When to Stay On Premise and When to Go Into the Cloud

As head of customer co-innovation for cloud solutions at SAP, I have the privilege of brainstorming with some of the leading corporations and talents of the world regarding next-generation applications.  Before joining the cloud organization, I was fortunate to hold a similar role for on-premise solutions.  In both cases, I found myself advising customers on the same burgeoning technologies albeit with much different frames of reference. Innovating with on-premise solutions is the classical example of building a plane in flight.  Innovating in the cloud, on the other hand, is like building a new spaceport.  Each discussion is understandably quite different.  In this blog, I’d like to explore the issue of technology without the forgone conclusion of execution.  Perhaps knowing the realm of the possible will help customers better decide which path to take.

Let’s begin with analytics.  In the on-premise world, we’ve made tremendous progress extracting data from production databases for slicing and dicing in data warehouses.  One of the shortcomings with this approach, however, is that you end up working with two separate applications: the transactional ERP system and the analytical business intelligence (BI) solution.  Furthermore, the data warehouse rarely has up-to-the-minute information.  Solutions being developed for the cloud not only have analytics built-in, they actually provide insights at key decision points to support real-time transactions.  In other words, rather than forcing the user to leave the application and perform a query on old data, cloud solutions anticipate queries and push the results based on up-to-the-minute information.  When you look up an account in a sales solution, for example, a graph is pushed to you illustrating the current sales forecast of the account – best tailored as your view of the situation.

Innovations in database technology such as in memory take the analytics discussion to a new level.  Both on-premise and cloud solutions are beginning to leverage these technologies now that the price of solid-state memory makes it feasible.  This has been leading to dramatic decreases in the time needed to execute a given query.  The killer application for this technology, however, is not getting the answer to the same question faster but rather asking more insightful questions that hitherto were impossible to answer in a relevant time period.

In other words, as you develop new cloud solutions based on in-memory databases, you have a radically different frame of reference as to what is a reasonable query to make on the fly.  Take traffic optimization in Japan as an example.  Taxis with beacons provide real-time data on traffic conditions so that a central dispatch can redirect traffic to less congested routes.  Before applying the quantum leap in speed of the in-memory database, there was simply too much lag time for the concept to work properly.  By the time data could be analyzed on say a broken down vehicle and alternate routes communicated, the vehicle might be on its way again.  As a result of the delayed optimization, the main route might be under utilized and the alternate routes overburdened.

The near ubiquitous access to networked computing via smart phones represents yet another ballooning opportunity for business applications – mobility.  In areas such as manager approvals, for example, applications are being developed for smart phones that enable managers to transact virtually anywhere, anytime.  In the case of on-premise back-end systems, the underlying application logic is mostly designed with the desktop interface in mind.  The challenge, therefore, lies in covering the same basic transactional steps on much less real estate.  In the case of cloud applications, you take the opposite approach.  You begin by designing the process for small smart phone or tablet screens knowing that there will be far fewer challenges later adapting the user experience for traditional desktop computers.

New avenues of collaboration are changing the vary way in which we work. Consider, for example, how email and online calendars have impacted the workplace. On-premise applications have evolved to take advantage of these new communication channels.  A typical use case consists of generating an automatic email to communicate an alert.  If, for example, the assembly line stops, the production managers receive an alert via email and/or SMS.  In the case of cloud solutions, the opportunity exists to develop new cutting-edge forms of collaboration that are baked into the solution – and tailored to the persons role.  Feeds are a good example of a revolutionary form of collaboration with far-reaching potential.  This new communication paradigm – we call it people centric – allows sales reps, for example, to interact with all account-related activity and participants in real time.  In other words, rather than waiting for the customer to complain about a late delivery, sales reps receive an automatic feed alerting them to the challenge.  This allows them to proactively reach out to the customer making them aware not only of the potentially late delivery but also of all the actions being taken to rectify the situation.

Collaboration and mobility are examples of a broader trend toward the democratization of information technology.  As more and more people gain both access to and reliance on software-based solutions, the focus of application development is shifting from the process to the person.  In other words, on-premise applications were largely developed at a time when a limited amount of power users were trained to ensure process integrity.  Today, the expectation is that everyone will use the enterprise cloud applications without training thereby ensuring process integrity.  On-premise applications can refresh user interfaces with newer technology such as copy and paste, drag and drop etc.  Nonetheless, it is cost-prohibitive to change the underlying application logic.  If the process was established to populate a series of fields in a database, the end user is going to feel like a data entry clerk.  To ensure a broad cross section of the company uses an application without training and complaining, a more user-centric approach is necessary.  Digital natives, in particular, are happy to rely on technology but only if it meets their high standards.  They expect solutions to be intuitive, helpful and efficient.  Therefore, cloud solutions devote a significant portion of development resources to supporting the needs of the end user.  Design thinking, for example, helps cloud developers gain empathy. Storyboards are used to not only convey the process that needs to be completed but also the end user challenges that need to be mitigated. Today solutions need to be desirable.

Lastly, there is the issue of speed of innovation.  On-premise upgrades have traditionally been so expensive and disruptive that companies tend to limit them to one every two to five years.  In other words, even if an on-premise application is able to develop new functionality to address one or more of the preceding trends, it may take a further two to five years until the customer can actually reap the benefits. A lot of efforts go into gaining speed here, for example to enable customers to consume enhancements in smaller chunks. But contrast this with cloud solutions where innovations can typically be consumed immediately. The user-centric approach creates intuitive functions that require close to no training.  The service delivery model means the software vendor does the actual upgrade once for all customers.  Add to rapid consumption, the rapid development that comes from lean development mythologies such as agile and the innovation cycle from initial idea to productive use may take as little as a few months.  In the on-premise world, the upgrade alone may consume a few months.

Despite the many advantages of cloud solutions, on-premise applications are likely to continue to perform the lion’s share of enterprise transactions for many years to come.  Typically, the main costs associated with these applications are already sunk.  If modern trends haven’t eroded their effectiveness, legacy on-premise applications can be robust, cost-effective workhorses.  Furthermore, the life of on=premise solutions can be extended with cloud extensions in a hybrid deployment.  Cloud solutions are most compelling when new investments need to be made either because the former solution no longer meets the needs of the business or a new business need arises. Regardless of which development paradigm makes the most sense, one thing is clear:  New technology trends make this an exciting time to co-innovate!