Achieving scalability requires the underlying application architecture to be scalable in order to fully leverage the highly scalable infrastructure services provided by AWS cloud.

Setting scalability objectives for an application will depend on many factors, such as the nature of the application, the number of users (peak and average), the growth rate, the business model (subscription based, free, paid, or freemium model), SLAs, the nature of customers (businesses or the general public), and so on. It is very important to set some scalability objectives, however, as operating on the cloud doesn't require you to be absolutely accurate about hard-to-estimate parameters. On the cloud, you can always respond quickly to your rapidly evolving business. So, while it is good to have sufficient information to guide you initially, you don't need to spend excessive time and effort trying to arrive at very accurate estimates.

At a high level, your application should respond proportionally to the increase in the resources consumed, and be operationally efficient and cost-effective. For example, executing a "lift-and-shift" strategy for migrating your on-premises applications to the cloud, followed by a traditional approach of increasing the sizes of your instances to meet increasing loads, will likely become very expensive. And your application’s increasing resource requirements will necessitate another move to even larger instances. In most cases, you will obtain the best results by splitting your application into smaller components, and then optimizing them using AWS infrastructural features and best practices. 

In all cases, an underlying objective is to always try and squeeze as much performance out of each service/component as possible before scaling and spending additional dollars.