Friday afternoon, the CFO of a logistics company in Lampung receives an email notification from the bank: an automatic payment of Rp 87 million to the cloud provider has been processed. He frowns. Last month the bill was Rp 64 million. The month before, Rp 45 million. No new projects he approved, no major app launches. He opens the cloud console and sees a long list of services, most with mysterious code names like "i-0f3c8a2e1b9d47c1" and "snap-9e7d21c4." Not a single one can be mapped to a project or team. One thing is certain: money evaporates every month, and no one can explain why.
This story is not a rare anecdote. It is the most common experience of Indonesian companies after moving to the cloud: bills that keep rising, unclear sources, and no one accountable. According to various industry reports, cloud waste at mid-sized companies can reach 30 percent of the total bill — and for an Rp 80 million monthly bill, that means Rp 24 million evaporating every month without producing anything.
The good news: cloud costs are not destiny. They can be managed, reduced, and optimized — without sacrificing the performance that made you move to the cloud in the first place. This article is the practical guide for the Indonesian context: how to read your bill, find waste, apply proven savings strategies, and build habits that keep costs under control.
Why Cloud Costs Feel "Out of Control"
Before discussing solutions, understand the root causes. Cloud costs differ from almost every other technology cost you know.
Physical servers have clear pricing: buy once, and the value is calculable. Cloud does not. It consists of hundreds of small components, each billed per hour or per gigabyte: instances, storage, data transfer, snapshots, load balancers, static IPs, logs, licenses, and so on. There is no single "server price." There is only an aggregate bill that is hard to trace back to its source.
Second, the cloud is too easy to "switch on." One click creates a new instance; nothing needs to be signed, no budget form to fill. Employees experimenting, developers forgetting to shut down test servers, or teams duplicating environments "just in case" — all add to the bill without anyone's formal decision.
Third, cloud pricing is designed for flexibility, not predictability. Per-hour prices look small — "just Rp 20 thousand per hour" — until multiplied by 24 hours, 30 days, 12 months, and dozens of instances. Deceptively small numbers are a hallmark of cloud bills.
Fourth, in Indonesia there is an additional factor: bills are often paid automatically by credit card or in foreign currency, so cost increases are not felt until they suddenly become large. Companies paying in dollars are also exposed to exchange-rate fluctuations, which can add 10-20 percent to costs in a single year.
Step One: Reading Your Bill Correctly
Optimization cannot start from guessing. The first step is understanding what you pay for. Fortunately, the major cloud providers — AWS, Google Cloud, Azure, Alibaba Cloud — provide built-in cost management tools, and most support rupiah in their reports.
Start with three questions for every line on the bill:
- What is this? Translate service codes into human-readable names. Is it compute, storage, data transfer, or an add-on service?
- Whose is it? Which project, team, or application uses it? This can only be answered if resources are tagged from the start — which we will discuss shortly.
- Is it still needed? When was it last used, and by whom?
To make this easier, categorize your spending:
| Category | Examples | Typical share of bill |
|---|---|---|
| Compute (instances/VMs) | App servers, database servers | 40-60 percent |
| Storage | Disks, backups, snapshots, object storage | 15-30 percent |
| Network | Outbound data transfer, load balancers | 5-15 percent |
| Managed services | Managed databases, CDN, monitoring | 10-25 percent |
Note: the percentages above are common patterns, not rules. What matters is not the numbers but the way of thinking: without categories, you cannot tell what is ballooning.
If you are just starting this journey, our cloud migration guide covers how to map your systems before moving — including establishing a clear cost baseline from day one, which will help greatly when the bills start arriving.
The Biggest Waste and How to Find It
Once you can read the bill, it is time to find the waste. There are six patterns we see most often in the field.
1. Idle instances
Servers running 24 hours but used only 2 hours a day are the most common waste. Development, staging, or trial servers are often forgotten once a project ends. How to find them: look at CPU and memory metrics for each instance over 30 days. If CPU is below 5 percent almost all the time, that instance is not working — it is just burning money.
The fix: shut down non-production instances outside working hours (many companies schedule auto-shutdown at 6 PM and startup at 8 AM, saving up to 60 percent), and delete instances that are no longer used at all.
2. Overprovisioning
The habit of choosing "just to be safe" instances produces high-powered servers that are never fully used. It is like buying a truck to deliver documents. How to find it: monitor peak CPU, memory, and disk usage for a few weeks. If peak usage is only 20 percent of capacity, downsize the instance to a smaller tier. This change alone can cut compute costs in half.
3. Accumulated snapshots and backups
Backups are a safety net, but backups that are never cleaned become a burden. Old snapshots of deleted servers, or daily backups kept forever, consume expensive storage unnoticed. How to find it: list all snapshots and backups, then ask of each: still in use, or already stale? A healthy policy: keep daily backups 7-30 days, weekly 1-3 months, monthly 6-12 months — then delete the rest.
4. Wrong storage class
Cloud offers storage classes with vastly different prices: hot (fast, expensive) for frequently accessed data, cold (slow, cheap) for archives. Many companies store everything in the most expensive class because they do not know the difference. Archive data like old invoices, old logs, or rarely opened files can be moved to a cold class, saving 50-80 percent on that portion.
5. Unexpected data transfer
Outbound data transfer (egress) is the cost that surprises people most. Every byte leaving the cloud — to users, to other systems, or to other providers — is billed. High-traffic applications, large syncs, or integrations that constantly pull data can balloon egress costs. The fix: use a CDN for static content (reducing direct transfer load), and design integrations so data is processed in the right place rather than shuttled back and forth.
6. Accumulated small services
Unused static IPs, load balancers for tiny traffic, logs stored without limits, alarms sending notifications — each is small, but dozens of small services can account for 10-20 percent of a bill. Audit periodically and delete what is unused.
Proven Savings Strategies
Once waste is found, apply the following strategies. They are ordered from fastest results.
Turn off what is unused (impact: high, effort: low)
This is the first step, because it is free and immediately felt. Shut down non-production instances outside working hours or delete them. Remove stale snapshots. Release idle IPs. One working day of auditing is usually enough to find a 10-20 percent cut.
Rightsizing (impact: high, effort: medium)
Size instances based on real usage data, not estimates. Monitor for 2-4 weeks, then downsize or upsize as needed. For fluctuating workloads, consider instance types that are cheaper for workloads that are not always at capacity.
Long-term commitments (reserved/savings plans) (impact: high, effort: medium)
All major cloud providers offer large discounts for 1-3 year commitments: typically 20-40 percent below on-demand pricing. Think of it like a lease: you commit to using certain capacity and get a better price. The condition: apply it only to genuinely stable workloads (production running 24/7), and start with a 1-year commitment before deciding on 3 years. For variable workloads, a combination of on-demand + reserved is usually optimal.
Autoscaling (impact: medium, effort: medium)
Autoscaling adjusts server count to traffic automatically: scaling up when busy, down when quiet. This prevents two problems at once: paying full capacity when quiet, and losing customers when busy. For web apps with fluctuating traffic — promos, seasonal moments like Lebaran or the new school year — autoscaling is almost always worthwhile. Important: set reasonable minimum and maximum limits, and test the configuration before peak season.
Moving to the right managed services (impact: medium, effort: medium)
Managed services (managed databases, managed cache, serverless) are often cheaper from a total cost perspective: you do not pay someone to maintain servers, and scaling is automatic. For sparse or unpredictable workloads, the serverless model (pay per use, not per hour) can cut costs by up to 70 percent compared to always-on servers. Calculate total cost of ownership, not just the instance price.
Using the right storage classes (impact: medium, effort: low)
Separate hot, warm, and cold data from the start. Move cold data to cheap storage; if rarely accessed, archive it. A clear storage policy is cheaper than guessing.
Building a Cost Control System (FinOps)
One-time savings only last if there is a system to preserve them. This is where the concept of FinOps enters: managing cloud costs as an ongoing discipline, not a one-time project. Four pillars to build:
1. Tagging and ownership
Every cloud resource must be tagged: project name, team, environment (production/staging/development), and owner. Without tagging, bills cannot be attributed to anyone, and waste is never found. Make tagging a mandatory policy: resources without tags must not be created.
2. Budgets and alerts
Set monthly budgets per team or project, and install automatic alerts: warnings at 80 percent of budget, 100 percent, and 120 percent. Cloud providers offer this natively. Working alerts detect cost increases in the same week, not six months later.
3. Regular reviews
Schedule monthly cost reviews: compare this month's bill with last month's, investigate increases above 10 percent, and ask why. Make this a standing agenda item, not a panic activity. A 1-2 hour monthly review is usually enough to keep bills healthy.
4. Procurement policies
Apply simple rules: new production instances need approval, development environments must shut down outside working hours, and no one may create resources without tags. These rules prevent waste before it happens, which is far cheaper than cleaning up afterward.
Realistic Savings Estimates
How much can you actually save? Our experience and industry patterns show consistent numbers:
| Action | Potential savings |
|---|---|
| Shutting down non-production instances outside working hours | 40-60 percent of those instances' cost |
| Deleting unused resources | 10-20 percent of total bill |
| Rightsizing oversized instances | 20-50 percent of compute costs |
| Reserved instances / savings plans | 20-40 percent off on-demand pricing |
| Right storage classes | 50-80 percent of relocated storage cost |
| Autoscaling for fluctuating workloads | 30-50 percent of compute outside peak hours |
Total: a company with an Rp 50-100 million monthly bill that has never optimized can usually save 25-35 percent in the first cycle, without touching performance. For an Rp 80 million bill, that is Rp 20-28 million per month — or Rp 240-330 million per year. A number worth fighting for.
Important: optimization is not shutting everything down until systems fail. The goal is to pay exactly for capacity that is genuinely used, with sufficient safety nets. Savings that break services are the wrong kind of savings.
Pitfalls to Avoid
Chasing discounts before understanding usage
Buying reserved instances before knowing how much capacity you actually use is a way to lock in waste for 1-3 years. The right order: audit first, downsize first, then commit.
Saving in the wrong places
Cutting security, monitoring, or backup budgets to save money is an expensive decision. A single security incident or an unrecoverable database costs far more than a year of savings. Security and backup are mandatory costs; optimization must target waste, not protection.
Ignoring people costs
The cheapest cloud is the one that needs the least manual maintenance. If your team spends 10 hours a week maintaining servers that a managed service could replace, their salaries may exceed the cloud price difference. Calculate total cost: cloud + people + risk.
Stopping after one optimization
Cloud costs are not static. New apps appear, traffic changes, provider prices change. Without regular reviews, waste grows back. Optimization is a habit, not an annual event.
When to Bring in a Partner
Cloud optimization can be done in-house, but many companies choose to involve a partner — especially in the first cycle, when the service map and tagging do not yet exist. Signs you need help: no one can explain the contents of the bill, waste has been running for months, or the internal team is too busy with other projects for cost reviews to happen. A good partner does not just cut the bill; they build the system — tagging, alerts, reviews — so costs stay controlled after they leave. This follows the same pattern as when you need an IT consultant in general: not when everything is broken, but when costs are silently eating away.
When choosing a partner, make sure they start with a usage audit, not with selling services. Promised savings must be traceable to specific bill lines, and the control system must keep running after the project ends.
A Monthly Checklist You Can Start This Week
- Open last month's cloud cost report, export to a spreadsheet.
- Group spending by service: compute, storage, network, managed services.
- Flag lines that cannot be identified — these are investigation priorities.
- Check for instances with CPU below 5 percent over 30 days.
- List snapshots and backups; delete stale ones.
- Set budget alerts at 80 percent and 100 percent.
- Establish tagging rules for all new resources.
- Record decisions: what was shut down, downsized, or scheduled.
Run this checklist once, and you will already be saving more than most companies that have "used the cloud for years."
Closing: Cheap Cloud Is Understood Cloud
Cloud costs are not the enemy. They are a mirror: a ballooning bill shows a system that is not understood, ownership that is unclear, and decisions that are undocumented. Conversely, a controlled bill shows healthy discipline. The good news is that discipline can be built — starting with honestly reading the bill, cutting waste, and maintaining a review habit.
If your cloud bill is starting to feel like a puzzle, or you want to make sure your infrastructure runs efficiently from the start, the Kartech team in Bandar Lampung can help: from cloud cost audits and rightsizing to building the FinOps system that keeps bills under control month after month. Reach us through the contact page or explore our services to see how we work.
