Recent experiences with a couple of clients have led me to question what I thought I knew about data modernization.
Specifically, the widely held belief that “cloud analytics will transform our business” is fiction; however, the notion that “analytics can transform our business” I can agree with.
Analytics don’t have to be cloud-based to produce transformative results – it’s not like companies haven’t been gaining value from their on-premises analytic environments for decades. There is also no guarantee that supplying analytics capabilities using the most whiz-bang technology stack will accomplish anything other than giving your IT staff some marketable skills.
If you don’t focus on the organizational change elements that will turn your corporate culture toward actually using the analytics, you won’t see any value – let alone transformative value.
So, is it worth it to make the move? The short answer is: it depends. Let’s walk through several important factors that can help you determine whether moving your analytics to the cloud will help you reach your goals.
Essentially all cloud platforms offer some pretty amazing capabilities: elastic compute, managed services, and seemingly unlimited storage. But simply doing a lift-and-shift of your existing analytics functionality into the cloud – without rethinking your broader ecosystem – is like buying a new Ferrari to do your grocery runs.
After some thought, retrospection, and study to answer my own questions, here are the key things to consider when moving to the cloud.
It has been said that data scientists spend 90% of their time gathering data or installing and configuring software, and they spend the other 10% complaining about how they spend that 90%. Modern cloud analytics frameworks can provide instant access to pre-configured workspaces with all the tools you need; however, it comes at a cost and does require some intentional design.
I try to be as transparent as possible about most things, but being transparent about cloud costs is a must. That first bill can be shocking, even revolting when you discover you left test environments running, opted for the wrong storage tier, or didn’t scale your environment as your usage needs waxed and waned. Being transparent about the costs means teams can finally see what their analytics costs are and optimize accordingly. Years ago, as a data warehouse manager, my team regularly absorbed what I called “CanYa’s”: Can ya do this? Can ya do that? We operated as a kanban team and just understood that every sprint we were going to spend half of our usual story points filling requests that we hadn’t yet received. Some amount of this is necessary when chasing those organizational change management (OCM) goals and demonstrating the full value of analytics. You still may deal with some of this, but it can be minimized by the next factor on this list.
“Self-Service” used to mean that you provided a business intelligence (BI) tool and user training to implement it. In a well-designed cloud environment, analysts can provision their own development workspaces within limits you can (and should) specify. They can more easily deploy the machine learning (ML) models, or even the simple “what happened and why” BI reports. They can share insights without clogging your email servers with spreadsheet attachments. And they can access data without generating IT tickets. This last point leads us to the next thing to keep in mind.
Traditional security often focuses on restricting access, while cloud security can enable work by automating access based on identity. In the old way you had to initiate a back-and-forth with an IT entity specifying who you are, what you need access to, why you need that access, having them manually adding you to a security group, and then waiting for that change to propagate through the environment.
In the new world, attributes about you and the resource to which you have requested access can be compared and access granted immediately. Authorization of your upstream manager and/or the resource owner can be more easily done as necessary. Shifting from dependence on manual actions to granting access and reviewing what access was granted reduces and catches errors. You are essentially enabling compliance through automation. Secure collaboration can also be automated via these same access mechanisms.
Cloud platforms let teams share more than just data. Back in my grade school days I was sometimes dinged because I didn’t show my work: 10 times any number is obviously just that number with the decimal scooted one place to the right; showing the long multiplication was obnoxious. In the cloud, showing your work so others can scrutinize it is easy: you can easily share development environments, code, models, and potentially spreadsheet free results.
While we’ve just reviewed a handful of things to keep in mind when considering a move to the cloud, there are also several important things to avoid:
When pushing your analytics capabilities to the cloud, success isn’t just about technology. And it isn't only about cost either — it’s about enabling your teams to work differently. Focus on your analysts, your data scientists, and your business users first and let that drive your technical decisions.
The goal is not to do the same things in the cloud—it is to do things that weren’t possible before. If that is not your goal, maybe it’s not the right time to move your analytics capability to the cloud. Nothing wrong with that!