Pricing Strategy for a New Amazon S3 Storage Class: Cost Floor, Value and Cannibalization

You are a product manager on Amazon S3, the AWS object storage service, and your team is preparing to launch a new storage tier alongside the storage classes S3 already sells. Walk through how you would set its price, from the structure of the charges to the actual numbers. ```hint Floor, ceiling and neighbors Treat the price as a range: something sets its lowest sensible value and something else sets its highest. Then look at how the new tier's charges compare with the classes a customer could pick instead. ``` ### Clarifying Questions - What does the new tier offer that existing classes do not: cheaper storage for rarely read data, faster access, different durability or availability, or a different location? - Which customers is it built for, and where does their data live today: another S3 class, a competing cloud, or their own hardware such as tape libraries? - Is the main aim to win new storage, keep existing customers by lowering their costs, or improve margin? - How much freedom is there to change the price after launch, or should the launch price be one the team can hold for years? ### What a Strong Answer Covers - The billing dimensions a storage tier needs (capacity, requests, retrieval, data transfer, minimum storage duration or object size) and the cost driver behind each. - A cost floor from the tier's cost to serve and a value ceiling from the customer's next-best alternative. - How the tier fits the existing ladder of classes: the access patterns where it is the cheapest choice, and the risk of cannibalizing higher-margin classes. - Competitive reference points and the trust cost of changing cloud prices after launch. - A launch and validation plan with metrics that show whether the price is right. ### Follow-up Questions - Three months after launch, most of the tier's growth comes from customers moving data out of a pricier S3 class. Is that a failure? - Would you charge different prices in different regions, and why? - How would your approach change if an automatic tiering option could move objects into the new tier without customers choosing it?