Have you ever spent weeks working on a machine learning project, dashboard, or analysis only to have to go back and redo your work because of feedback from stakeholders? I know I definitely have. So what’s the solution?

Ship Early and Often

During my time as a data scientist at Apple, I realized that the idea of going into a cave to work on a project for several weeks is the antithesis of what a good data scientist should be doing. This type of approach leads to longer time lines for deliverables since you’re not integrating feedback early on when it’s the easiest to do so and you’re also missing opportunities for building closer collaboration with your stakeholders. Instead, good data scientists are proactively communicating by shipping early and often to solicit feedback and engage in dialogue with their stakeholders.

Here are three example scenarios that encourage the ship early and often mentality that you can apply in your future projects:

Machine Learning

The majority of your time in a machine learning project is spent on building a training dataset and feature engineering. As a result, the quickest way to prototype something is to only gather a few key features and spend minimal time with feature engineering. Taking advantage of AutoML solutions like TPOT (http://automl.info/tpot/), can also greatly reduce your time to an initial prototype. With a prototype in hand, you can create a backtest that you can take to your stakeholders to start conversations about whether you should invest further. 


Doing any kind of work in Tableau can be a frustrating experience because the process of revising and adjusting your views tends to be incredibly manual, which makes the prospect of revisions cumbersome. Instead, tools like Sketch or Figma allow you to create quick mockups, saving you countless hours in the process. The main thing here is to get feedback early, mockups are always easier to comment on than a text-heavy requirements document (though these are also important).


With analysis, you can benefit from creating a rough story of what you want to say. Sometimes you don’t know until you actually do the analysis, but other times you can paint with broad strokes the approach you plan on taking and the questions you’d like to answer. You can even get down to the level of detail of what type of visualization you plan on showing.