In recent conversations with potential customers, I regularly notice a misunderstanding: many organizations find that they are already doing FinOps as soon as they use tooling to provide insight into costs. However, despite these tools, they often struggle with controlling and actually reducing their spending. While providing visibility into costs is important, it doesn’t automatically mean that an organization is managing its cloud usage optimally. After all, the main goal of FinOps is to use the possibilities of the cloud as efficiently as possible, with financial reporting being only one part. The most essential thing in FinOps is to understand the cloud usage and associated costs in order to align it with the value it brings to an enterprise. If this understanding is lacking, it is impossible to manage costs, no matter how insightful they are and how beautiful the reports are.

Tooling
The use of tooling to provide insight into cloud usage and costs is essential for FinOps. In my view, it is the starting point to be able to implement FinOps. Tooling is an enabler for FinOps. The real work within FinOps is understanding cost (structures), knowing how they can be influenced and working together to achieve an optimal result. An important part of tooling is enabling showback or chargeback; having insight into which team or application costs how much. From the FinOps point of view, this is essential so that the department that incurs these costs is also held responsible for this and can/will steer to optimize these costs. In practice, I often see that the costs are mainly used as a starting point for budgeting without looking at how they actually are these cloud costs. Teams are not challenged to control or reduce their costs. An absolute amount does not provide insight into the extent to which the costs are proportional to the value of the service purchased.
Costs within budget, still 30% too high |
---|
In this example, we did a quick scan at a customer to investigate which quick wins could be achieved. This customer had contracted a third party to optimize costs in the cloud, which also provided reports on the costs per application. The cloud spend was fairly stable and within budget. Of course, our client liked this from the point of view of predictability, but still wanted a second opinion. When we delved into the cost per application, it turned out that the production environment was well optimized from a cost perspective. However, much less attention had been paid to the non-production environment. The VMs in these environments were running 24×7, the environments had not been cleaned up, so there were many (orphaned) resources active that were no longer in use at all. In the end, we reduced costs by 50% on the non-production environments. In the end, the total costs for this application were reduced by about 30%. When only looking at costs versus budget, it was not clear that the costs for this application were far from optimal, despite the cost optimizations on the production environment. |
As mentioned above, insight into costs is essential, without this insight it is not possible to give control of the cloud use to the people who actually make / influence these costs. Given the size and volatility of the data on cloud use, a tool is therefore an indispensable tool. However, it takes more than a tool or report to actually optimize cloud usage and the associated costs.
Recommendations
In addition to cost insight, FinOps tooling also provides a lot of recommendations on how to take cost-saving measures. However, with the use of tooling, there is also the risk that this advice will be followed up immediately. Also, the insight into the costs of a subscription, department, team or application does not say anything about how (cost) effective this is. That is why it is very important to understand why the advice is given so that it can also be validated.
Resizing before purchasing reserved instances |
---|
In this real-life example, a cloud management tool indicated that our customer could save a lot by shutting down Reserved Instances. Costs could be reduced by 60%. If you look a little deeper into this advice, it turns out that these are only the compute costs. Since the machines in question were running on Windows OS, the cost would effectively only be reduced by 34%. When we zoomed in a little further on these machines, the memory and cpu usage turned out to be very low. It turned out that they were sized 1-to-1 at the size they had in a previous VMWare cluster without looking at the actual usage. We therefore proposed to convert the VM to 2-core CPUs. In case of performance problems, you can always easily scale up again (this is one of the advantages of the cloud). This would save us 69% tons compared to the current cost. Finally, this customer still had a large on-premise environment with Windows Datacenter and there was enough space to cover the machines with Azure hybrid benefit, which also saved the license component. This ultimately resulted in a 91% reduction in the cost of the VMs (from €280 to €25). If only the advice of the tool had been considered, a neat saving of 34% would have been achieved. However, by looking a little further than just the advice of the tool, a saving of 91% has now been achieved. |
VM | Pay-as-you-Go | 3 YR RI | Saving |
---|---|---|---|
D4s v4 | € 279,47 | € 183,21 | 34% |
D2s v4 | € 126,91 | € 86,48 | 69% |
D2s v4 AHB | € 64,80 | € 24,38 | 91% |
Risks
As mentioned, tooling for insight into costs and recommendations on savings is an essential part of FinOps and cloud cost management. However, it is important to have a good knowledge of the cloud cost models in order to be able to estimate the costs and advice. The tooling provides input for potentially cost-saving actions. However, these actions will first have to be properly analyzed. After all, Tooling only looks at historical consumption. For example, based on historical consumption, it can be very interesting to buy RIs where the optimal savings are on a 3-year RI. But when a workload will be terminated in the (near) future or that the architecture will be further optimized to, for example, containerization or a serverless solution, a 1-year RI or even no RI at all may be much cheaper.
In practice, we see that companies follow the advice of tools to purchase reserved instances and savings plans on VMs without thinking too much about it. In some cases, this has saved a lot of costs (rate optimization), but these costs could have been much lower if the sizing of these VMs (usage optimization) had been looked at first.
Another ‘problem’ with the reckless follow-up of the advice is that it is often not looked back whether those recommendations actually turned out to be correct. FinOps is a continuous process, cloud usage is very dynamic and changeable. You can make use of this dynamic, but you will have to pay continuous attention to it.
Continuous monitoring of actions that have been carried out |
---|
Based on the tooling, a considerable number of RIs of the same type of machine were purchased by a company. Our Quickscan showed that these were only used for 5%. After a further analysis, it turned out that 2 months after the shutdown of the RI, many machines were converted to a different, more memory-intensive VM family. As a result, on the one hand, money was paid for RIs that were not used and, on the other hand, (more expensive) machines were purchased on the basis of PAYG. Because this customer actually only looks at the advice once every 3 months, this has cost an unnecessary amount of money. |
We have seen another example with a party that had purchased prepaid Azure Databricks Units for 1 year with a substantial discount. They had already used the units after 5 months, so they started paying PAYG prices after that. They thought that the discount was valid for 1 year, while the pricing model works differently. As a result, the costs for Databricks turned out to be much higher, partly because the discount was no longer applicable. On the other hand, due to the higher usage, a higher discount could be agreed. |
Native or 3rd party tooling
In my opinion, the most important thing is that cloud users understand how to interpret the data in order to make good decisions about cloud usage. This is independent of the tool that is used. That’s why we often advise customers to start using Native tools. Many of the initial requirements regarding cost overviews, cost transparency, cost recommendations are already offered free of charge, while 3rd party software entails additional costs that are not always proportional to the added value. As customers have a better understanding of how it works, there is often a specific need (e.g. more insight into container costs, more options for budgeting, further automation, multicloud reporting) for which an additional 3rd party tool is needed/useful. It is essential to get a good idea in advance of what is expected of the tooling and what it yields (also compared to the free tools that are already available).
Points to consider when using FinOps tooling
The following things are important:
-
- Properly configure the tooling so that the cost overviews fit the structure of the organization and (financial) processes.
- Discuss the recommendations with the technical teams and functional owner to determine whether it is indeed a good recommendation and whether it fits in with future developments.
- Look beyond automated recommendations
- Monitor afterwards whether the implemented actions also (continue to) yield results.
- Implement continuous monitoring to identify anomalies and make adjustments on a continuous base.