A Cloud Data and Analytics Platform Has Major Benefits, But That Doesn’t Mean it’s Easy to Manage
As the phrase “move to the cloud” becomes ever-present, it’s critical to understand that just like any data engagement, a cloud data and analytics program needs to be treated with the same attention to strategy, process and testing. While we’d like to believe that it’s as simple as spinning-up a new environment and moving data, doing the project correctly (and planning for future growth and success) is actually far more involved.
There are three things to consider when moving your data and analytics to the Cloud. In an increasingly digital world, we’re seeing a common trend with our clients: the misperception that cloud equals easy. As the phrase “move to the cloud” becomes ever-present, it’s critical to understand that just like any data engagement, a cloud data and analytics program needs to be treated with the same attention to strategy, process and testing. While we’d like to believe that it’s as simple as spinning-up a new environment and moving data, doing the project correctly (and planning for future growth and success) is actually far more involved.
Understand that data management and data and analytics patterns have changed in the cloud
The cloud has brought about the separation of storage and computing. That fact alone means that the old modes of operating have fundamentally changed. As an example, the shift from extract-load-transform (ETL) to an extract-load-transform (ELT) pattern. With the cloud you have instantaneous scalability, so you don’t want to bog-down the initial load of that data. The cloud reduces your costs and load times and allows you to centralize both data and logic – but you have to understand how to do it. Knowing the rules of this “new game” are key to building your architecture for success, and operating in this space requires data strategists who know the cloud. When it comes to this skill set, not all data teams are the same.
Your cloud data and analytics platform still needs to connect to your broader ecosystem
Your cloud data platform is not going to be in a vacuum – you’re still going to have a number of systems that exist on-prem, and your cloud data platform needs to understand that the business and operations still need to function. As you take steps to move your data and analytics processes into the cloud, you have to understand what the current systems do and how to swap them out with limited to no interruption to the business itself. These internal systems need to be updated to account for the introduction of a cloud-based data analytics strategy, and the team needs to be trained on how to incorporate this new infrastructure into their existing processes. There is going to need to be change, and that change needs to be planned for, documented, and managed.
Balance process and speed when delivering data and analytics to the business
Can a data and analytics move to the cloud be fast? Absolutely. We can move quickly and get an initial analytics output, but if the initial steps weren’t well-thought-out, you’re not building something to last. Typically what we see is that a rapidly deployed cloud solution tends to run into the same challenges: the business likes what they see, and they want more data and more use cases. They’re getting value, increasing revenue, increasing efficiency, and they want more of it – at the same speed. But they can’t scale fast enough because the foundation isn’t built for that speed of expansion.
Another key drawback to speed at all costs is that oftentimes the team who is tasked with maintaining that pace is learning as they go, and as a result, mistakes are made:
- Security models get missed or aren’t defined: as you scale, privacy concerns become more of an issue as you bring in more data sources. A good security model needs to be built from the beginning. You’re giving people access to data from across many systems, and it’s much more critical that you manage the security of your platform.
- No clearly-defined data ops process: how do you control your versions, source code management, and deploy and merge changes from staging into production? If not defined, those basic systems all have to be rebuilt later, and it ends up being a far more difficult (and slow) process than if they were defined from the start.
- One person owns it: we see this all the time. A singular person is tasked with owning the cloud data and analytics architecture, and they have the keys to the kingdom. No one knows how to do what they do because nothing is systematized. When they go on vacation and something breaks, it’s catastrophic.
Remember, data and analytics in the Cloud is ‘new,’ but the approach isn’t. It’s critical to find the right mix of value to the business and value to IT. We can’t forget the lessons of the past and best practices of on-prem data processes. This is simply about translating those processes and evolving them to work in a new environment.