Jun 14, 2026

16 min read

AWS Cost Management: A Complete Map of Services to Track, Control, and Optimize Spend

End of the month, the AWS bill arrives. The number is 40% higher than expected. You open the console and face a barrage of questions:

“Where did the money go? Which service, which environment, which team spent the most?”
“Why did it suddenly spike this month? Did someone leave an EC2 cluster running all weekend?”
“How do I get a heads-up next time before I blow past the budget, instead of finding out at month-end?”
“And most importantly: how do I pay less for the same amount of work?”

Here’s the interesting part: cost in the cloud is not a fixed number you receive at month-end — it’s a variable you have to actively manage. The pay-as-you-go model is great for flexibility, but it also means a single misconfiguration can quietly burn money until the bill lands.

To solve this, AWS doesn’t give you one tool — it gives you a whole suite of services, each tackling a different stage of the cost management lifecycle. And this is the SAA exam’s favorite trap: it presents a scenario (“get an alert when cost exceeds X”, “the most detailed report pushed into a data warehouse”, “automatically detect abnormal spend”) and asks which service to use — while several of the other options all “sound right” too.

This article is the map that helps you draw clear lines between those services. For each one, we’ll cover: what it does, its core features, real-world use cases, and the points that tend to come up (free or paid, level of detail, data latency).

Note: This is an overview meant to build a mental model and help you recognize services quickly in the exam room. Each service here could be a deep-dive of its own; this post focuses on the boundaries between them and why each exists.

1. The big picture: 5 stages of the cost lifecycle

Before the details, pin down this framework. Cost management on AWS goes through five stages, each answering a different question with its own set of tools:

Stage	The question it answers	Primary services
Estimate (before)	“How much will this cost before I build it?”	Pricing Calculator
Understand & analyze	”Where did the money go?”	Cost Explorer, Cost & Usage Report, Billing Console
Alert & control	”Tell me when I exceed a limit / something’s off”	AWS Budgets, Cost Anomaly Detection, CloudWatch Billing Alarm
Organize & allocate	”Split cost by team/environment/project”	Cost Allocation Tags, Cost Categories, Consolidated Billing, Billing Conductor
Optimize	”How do I pay less?”	Savings Plans, Reserved Instances, Spot, Trusted Advisor, Compute Optimizer, Cost Optimization Hub

Keep this framework in mind; now let’s walk through each stage.

2. Understand & analyze cost (Visibility)

This is the group that answers “where did the money go”. All three tools below look at cost that has already been incurred, but at three different levels of detail and for three different audiences.

2.1. AWS Cost Explorer — the visual analysis dashboard

AWS Cost Explorer is the visualization tool you use to view, filter, and analyze cost and usage over time. It’s the first place to open when you want to answer “what cost the most this month”.

Core capabilities:

Filter & group by: view cost by service, by region, by linked account, by charge type, or by cost allocation tag. For example: “EC2 cost in us-east-1 for the production account, grouped by the Environment tag”.
Granularity: by month, day, or hourly/resource-level (hourly and per-resource granularity must be enabled separately, carries an extra fee, and is limited to a 14-day window).
Forecast: uses history to predict upcoming cost (up to 12 months ahead).
Built-in optimization reports: Cost Explorer also ships with reports on the utilization and coverage of Reserved Instances and Savings Plans, along with purchase recommendations and rightsizing suggestions (adjusting EC2 sizes to fit actual demand).

Note:

Historical data reaches back up to 13 months.
The console UI is free; if you call the Cost Explorer API, it costs $0.01 per request.
The data is not real-time: it’s typically delayed by around 24 hours.

Think of Cost Explorer as an “interactive dashboard” for a human to sit and explore cost — pretty and fast, but not the place to run SQL or feed your own BI system. When you need that, you move to CUR.

2.2. AWS Cost and Usage Report (CUR) — the most detailed data

Although they draw on the same underlying data, Cost Explorer is for turning that data into charts. The Cost and Usage Report gives you a far more granular view.

