Introduction
I recently picked up a book given to me as part of my Tailwarden onboarding welcome kit. It came with sizeable boots to fill since our CEO scribbled a note on the inside cover saying it was one of the most fascinating books he had ever read, I had to see for myself.
He wasn’t wrong, “You are about to make a terrible mistake” explores the fascinating instances of how cognitive biases distort our decision-making capabilities. Countless examples illustrate how our implicit biases affect us all, gaining awareness of their presence is the first step towards anything resembling an antidote. “Confirmation bias”, “Storytelling”, “hindsight” and “present” bias to name just a few invisible obstacles we are up against. There are so many ways our intuitions can fail us, it’s a miracle we can ever get anything done in the first place. Granted there are a multitude of biases that we could explore in this article, but in the context of cloud cost optimization and FinOps there is one bias I found to be particularly salient. A bias that if taken into account can unlock eye-watering cloud resource savings. We are talking about the Anchoring bias.
What is anchoring bias anyway?
It’s a commonly occurring bias that emerges when individuals rely too heavily on the first piece of information they receive, using it as a reference point or "anchor" for making subsequent judgments or estimations. This initial information, even if irrelevant or arbitrary, can have a powerful influence on their final decisions.
For example, let's say you are looking to buy a new laptop, and the first one you see has a very high price tag. Even if it's not the best laptop for your needs, that initial high price might become your anchor. As you continue to explore other options, you might unconsciously compare their prices to the initial high price, perceiving them as cheaper or more expensive relative to that anchor. As a result, you might end up making a purchase decision based on the anchoring bias, even though there might be more suitable and reasonably priced options available.
In the context of cloud cost optimization and FinOps, if a team sets a high initial budget for a project, they might be anchored to that figure and end up overspending on cloud resources throughout the project's lifecycle.
Recognizing anchoring bias and being mindful of its influence can help individuals and teams make more objective and data-driven decisions, ensuring that they don't get unduly swayed by initial, arbitrary, or irrelevant information.
The Pitfalls of Anchoring Bias in Cloud Optimization
Settling for modest cost reductions due to fixation on initial data
Many times in complex cloud environments you are going to encounter large and many times hard to understand bills. In most cases, a few services have an outsized weight on the overall bill, such as compute instances, databases, or complex managed services. The main cost savings opportunities will involve these services. However, here's the catch: as you delve into your cost-cutting endeavors and pinpoint the exact expenses for each service and resource, there's a risk. You might inadvertently anchor your focus on these initial cost figures. Consequently, the danger lies in settling for reductions that, while certainly noteworthy compared to the starting costs, might pale in comparison to the much larger savings potential that could realistically be achieved. In other words, you might be leaving “money on the table”. This fixation on the anchor values blinds you to the more significant optimization possibilities that await deeper exploration.
Overlooking significant savings opportunities in the long run
The risk of narrowly trying to reduce the cost of the resources that are provisioned may lead to larger overhauling and re-structuring efforts that might lead to monumental savings being overlooked. The danger is that you focus too much on the tree and miss the forest. For example, you might only ask questions like, how can I get a 10% reduction of that compute instance? and not questions like, is this the right instance type for the job, why are we not using spot or reserved instances instead? Are my compute instances scaling horizontally in an efficient manner? How long do we actually need the Cloudwatch logs to be persisted? Is serverless an option for the use case?
The impact of anchoring bias on strategic cloud resource allocation
You will find that anchoring bias greatly influences the questions you ask. And during any strategic planning effort, it's essential to frame the conversation correctly and base it on premises that can set you up to win. For many managers and divisions inside an organization, the budget allocated to the team is a proxy for how much the team is valued in the company. This leads to incentives that align with budget conservation. For those who tie their teams identity to the budget that is allocated to them results in very little variance from quarter to quarter. To be the person in the budget meeting to propose slashing the budget in half can win you foul looks from colleagues. This has to be completely thrown out of the window when in the context of cloud resource cost allocation.
