Is Analytics as a Service the Right Approach for You?

It seems everything is available as a cloud hosted service today, creating a tremendous amount of noise. How can you be sure what’s right for your organization?

There is a lot of confusion out there when it comes to the “as a service” market—it’s not hard to find analysts and publications talking about the problems of “confusion as a service,” and the notion is not new. It seems virtually every business need can be met by a cloud service which promises to save you money.

The key to cutting through the noise is to first understand exactly what kind of service you need. We’ve talked about the need for flexible software architecture that is cognitive-first, allowing you to use the technology you need when you need it. Analytics are a critical component of any organization’s technology stack, so let’s now drill down to the analytics-as-a-service market.

Where it all Began

Organizations have been trying to get out of the data center business by going to the cloud for services for a long time. It started decades ago with outsourcing. First the environment, then IT and then application development were all outsourced to private service providers. Next came Salesforce, now a household name, offering SaaS (Software as a Service), followed quickly by providers offering IaaS (Infrastructure as a Service). These were helpful innovations—for example, instead of relying on a service provider to manage your environment, IaaS used virtualization, multi-tenancy and other new technologies to share infrastructure and cut costs.

It didn’t end there though. IaaS and SaaS providers moved up and down the stack to offer a PaaS (platform as a service), providing an application development platform in the cloud. An avalanche of services built from there as demand increased. Seemingly everything became available as a service, including analytics. The definition of a “service” began to get murky.

What Exactly is a Service?

There are many different types of services, so the term can mean different things depending on the implementation. SaaS users will interact with an application, IaaS users with a management interface, PaaS users with a platform and development tools. On the other end of the spectrum, one might interact with the API that the service uses to allow integration, such as OData for databases, REST for the web, or a private API.

Without one single definition to rely on, the best approach to reviewing a cloud service starts with knowing your exact requirements.

Do You Really Need It?

Though it’s marketed as simple, aspects of the as-a-service market are complex. If you don’t know exactly what you need from your Analytics, it’s easy to pick a service that is wrong for your organization.

For example, do you need a platform that will manage all your analytics infrastructure? Or would a smaller pre-packaged service suffice, such as one that can analyze something specific like a set of images? Chances are you may need both. In both cases you will leverage Machine Learning to accomplish your goal, but making predictions from a static set of data (like those images) requires an entirely different scenario than dynamic time-series data – think fraud detection, healthcare diagnoses, or predictive maintenance.

For packaged services, you provide the data in a pre-defined pattern that has already been modeled… in other words, all of the heavy lifting involved with Machine Learning has been done and has been baked into the service.

For the complex scenarios, it will take a lot of manual work by data scientists to properly get, prepare and label data sets, and more still to identify what combination of analytical models will provide the most accurate result. While you may be able to use a cloud service to support this work, it’s much closer to a hosted platform than a packaged service.

This type of service offering can provide significant benefit, but don’t overlook what is required to make your analytics team successful. There is a huge difference between outsourcing the infrastructure vs. selecting an option that will automate the data science lifecycle. Selecting the wrong option could leave you stuck in a way you didn’t expect.

Ask Yourself Why

The as-a-service market is broad and messy, and the word analytics is hardly less so. An “analytics service” could be any number of things, and they are all for sale. Whether you need a small specific service or a complex one based on time-series data, or whether you need a pure infrastructure platform or one that will alleviate the complexity for your data scientists (like DataRPM), the key is to start by knowing exactly what you mean to solve. Once you have that, as long as you’ve started with a flexible and cognitive-friendly software infrastructure, it will be much easier to find and implement the solution your organization needs.