AWS Cost and Usage Report (CUR) is the most detailed and comprehensive cost-and-usage dataset AWS provides. Where Cost Explorer is charts to look at, CUR is raw data for you to process yourself.

Characteristics:

Each row (line item) describes usage at a very fine grain — down to the hour, down to the resource — with full tag, pricing, credit, and RI/Savings Plans information.
CUR is delivered to an Amazon S3 bucket you specify, in CSV or Parquet format, and is updated up to several times a day.
From S3, you wire CUR into the analytics ecosystem: query it with Amazon Athena, load it into Amazon Redshift (a data warehouse), or build dashboards with Amazon QuickSight.
CUR 2.0 is the newer dataset (delivered via AWS Data Exports), with a more consistent schema that’s easier to work with for analysis.

When do you use CUR instead of Cost Explorer? When you need deep, highly customized, or automated analysis — for example, a finance team that wants to build its own chargeback reports, or you want to join cost with internal data. The mnemonic: if the question says “the most detailed cost report, pushed into S3, queryable with Athena” → the answer is almost always CUR.

2.3. Billing Console — this month’s bill

The AWS Billing Console (the Bills page) is where you view the current bill: how much has been incurred month-to-date, broken down by service and by account. It’s simple — it doesn’t have Cost Explorer’s deep analysis — but it’s the fastest place to glance at “how far along is this month’s spend”. It’s also where you enable options like receiving cost alerts and activating cost allocation tags.

3. Alert & control (Monitoring & Alerting)

Analysis requires you to actively open something and look. This group is the opposite: the system actively notifies you when something noteworthy happens — and in some cases even takes action automatically to stop the bleeding.

3.1. AWS Budgets — set a budget and get alerts

AWS Budgets lets you set a budget threshold and get alerted when actual (or forecasted) cost or usage hits it. It’s the standard “defensive” tool to avoid end-of-month surprises.

There are four budget types:

Cost budget — a limit by dollar amount (e.g. “no more than $1,000/month”).
Usage budget — a limit by amount of usage (e.g. “no more than 1,000 EC2 hours”).
RI utilization/coverage budget — tracks the utilization and coverage of Reserved Instances.
Savings Plans utilization/coverage budget — the same, for Savings Plans.

When a threshold is hit, Budgets sends an alert via email or via an Amazon SNS topic (and from there fan out to Slack, Lambda, etc.).

Its signature strength is Budget Actions: instead of just notifying, Budgets can react automatically when a threshold is crossed — for example attaching a restrictive IAM policy / SCP (blocking new resource creation), or stopping running EC2/RDS instances. This is the mechanism to actually stop overspending, not just learn about it.

Budgets data is refreshed up to several times a day (not real-time).

3.2. AWS Cost Anomaly Detection — ML-based anomaly detection

Budgets requires you to know the threshold to set in advance. But what if cost spikes somewhere you didn’t anticipate? That’s the job of AWS Cost Anomaly Detection.

This service uses machine learning to learn your normal spending patterns, then automatically detects abnormal increases (anomalies) — complete with an estimate of the impact and the likely root cause. You create cost monitors to scope what’s being watched, by:

All AWS services, or
Each linked account, or
Cost category, or
Cost allocation tag.

Alerts are delivered via SNS or email, either as individual anomalies in real time or as a daily/weekly summary. Important for the exam: Cost Anomaly Detection is free.

The core difference from Budgets: Budgets catches a fixed threshold you define; Anomaly Detection catches deviation from a pattern that ML learns on its own — you don’t set any number. The scenario “detect abnormal spend without setting a manual threshold” → the answer is Anomaly Detection.

3.3. CloudWatch Billing Alarm — the old-school billing alert

Before Budgets existed, the classic way to get a cost alert was to create a CloudWatch alarm on the EstimatedCharges metric. The mechanism: enable “Receive Billing Alerts” in Billing preferences, then create an alarm that fires SNS when estimated charges exceed a number.

