Why Ternary? Why Now?

A couple of months ago, we formed Ternary to address what we see as the central reason why data science is slow to adopt in most companies - they simply aren’t ready. Data is hard. Plain and simple. Ironically, a lot of the challenges of succeeding with data science have little to do with data science itself. The main challenge has to do with a lack of a solid data foundation. Here are some reasons we’ve observed along the way.

The fog of data science

Data science and machine learning are sexy. It’s easier than ever to learn data science skills. Usually these skills are taught in a vacuum, without the proper context of the challenges of real world implementation.Meanwhile, companies face a data science arms race where they may fall behind smarter, quicker competitors making data driven decisions. Data science talent is very scarce; data scientists are expensive. Every company with the data and the budget is taking part in a data science land grab, gathering talent before wages rise even further, or talent disappears altogether.Business leaders bring in the newly hired talent and wait for the magic to happen. And wait. And wait…

Unrealized value

A common scenario looks like this. An ecommerce company hires a data scientist. Then newly hired data scientist opens their new Macbook Pro, pulls data from an EDW system, identifies a fairly simple algorithm that might yield interesting results, installs the relevant Python library, runs the analysis and voila! Site trends fall out like magic. Or customer segments. Or cross-sell recommendations. Great stuff so far.Our intrepid data scientist wants to serve the model on the website to deliver value to the business. She talks to the ops and engineering teams, but gets stonewalled. Turns out the company has never run Python code in production before, and current processes make new language deployment slow and arduous. It will take six months for Python 2.7 to be ready on the production Redhat VMs. Forget about Python 3.6. In addition, the EDW system can’t talk to the production network.The ecommerce company hired a dozen data scientists...results arrive at a glacial pace, if ever...

Rethinking Data Engineering, Data Architecture and DataOps

Eliyahu M. Goldratt’s classic book, The Goal, is a novelization of the Theory of Constraints. The core idea is that a factory is prevented from achieving its goal of manufacturing product in a timely and cost effective manner by a small number of constraints or bottlenecks. Inventory and partially finished product piles up on the factory floor; capital is tied up in inventory; labor is underutilized when workers are idle waiting for other parts of the manufacturing process.The Phoenix Project (Gene Kim, Kevin Behr, George Spafford) applies these ideas to code and DevOps. Specifically, conflict between development and operations within an organization creates a constraint that prevents deployment of code and delays delivery of value to the business. Merging the two into a DevOps organization and moving to a continuous code deployment model removes the constraint and speeds time to value.We’re now seeing the same pattern with enterprise data. Data sits in data warehouses gathering dust. Models and algorithms grow stale on laptops. Again, the business continues waiting…

The Dunning-Kruger Effect

Companies tend to over-assess their capabilities and resources. Especially with the data science gold rush, it’s tempting for companies to jump head first into data science projects.Our observation is that enterprises tend to lack the competencies to assess their own limitations and execute organizational change. In many cases, enterprises are operating according to outdated best practices and lack awareness of production data science requirements.Meanwhile, the business continues waiting to derive value from data science...

Why we founded Ternary Data

We are recovering data scientists. Data science has tremendous potential, and we have first hand experience with the struggle to deliver value to the enterprise; systems and processes are the key to creating products that impact the business. But doing data science is also incredibly difficult for many businesses. We grew tired of seeing businesses waiting to derive value from their data science initiatives.Unlike some data scientists, we believe that businesses ultimately care about ROI, not just cool AI. Ternary Data reflects our belief that companies can do better with data. So, we are putting data science on the back burner, and helping companies build a solid data foundation. Welcome to Ternary Data.

Thanks,

Joseph Reis and Matthew Housley