Cloud platforms all have their own advantages, and the right cloud deployment strategy can meld those benefits into a unified platform. But IT may not be able to define which option is best for new deployments because, in many cases, the choice is not clear. It’s important to get a real analysis of usage and cost across enterprise cloud environments in order to maximize usage and keep costs under control.

It may be possible to move every enterprise app to a cloud environment and nearly eliminate on-premises systems, but like every other groundbreaking technology, there are limits to what should be done versus what can be done. The second decision IT leaders need to make after first establishing that they will pursue a cloud-based strategy is to evaluate what systems and applications to migrate to different types of cloud environments.

Where in the cloud?

 Your cloud deployment strategy is likely to include a variety of platforms, including local storage, private cloud systems, and public clouds to make up a hybrid cloud. Each cloud system offers different advantages along with their own cost structures. Putting your data and apps in the right cloud will make the difference not only in how responsive the resulting system is but also in your ROI. Here are the main issues to consider as you assemble your cloud deployment strategy.

Specific featuresEach cloud environment has its own set of features, making it imperative to understand resource requirements for the applications you deploy to them. The first step in making this decision is to understand what, if any, special features are required by the app you intend to host. For example, the ability to have full control over app performance tuning and the security of its data may be reasons to consider local or private cloud infrastructure rather than a public solution.

Performance and service restrictions—Cloud services typically have financial advantages when it comes to capital expenditures, shifting the cost to ongoing expenses based on usage. This change can require an adjustment from an infrastructure in which the cost of performance has been based on the capital cost of the number of CPUs and storage the company owns. Cloud providers charge incrementally for the use of resources, and each application uses a unique combination of resources based on the program code and how the app is used. That means the calculation may not be fully understood until after the instance has been up and running in real life conditions for some time.

Initial and ongoing costs—Enterprises can achieve savings beyond the reduction of capital outlay by shifting ongoing maintenance costs to the cloud provider. Those costs can be predicted by carefully estimating what resources will be used based on historical data. But IT should perform ongoing monitoring and analysis to assure budgets are within their initial estimation and be prepared to make changes when warranted.

SLA / provider skill sets—Valid reasons for pursuing a cloud deployment strategy include taking advantage of capabilities and skills not readily available within the enterprise. New services like bots, analytics, and machine learning require specific skill sets and some cloud providers can include both initial and ongoing services as part of a cloud deployment agreement. CIOs should evaluate the extent and duration of their dependency on external services and weigh their costs against developing some or all of them within the company.

Regional deployment—Global networking has reduced the importance of the location of data and computing power, but where data is stored can still have a significant impact on a company because of legal and regulatory compliance issues. IT needs to understand related features offered by cloud service providers and how they can be managed as mandated by shifting regulations around the world.

Configuration flexibility—Comparing two or more service providers requires careful evaluation and estimation of the applications to be deployed and the conditions under which they will be used. Even after diligent study and selection of configuration options, actual costs are likely to be unknown until after deployment and some period of usage.

After initial decisions are made and the first set of applications are deployed on cloud resources, IT must continuously monitor and evaluate performance against initial expectations. Companies have the option to employ a cloud deployment strategy that calls on multiple cloud service providers with unique strengths, but maintaining visibility and control over a diverse environment can be difficult. IT leaders need to add the right tools to their strategy to simplify continuing evaluation and decision making.

In short, there is likely no single cloud deployment strategy that will deliver the best ROI. It’s incumbent upon IT leadership to evaluate their options to understand how they will be supported best by various cloud technologies. Testing, experimentation, and analysis will provide the answers that guide a final mix of solutions that does deliver the best ROI, and that combination is likely to be different for every enterprise.

Scott Koegler has more than 20 years experience as a technology journalist and has written for publications including Network Computing, Forbes, Internet Evolution, and many others. He also practiced IT as a CIO for 15 years.

»Read next: IT isn't looking for an all or nothing approach to cloud, an interview with Red Hat CIO, Mike Kelly