At my current job (we’re hiring), I work with a number of companies that are working to become more data-driven. It’s hard work. Not just the technical aspects, but the cultural changes that are either a pre-requisite or a result of data-driven decision making.
In this post, I want to talk about the improvements that are possible with a proper data culture as well as how to start making the changes necessary to implement a data-driven culture.
Where to Start?
There is a huge disconnect between the ideals some people have of data-driven organizations (We’ll use AI! It will be magic!) to the reality: a lot of work by a lot of engineers to make the data consistent and usable first.
The key to success with data projects is having a crystal clear idea of the business problems you want to solve and then focusing your efforts on the most approachable ideas.
Things that are (relatively) straightforward to improve with data:
- Internal operations
- Customer segmentation
- (Digital) product improvements
These are broad categories, but most companies can find straightforward ways to use data to have a measurable impact on results. And more than almost anything else, being able to measure the impact and ROI of data projects is critical.
How can we approach these topics and make an impact?
We want to build a time-boxed experiment with the key goal of building internal skills around data. It’s critical that the learning and growing aspect of these projects is stressed. Initial attempts at being data-driven may not have huge ROI payoffs in the short term. Obviously, if they do, that’s ideal, but there are many reasons why these returns don’t happen immediately.
As you do these experiments, it’s critical that you move business metrics from being descriptive to predictive.
Evaluate your Data Maturity Level
Every company, even every division within a company, is at a certain level of data maturity. Data maturity is a concept that measures your ability to use data to drive your business success.
I would say most companies are, at best, level 3. And typically it is only parts of the company that have even gone that far such as marketing or finance where analytics have become normalized.
©2017 TM Forum
Data maturity has many sub-domains. In order to be successful, the business needs to mature across all the different sub-domains. Technology skills are not sufficient. It’s critical to have the right processes, controls, and interpretation skills to really get value out of data projects.
If you aren’t sure where you in terms of data maturity, here are two quick “gut check” mechanisms:
Once you’ve picked your project, you need to align stakeholders across the organization around the necessary technical, procedural, and governance aspects of your project.
The typical sequence of needs is Data Infrastructure (cloud) > Data Engineering > Data Science > Visualization/reports/metrics.
Suggestion: If you aren’t already using cloud tech, start.
Data storage needs almost never get any smaller. It’s a lot easier and cheaper to scale using one of the cloud computing providers. Beyond storage and processing, AWS, Google Cloud, and Azure all provide a variety of data analysis tools that are mostly “plug-n-play” once your data is in the right place.
Build an organization-wide data architecture
The maximum value in data projects is realized when data is shared and utilized across the organization. It’s best if some thought is put towards a consistent data architecture for the organization from the beginning.
The actual technologies used can vary depending on the nature of your company and your data sources, the cloud provider you use (or if you do not use one of the cloud options, the on-prem options). This is why it’s important to find the right people who can guide your organization, whether internally or externally.
- Begin moving to the cloud for scale and superior tooling.
- Evaluate and document your key business cases and data maturity.
- Evaluate your organization’s internal skills and capacity for change. Make plans to build those skills or hire or them.
- Create a set of predictive experiments that can answer your questions.
- Measure results, get some wins, iterate.
If you need help, don’t be afraid to reach out to Wizeline.
This blog was originally posted on Medium. For more posts from Director of Engineering, Brenn Hill, please follow him on Medium: https://firstname.lastname@example.org