3 min readMar 5, 2021


Minimum Viable product and Minimum Viable Data

Minimum Viable Product

What is an MVP?

Minimum Viable Product called as MVP in short, is a version of product with a bare minimum features. These are put together in ways which makes the product usable for early customers.

What is the advantage?

  1. This validates the product against the need of the market.
  2. The feedback from early customers are used to develop further.
  3. It also helps in planning for incremental features
  4. Also cut shorts the time and efforts emancipated to release the product
  5. Helps in learning about the customers with least efforts

While we talk about digital products, we are likely to talk about AI and data. In some cases, they become the core offering of a product.

AI and Data in our products

So, where do we start off? It usually takes a period of time to have ample sets of data to run a machine learning model. So how do we define an MVP for AI and data products? Let’s first look at what do we need to build a piece of intelligence?

  1. Develop a user base
  2. Collect Usage Data
  3. Processing and Cleaning the Data
  4. Build Data Model
  5. Training and re-training the model
  6. Testing

What is MVD?

MVD is the smallest amount of data required to create a potential AI product. Building a model with the least amount of data and data infrastructure as possible. It is about leveraging methods that require very minimal data and over time when data is collected, develop another data model with a method that needs more data. One can start off with a regression analysis, moving on to a decision tree model and then building a neural network as the data increases over a period of time.

What are the advantages?

  1. A sweet spot where data can deliver some value to the customer
  2. Validates the model with feedback loop
  3. Helps build the incremental models better
  4. Enhances the learnability of the user too
  5. Also optimises the time and effort being put

The advantages are pretty much similar. The only thing that remains unchanged is that it has to be lean and it has to be iterative.