Two things you must remember:

The billing metric only exists in us-east-1 (N. Virginia) — you must create the alarm there even if your resources live in other regions.
It only does one simple thing: alert when estimated charges exceed a fixed threshold.

These days AWS Budgets is far more flexible (multiple budget types, usage/RI/SP tracking, Budget Actions) and is usually preferred. But the exam still brings up the CloudWatch billing alarm as an option — and the “us-east-1 region” trap shows up a lot.

4. Organize & allocate cost (Organize & Allocate)

As an organization grows with many teams, environments, and accounts, the question stops being “how much” and becomes “which team / which project is responsible for which charge”. This group lets you slice the bill along whatever dimension you want.

4.1. Cost Allocation Tags — label resources to trace cost

Cost Allocation Tags is the mechanism that uses tags (key–value pairs attached to resources) to break cost down along dimensions you care about. Attach Team=payments or Environment=prod to your resources, and you can then group cost by those tags in Cost Explorer, Budgets, and CUR.

There are two types:

AWS-generated tags — generated by AWS, with the aws: prefix (e.g. aws:createdBy). You can’t edit these.
User-defined tags — created by you, with the user: prefix.

Three points that commonly trip people up:

A tag existing on a resource is one thing, but to use that tag in cost reports you must activate it in the Billing console.
After activation, the data can take up to 24 hours to become available.
Activation only applies going forward — it’s not retroactive for past cost.

4.2. AWS Cost Categories — group cost into logical buckets

Tags are very flexible but also easy to make a mess of — inconsistent spellings, inconsistent across accounts. AWS Cost Categories sits one level above: you define rules to group cost into buckets that make sense to the business — for example a Department category with values Engineering, Marketing, Data.

Rules can be based on multiple dimensions: by account, by service, by tag, by charge type, and so on. Once defined, the category is usable throughout Cost Explorer, Budgets, and CUR. Think of Cost Categories as a “mapping layer” that turns a messy technical structure into a tidy financial one.

This service is free.

4.3. Consolidated Billing (AWS Organizations) — merge bills across accounts

With multiple AWS accounts, you should bring them under an AWS Organizations and enable Consolidated Billing. Then all member accounts share a single bill, paid by one management (payer) account.

The benefit isn’t just tidiness — it’s saving money:

Aggregate usage to reach cheaper pricing tiers (volume discount): many services have pricing that drops as usage grows (S3, for example). When you pool the usage of all accounts together, the combined total is larger and falls into a cheaper tier — like buying in bulk.
Share Reserved Instances and Savings Plans benefits: an RI/SP bought in one account, if that account doesn’t fully use it, automatically applies its leftover to other accounts in the organization (this sharing can be toggled on/off).

Consolidated Billing is free. This is one reason multi-account architectures almost always come with AWS Organizations.

4.4. AWS Billing Conductor — custom billing for resellers/MSPs

AWS Billing Conductor is for a specialized situation: you’re an MSP, a reseller, or an enterprise that needs to recompute/redistribute cost at custom rates for each business unit. You create billing groups and apply your own pricing rules (markup, discount, and so on) to produce pro-forma billing data that reflects exactly how you want to charge back to customers/departments, instead of the raw AWS rate. It’s a niche service; for SAA you just need to recognize “needs custom rates / chargeback for a reseller’s customers” → Billing Conductor.

5. Estimate cost before deploying (Estimate)

5.1. AWS Pricing Calculator

Everything above is after you’ve spent money. AWS Pricing Calculator is the only tool that looks forward: it estimates the cost of an architecture before you build it. You pick services and configurations (instance type, capacity, region, traffic, etc.) and it produces a monthly estimate.

Characteristics: free, no AWS account login required, lets you save and share estimates and group multiple services together. Use it to compare architecture options, build project budgets, or make the case to your boss for a budget. The scenario “estimate cost before deploying” → always Pricing Calculator.

