Today’s businesses collect more data from more sources than ever before. From weblogs to transactional data, the Internet of Things (IoT) and everything in between, I don’t think any company today would say they have a data problem.
But I do think many companies, whether they like to admit it or not, have a data value problem. That is, they struggle to gain real business value from all of the data (or any of the data) they’re collecting.
In the beginning, digital-native tech companies (like Google, Apple, Facebook, and Amazon — GAFA) created value from data essentially by applying advanced machine learning techniques to a few key problems (how to make ads relevant, recommendations effective, etc.). Their problems were technically fairly challenging, but the means to solve them was fairly simple: hire 50 PhDs and talented engineers, and you’re probably bound to succeed.
Traditional enterprises, on the contrary, have a murkier path to success because they have to transform and optimize existing products and services step-by-step, and their business problems are not only technically difficult, but also difficult to work with. As a consequence, these organizations need to adopt a more systemic approach, looking for productivity gains in the same way one looks for productivity gains in a factory:
- Set Up a Reuse Methodology. This means setting processes so that data and results can be effectively shared from one project to another. In a typical organization, 80 percent of data projects are started from scratch in an attempt to control what’s happening. That’s because reuse requires some documentation and discipline. But come on! You can do it.
- Multiple profiles working on data projects together. This means data scientists, of course, but also analysts and business people. Data science (and business-impacting insights) don’t happen in a vacuum. Just as those on the business side don’t know how to do hard-core data science, hard-core data scientists are usually not so in touch with the business. So collaboration is key.
- A way to deploy to production at scale. Again with the vacuum — data science can’t happen inside one. Data teams have to be able to regularly push things out into the world instead of working in a sandbox and not having any real impact. Being able to do this seamlessly and quickly is critical to scaling insights from data.
Without a doubt, the amount of data will only continue to rise. And those that scale their insights along with it will come out on top.
Join 30,000+ people who read the weekly 🤖Machine Learnings🤖 newsletter to understand how AI will impact the way they work and live.
Companies don’t have a data problem, they have a data value problem was originally published in Machine Learnings on Medium, where people are continuing the conversation by highlighting and responding to this story.