How Madrona Venture Labs Used
Slack Data to Visualize Employee Engagement & Manager Effectiveness




  • Testing & validating an idea that you can mine company communication data...and predict employee attrition


  • Data visualizations on relationship dynamics between employees and managers
  • An information infrastructure to receive data from a publicly traded company


  • Pursuing their 1st round of funding


  • Human Resources


Madrona Venture Labs

Madrona Venture Labs (MVL) is a startup studio based in Seattle, WA.

MVL is housed inside Madrona Venture Group, a venture capital firm. Their charter is to think up startup ideas that align with the firm's investment thesis.

Through a systematic vetting process, MVL will try out new ideas...

And ask lots of questions about the market, the customer, and the product.

Lots of ideas get killed in the process. 

But typically, they will spin out 2-3 companies per year with founders and financing from Madrona Venture Group and other outside investors. 



Testing and validating the idea that you can mine company communication data...and predict employee attrition

In the summer of 2017, MVL went to work on a startup idea that helped predict attrition for tech employees...

But quickly pivoted to a higher level goal of understanding employee level engagement and manager effectiveness.

Based on their initial market research, they knew that survey-based companies were antiquated and flawed, and wondered if they could do something more interesting by mining communication data. 

Thanks to an initial test from a small technology company and 22 years of startup experience, CTO Jay Bartot knew natural language processing was going to be a big part of this project...

And then recalled a conversation we had last year. 


Pretty much every data scientist has some, usually, interest or experience working with text as an input form...

Just a quick chat with Matt told me he was pretty up-to-speed on what the latest going ons were in this natural language understanding space.




Madrona Venture Labs



Data visualizations on the relationship dynamics between employees and managers...and an information infrastructure to receive data from a publicly traded company

As soon as Madrona Venture Labs had data to work with, I got to work unearthing signals in the data to see how people within a company communicate. 

Despite our progress, we quickly ran into an issue when we were preparing for a data dump from a pilot customer (who happened to be a publicly traded company). 

So before we even retrieved had the data, Jay knew that...

  1. Infrastructure had to be created
  2. And it needed to be secure and buttoned up. 

So he contacted a friend (who happened to be an AWS guru), and was able to get great advice and assistance on setting up a secure environment.

But what surprised Jay the most was when I stepped in to configure the server.


Matt also had a number of recommendations on how we should set up different servers and configure the environment and so forth.

That's when I was like, "Oh, this guy knows more than just data science."



Madrona Venture Labs is currently working with their 2nd customer and pursuing their 1st round of funding

Nearly a year later, MVL is ready to spin off the startup idea into a separate company. 

They are assisting with the pitch presentation process, using approximately 150 different micro signals discovered in the Slack data.

Jay knew there was a risk utilizing fairly expensive data science resources to validate this startup idea, yet believes his investment has paid off. 

I continue to work with the MVL team to transition this startup idea into an independent company. 

There's some data scientists or engineers, too who you have to  tell them what to do. 

Whereas, the kind I like to work with the most, just given the chaos of early-stage companies, is people who can be self-directed and strategic and can think independently. And Matt's all those things.


Test your startup idea like
Madrona Venture Labs

Why not use data to validate your idea?