it have structured formatting?)
  • What is data quality? Are they sufficient for analysis?
  • Accesses vs. Availability — Do you have access to the data? Are they available?
  • Process / Transformation — Is it necessary to establish a process for reading the data? Will there be any transformation process?
  • Outputs — What are the output formats?
  • Test / Training / Validation — Are there any strategies or assumptions about test, training and validation data?

4. Which hypotheses will be tested

Indentify the hypothesis
What hypotheses will be tested?
  • What are the hypotheses we want to test?
  • What are the expected responses for each of them?
  • What to do from each answer? In other words, what strategy should we follow?

5. Identify all actors (customers and stakeholders )

DPC — how to identify all actors?
How to the identify the actors?
  • Who is the sponsor?
  • Who is the final customer of the product?
  • Who are the interested parties and stakeholders ?
  • Who will use the solution?
  • Who will consume the solution?
  • Who will be impacted by the solution?

6. Plan the strategic actions that will be implemented using the solution

How to plan the actions?
  • What actions will be used?
  • Which campaigns should be created?
  • How to generate value for the business from the use of the data product developed?

7. Create the KPIs that will be used to monitor the entire product

DPC — how to crate KPIs.
How to create the KPIs to monitor the product along the journey?
  • How to evaluate the quality of the finished product? (Ex. if it is a Machine learning model we can use its accuracy rate.);
  • What metrics should be used?
  • How to measure the results of actions?
  • If it’s with A/B testing , how?
  • How much uncertainty can we deal with?

8. Estimate project values

DPC — Estimating product values
How to estimate values.
  • How big is your problem ?
  • What is the baseline?
  • What is the expected gain or savings from using the product?

9. Map the risks

DPC — Mapping the risks.
How map the risks.
  • What are the risks?
  • What could these risks block during product development?

10. Identify the performance and impacts that the product will generate for the business

DPC — Identify performance and impacts.
How to identify performance and impacts.
  • What is the impact for the business?
  • How to measure it?
  • Where and how can we see this improvement or impact/performance?


Thus, after presenting the 10 blocks that define the roadmap of a data product using the Data Product Canvas framework, we have reached the end. I hope this description makes you feel more comfortable and confident in creating high-performance data products. And remember: avoid creating the right solution for the wrong product.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Leandro Carvalho

Leandro Carvalho


Data Science Manager | Data Scientist | Machine Learning Specialist | Professor | IT Manager. LinkedIn: