Avoiding the Cloud Cost Apocalypse: A Sustainable Approach to Public Cloud Financial Governance

A typical public cloud journey sees an enterprise experiment with cloud resources and services and set the initial use cases into production. Once the first value has been generated by these initial use cases and the enterprise proceeds to scaling up the transformation efforts, it often experiences a sharp rise in its cloud costs, prompting management to seek explanations.

At this stage, financial teams find themselves overwhelmed by new cost management challenges in public cloud scenarios, such as:

• On-demand availability of resources, defying a traditional understanding of cost sources
• New cost drivers and pricing models (hourly charges, per-use API charges, etc.) using one cloud provider and significantly worsening in multi-cloud setups
• Overwhelming amounts of available granular cost data (in contrast to large overhead charges with estimated attribution keys in an on-premises world)

We’ve observed that cloud economics are a focus point in estimating investment requirements upfront while planning and kicking off the cloud transformation. However, the primary attention moves to other focus points once the program scales up, such as committed upfront consumption targets with vendors, a lack experience in cloud architectures, or tight project timelines. This leaves financial governance as an afterthought—until costs escalate, as outlined in the typical journey above.

We believe that avoiding this pitfall and ensuring a sustainable, value-generating public cloud transformation requires early implementation of continuous and embedded financial governance.

Mastering emerging cloud cost challenges through dedicated public cloud financial governance

Together with clients from the insurance industry, we have developed an end-to-end view on financial governance in public cloud setups.

Embedded in the cloud governance organization, the cloud financial governance function operates in three core dimensions, each linked with a set of activities that contribute to a controlled and cost-conscious culture.

1. Cost visibility and accountability

Public cloud charging and pricing structures enable a more granular level of available cost data compared to traditional on-premise or private cloud settings. By implementing rigorous cost ontology structures with clear accountabilities and owners, enterprises can apply real unit-based modeling approaches (e.g., per net premium) with automated metering and chargeback capabilities. Common challenges of financial IT reporting, such as cost data collection, application- or project-level reporting, or long reporting cycles, are thus addressed by this dimension. Furthermore, it helps ease the transition to an as-a-service IT model (from capital expenses to pure operating expenses).

The main activities include implementing an unambiguous (billing) account structure and cloud resource tagging strategy. The use of separate accounts and client-specific resource tags enables exact tracing and, as a result, chargeback of costs to consumers. Setting up automated reporting infrastructure and processes supports the continuous monitoring of cost efficiency and value-based cloud KPIs.

2. Continuous cost optimization

The “one click away” availability of resources inevitably leads to cloud waste. Analysts estimate that up to 35% of all cloud spending is waste (source: RightScale, 2019 State of the Cloud Report). While we’ve experienced different levels of actual cloud waste in our client work, a small reduction in waste can already have a significant impact on costs. Common sources of cloud waste include:

• Unused environments, infrastructure, or storage capacity (e.g., DEV environments running unused overnight)
• Oversized virtual machines, due to a “better too much than just enough” mentality originating from on-premise infrastructure commissioning
• Failure to leverage discounts for reserved instances for critical workloads, even though a steady need for cloud resources can be reliably predicted
• Inexperience with cloud-native architectures, leading to unnecessary overhead (e.g., high-frequency API calls)

Continuous cost optimization aims to detect and counterbalance cloud waste.

The main activities include detecting cloud waste (in real time) and reducing it by implementing several cloud cost optimization strategies with different levels of implementation complexity and savings impact.

In our experience, the most effective strategy in the short term is to eliminate idle resources or leverage intelligent tiering systems of cloud providers while matching committable capacity demand with discounted reserved instances (“saving plans”/“credits”/“budgets”). Intelligent tiering refers to the automatic sizing of cloud resources by providers based on usage patterns. In the long term, sensitivity to right-sizing virtual machines and cost-aware, cloud-native architectures will have a more significant impact.

