If you missed the IMPACT Virtual Conference 2022 then this briefing is for you. As a Tech Value Creation follower, you might be surprised to learn about the trials and tribulations of putting a $ value on cost optimisation for your portcos.
Here are three insights you’ll get from this write-up:
- Why value being left on the table at the exit for sellers is an opportunity to jump-start technology value creation for the buyer.
- How one firm tuned the cloud cost equation on its head and the upside they achieved as a result
- First-hand experiences of working with some of the most extensive and most expensive cloud estates in the world
Let’s dive in.
Why this is a hard problem to solve for your portcos, especially in 2022.
- Statistics suggest 50-70% firms utilising cloud infrastructure overspend 70-90% for those firms in their own data centre.
- Most firms approach an increase in demand with step changes to capacity.
- It is not as straight forward as simply ‘cutting spend’.
- Issue with overspending is this missed opportunity to allocate money more effectively.
- For example, to grow development teams, bolstering product offerings etc.
- More effective cloud cost model developed based on the idea that enterprise firms are typically over provisioning infrastructure and services.
Data Centre versus Cloud Infrastructure
- Data centre = Scale for peak periods for capacity management
- Cloud infrastructure = scalable on demand and flexible
We’ll look at an example scenario as a launching point to drive conversations with your business leaders and colleagues in your business context.
Example Scenario: SaaS service provider with $100m annual revenue
- Typically, 20% of revenue spent on technology, including licensing, people, services and infrastructure ($20 million).
- Assuming 30% of $20 million is allocated to technology infrastructure and services ($6 million spend).
- Referencing the statistics of 50-70% overspend (cloud) and 70-90% overspend (data centre) that equates to $3m-$4.2m p.a and $4.2m-$5.4m p.a respectively.
- Further, for some companies this money would be allocated to EBITDA – and therefore a multiplier would apply to these figures and savings further maximised.
Case Studies
Case Study 1: World’s largest Online data matching Company
- Spent $20m p.a. in their data centre to $60m p.a. after migrating to the cloud.
- Company was already overspending, and this was compounded by an issue in how they sized the workload based on a calculation error.
- Challenge:
- Identifying where to cost save.
- Proposed:
- Rationalising the process that was followed with each new user.
- Outcome: Spend was reduced to $25m p.a with a growth upside of 40%.
Case Study 2: World’s largest messaging platform
- Challenge:
- High level of service incidents (200+ per week)
- Demand for instances could not be met.
- Proposed:
- Identifying patterns within the incidents to be able to target their approach.
Case Study 3: Europe’s 2nd largest finance SaaS platform
- Challenge:
- High management report load
- Siloed development teams
- Linear increase in cloud cost as growth occurred.
- Proposed:
- Analysing underlying data to identify what was being used (and why) and how users reacted to the provision of additional capacity.
Addressing cloud cost optimisation in your business requires breaking things down into:
- The technical component
- The business operations component
Technical Component
Efficiency: Measure where to invest technical time.
- Consider:
- What operations are essential to users?
- What needs to happen now or later?
Performance: How fast your platform is.
- Consider:
- Understanding how it compares to competitors.
- Explore how this maps to your user’s expectations.
Stability: What happens when more users are on your system.
- Consider:
- Does it impact speed?
- Do you have intermittent outages?
- Do things start to fail?
- Suggested approach:
- Look at your data to get an idea of the efficiency, performance and stability.
- Good place to start is outside your APM or monitoring tool.
- Can you gain access to CPU, Disk (I/O), Disk read/write and memory statistics?
- Do you have this monitoring set up already? This is priority if it doesn’t currently exist.
- Aim to access data for a 2-week period (at a minimum) and a 1-minute granularity.
Business Operations Component
- Take the technical data and map it directly to your business metrics.
Examples include:
- Orders per second
- Customers on your platform
- Concurrent users
- Once you have a handle on performance, efficiency and stability you can make changes without the need to worry about underlying technology.
- The outcome = reduced cost.
You can sign up for more ‘unconsidered needs’ below.