Examples of Anchoring Bias in Cloud Management
Case Study 1: E-Commerce Scale-Up
Scenario: An e-commerce company experiences a surge in website traffic and subsequently sees a significant spike in its cloud expenses, particularly related to database read/write operations.
Anchoring Bias: The DevOps team fixates on the initial data that indicates database operations as the primary cost driver. They focus their efforts solely on optimizing database-related expenses.
Impact: The DevOps team successfully reduces database costs by 10% through query optimization. However, they fail to notice that a considerable portion of the expenses now comes from auto-scaling server instances required to handle the increased traffic. They also fail to ask themselves if the current database architecture is still appropriate if similar surges are to be expected.
Case Study 2: Media Streaming Platform
Scenario: A media streaming company notices a gradual increase in its cloud expenditures due to higher demand for streaming content. They identify that a specific type of storage service, used for storing user data and media files, is the costliest component.
Anchoring Bias: The operations team becomes anchored to the initial data that points towards the expensive storage service as the main issue. They dedicate their efforts to optimizing the usage and pricing of this storage service.
Impact: The operations team indeed manages to optimize the storage service, achieving a 20% reduction in costs. However, during this process, they overlook the fact that a significant portion of costs are attributed to transcoding servers responsible for converting media files into various formats. The team's fixation on the initial anchor prevents them from considering a more extensive overhaul of the transcoding architecture, which might involve adopting more efficient codecs or utilizing specialized hardware. As a result, the potential for substantial savings from a more significant transformation remains untapped, highlighting how anchoring bias can lead to suboptimal decisions.
Combating Anchoring Bias with Tailwarden
Awareness is half the battle
Awareness of the anchoring bias within us marks a crucial starting point. Effective cloud cost reduction strategies thrive on data-driven decision-making, incorporating various metrics. By expanding your analysis beyond singular resource costs, you create space for robust choices. Tailwarden's Cost Explorer feature aids in revealing historical cost trends, highlighting resource-heavy areas. Yet, it's not only about understanding high costs, it's about contextualizing their role in your architecture. Regular evaluations of alignment with budgetary needs are key. Tailwarden cloud agnostic resource management features can act as your guide, providing insights for well-informed decisions.
Collaborative decision-making to counter individual biases
One of the main takeaways from the book is that from an individuals perspective it's extremely hard if not impossible to be 100% aware and consious of ones biases at all times. It's even excruciatingly hard to be aware of ones biases at all for that matter. The only way to counteract the biases of the individual decision maker is to share the load, focus on the process itself of making decisions and engaging in dialogue, a lot of dialogue.
Olivier Sibony, the author of the aforementioned book, raises concerns about other biases surfacing within group dynamics. Topics we can explore in another article, include Groupthink, Information cascade, and Group polarization. What's evident, however, is that a group united by the goal of becoming a data-driven entity, unafraid to question past strategies and rally behind bold cost-cutting aspirations isn't just the best way to counteract biases, it's pretty much the only way.
Regularly update assumptions and strategic plans
As you use Tailwarden to gain deeper insight into the inner workings and behavior of your cloud environments, use the data points and granular cost data to inform regular conversations and working sessions with the explicit objective of challenging assumptions about how your budget is structured and how it is currently allocated. Build internal processes that help make the best decisions, use clear and concise data to inform the group conversation, encourage dissenting voices and set a deadline for a decision to be made.
Conclusion
Simply put, if anchoring bias goes unchecked it can hold you back and deny you major cost savings in the long run. The stakes are usually quite high, with ever-growing cloud environments, cloud bills quickly stack up and become the driving expense of many companies. As cloud engineers and developers, you have an outsized role in the impact on your company's bottom line. By maintaining a clear focus on the importance of unbiased cloud decision-making, and always remembering the potential that any single decision maker has to be a biased single point of failure, you are setting yourself and your team up for long-term cost savings and efficiency.