3. Cost planning and forecasting

Budget setting and IT spending forecasts are well-established practices in enterprises. However, when it comes to public cloud environments, the task becomes much more challenging. Vast increases in the number of trackable resources and the inherent change in cloud infrastructure are two examples of novel complexity drivers. Cost planning and forecasting aims to proactively control cloud cost developments and make the task more manageable through intelligent automation.

Modelling the expected consumption levels is one core activity in this dimension. While the accuracy of predictions will be hit-and-miss at the beginning of the cloud journey, it will increase over time due to experience gained through iterative cost estimation for different workloads. Cloud providers also offer a variety of services to help enterprises with these initial estimations (e.g., AWS Cost Explorer). Based on the predicted consumption, automated protocols are set, such as spending limits and linked overdraft alerts as well as enforceable spending policies. This leads to cost owners and projects re-evaluating budgets with management in case of overspending, contributing to cost-aware service design while providing safeguards against costs overtaking cloud goals.

The bottom line: Practical takeaways

Throughout our project work with diverse sets of clients, we have compiled a set of practical takeaways to build on when implementing financial governance for public cloud environments: 

1. Cost visibility and accountability

a. Enable single cost item experts on day one: Collecting and storing your cost data centrally allows for reporting or debugging at any point in time and provides valuable input for forecasting.

b. Define a global cloud resource tagging/cost ontology: Use of project structures to establish boundaries between workloads, users, and costs will replace traditional “application thinking,” while tagging will help to report on cross-project costs (e.g., shared database usage). Always connect cloud usage data to accountable users.

c. Prefer active cost reporting over retrospective invoice inspection: Use of the right resource tagging and project strategies to track costs in real time will yield more insights on true cost drivers than just reviewing invoices.

2. Continuous cost optimization

a. Manage your cloud costs in a continuous effort: Rather than setting fixed review intervals, implement automated reporting structures and rules along with a permanent cloud financial operations team to help continuously monitor costs and take immediate action.

b. Keep track of unattributed spending: Set up focused reporting on unattributed cloud spending (e.g., resources that aren’t associated with projects and aren’t tagged) to identify abandoned, idle, or unintentionally commissioned resources.

c. Reserved instances (“saving plans”/“credits”/“budgets”) are a game changer: Investing in reserved instances yields lucrative discounts based on time and payment committed upfront. Start by purchasing resource family plans for key workloads, analyzing usage patterns, and iteratively adjusting the reserved instance volumes to mitigate the risk of over commitment.

3. Cost planning and forecasting

a. Automate everything: Set up automated policy enforcement to help plan costs and avoid overspending for no good reason. Many cloud providers offer policies for different resource types: For compute and storage, automated availability schedules, auto-scaling capabilities, and utilization alerts limit waste, while for governance budgets, policies and alerts ensure that organizational, owner, or project accounts are compliant with enterprise guidelines.

b. Accept iterative refinement of cost planning and forecasting: Even though cloud costs can be reported at a granular level once they have been incurred (using the right visibility and accountability practices), planning and forecasting in public cloud environments is closer to agile estimation methods due to the dynamic nature of resources and the encouragement to experiment. Therefore, enterprises that accept the initial planning inaccuracy and focus their efforts on enabling technical and governance teams to learn together rather than play the blame game will be more proficient and accurate in subsequent iteration cycles. 

c. Give accountable stakeholders access to cost data beyond reports: Rather than maintaining an information monopoly on cost data, finance teams should openly share granular cost data with resource owners to enhance planning forecasts. Next to educational effects, enterprises can create gamification systems to incentivize cost-conscious mindsets for public clouds.

If you recognize the challenges relating to cloud transformations we’ve outlined in this article, please do not hesitate to contact us. We would be happy to share our experiences with you and discuss your cloud transformation journey.

About the authors

Dr. Simon Kirstein (Associate Director) and Nico Müller (Senior IT Architect) are part of the Insurance practice at BCG Platinion and are experts in IT transformations and cloud computing. The experiences shared in this article were compiled from a diverse set of insurance client engagements.