When we developed the concept of Effective Savings Rate (ESR) in 2019, the goal was to create an outcome-focused KPI to measure the ROI of rate optimization, and enable ProsperOps to quantify its value. Prior to ESR, the standard KPIs to measure rate optimization, such as discount coverage and utilization, were input-oriented. Optimizing for either of those variables in isolation could result in a suboptimal economic outcome.
We operate in a world where the rate optimization work of FinOps practitioners can be understood in economic terms, trended, and benchmarked. It was clear from the start that there was more to cloud cost management than rate optimization, but ProsperOps limited its focus to the scope of its services. This year, ProsperOps launched its first workload optimization capability. Naturally, we revisited the question “Is there a workload optimization analog of ESR?”
Here, we introduce Effective Avoidance Rate (EAR), a new KPI that complements ESR and measures the ROI of workload optimization. In part two of this blog series, we will introduce an additional overarching KPI that integrates the results from rate and workload optimization. In part three, we will dive deeper into the definitions and provide examples. There are several considerations related to each metric that will benefit from scrutiny of the FinOps community, but we are confident these concepts represent important steps forward.
Measuring Rate Optimization
ESR is computed by dividing the savings generated via discount instruments by the on-demand equivalent spend. Said differently, this equates to the price you would have paid without rate optimization, minus the price you actually paid, over the price you would have paid without rate optimization.
Using the same logic, cost avoidance due to workload optimization can also be measured.
Measuring Workload Optimization
In the public cloud, cost reduction is the measure of workload optimization success, assuming constant application performance and workflow efficiency. This cost reduction can be measured using Effective Avoidance Rate (EAR). EAR is equal to costs avoided divided by pre-workload optimization spend, or the price you would have paid without workload optimization, minus the price you actually paid, over the price you would have paid without workload optimization. This is conceptually identical to ESR.
Costs Avoided
By Costs Avoided, we mean the costs that are actually avoided via workload optimization (e.g., stopping, deleting, resizing, reconfiguring, etc.); if rate optimization would have otherwise discounted the resources that were workload optimized, Costs Avoided should be calculated post-discount. Costs Avoided is therefore defined as the cost of resources pre-workload-optimization after all discounts are applied, minus the cost of resources post-workload-optimization after all discounts are applied.
With ESR, the numerator (Savings Generated) can be measured easily since the baseline price you would have paid without discounts is known. In workload optimization, the price you would have paid without optimizing resources is more subjective and could be determined differently based on the specific action(s) taken and assumptions made. To illustrate this point, consider two scenarios.
Scenario 1
A virtual machine is powered off after business hours, and powered on the next day. Costs Avoided is calculated by subtracting the cost that was incurred while the machine was off, from the cost that would have been incurred if the resource had been on. At face value, this sounds simple. However, assume commitment-based discounts exist in the environment. It can be difficult to know exactly how these discounts would have applied if the VM stayed on.
Scenario 2
A storage disk unattached from a virtual machine is deleted. Costs Avoided is computed by subtracting the cost after the resource was deleted from the cost that would have been incurred if the resource continued to exist. Again, it is not difficult to understand the hourly cost you would have incurred if you hadn’t deleted the resource. However, a question remains about how long you should account for the hourly savings. Should the savings be realized for one month, one quarter, one year, or forever?
The challenge in measuring Costs Avoided is that you must assume what would have happened if actions were not taken, so you can compare against the actual outcome.
Conclusion
While the definition of EAR is simple, calculating it consistently requires a standard set of assumptions and methods. We invite the FinOps community to collaborate with us in defining these standards, and hope to apply this framework to different scenarios together. In subsequent posts, we will explore the relationship between ESR and EAR, and dive into the calculation of EAR in the context of specific workload optimization techniques.
ProsperOps remains focused on helping organizations prosper in the cloud. By providing a framework for quantifying workload optimization, we hope to enable FinOps practitioners to measure holistic cloud optimization outcomes. Stay tuned to learn more about how you can use EAR to understand and communicate the value of your cloud optimization strategy.