6. Optimize cost (Optimize)

The final stage: pay less for the same workload. There are two directions: choose the right pricing model, and eliminate waste based on recommendations.

6.1. Pricing models: On-Demand, Reserved Instances, Savings Plans, Spot

This is the most fundamental knowledge about compute cost optimization, and it shows up densely in the SAA exam. For the very same EC2 instance, you can pay in four very different ways:

Model	Commitment	Discount	Characteristics & use case
On-Demand	None	0% (list price)	Maximum flexibility, pay per hour/second. For unpredictable, short-term, experimental workloads.
Reserved Instances	1 or 3 years	Up to ~72%	Commit to a specific configuration. For steady, always-on workloads (e.g. a production DB).
Savings Plans	1 or 3 years	Up to ~72%	Commit to a $/hour rate, more flexible than RIs. For steady workloads whose configuration changes often.
Spot Instances	None	Up to ~90%	Use spare capacity, can be reclaimed (2-minute warning). For interruption-tolerant workloads.

A few distinctions to keep straight:

Reserved Instances come in Standard (deepest discount but least flexible) and Convertible (can be exchanged for a different configuration, smaller discount), with No / Partial / All Upfront payment options (more upfront = deeper discount). RIs apply to EC2, RDS, Redshift, ElastiCache, OpenSearch, and more.
Savings Plans come in three flavors: Compute Savings Plans (the most flexible — applies to EC2, Fargate, Lambda, across region/family), EC2 Instance Savings Plans (a deeper discount but locked to a family within one region), and SageMaker Savings Plans. Compared to RIs, Savings Plans are preferred because the commitment is in dollars per hour rather than to a specific instance.
Spot only fits interruption-tolerant workloads: batch processing, CI/CD, rendering, big data, or stateless workers in an Auto Scaling group. Absolutely avoid it for anything that must run continuously without interruption, like a production database.

A quick rule for the exam: steady/long-running workload → Savings Plans or RIs; interruption-tolerant, cheapest possible → Spot; short/unpredictable → On-Demand.

6.2. AWS Trusted Advisor — surface waste and risk

AWS Trusted Advisor is a service that automatically scans your environment and offers recommendations across several pillars: Cost Optimization, Performance, Security, Fault Tolerance, Service Limits (quotas), and Operational Excellence.

The Cost Optimization group in particular finds the classic kinds of waste: EC2 running at very low utilization, idle load balancers, EBS volumes not attached to anything, unused Elastic IPs, idle RDS, and Reserved Instance purchase opportunities.

A frequently-tested point: only accounts on a Business or Enterprise Support plan unlock the full set of Trusted Advisor checks. Basic/Developer plans get only a small subset of checks (mostly basic security and service limits).

6.3. AWS Compute Optimizer — ML-based rightsizing suggestions

While Trusted Advisor flags waste broadly, AWS Compute Optimizer specializes in rightsizing — it analyzes real CloudWatch metrics to suggest the right resource size for your needs, with estimated savings and performance impact.

It supports: EC2 instances, Auto Scaling groups, EBS volumes, Lambda functions, ECS services on Fargate, and RDS too. For example: “this m5.4xlarge averages only 8% CPU, so drop it to m5.xlarge to save ~70%”. The service is free, you just opt in.

6.4. Cost Optimization Hub — all recommendations in one place

The problem with having many recommendation sources (Cost Explorer suggests rightsizing, Compute Optimizer suggests sizes, there are Savings Plans purchase suggestions, etc.) is that they’re scattered and sometimes overlap.

Cost Optimization Hub is a centralized page in Billing and Cost Management that gathers all optimization recommendations in one place — rightsizing, deleting idle resources, purchasing Savings Plans/RIs — already deduplicated and normalized to estimated savings on a single, comparable measure, across all accounts in the organization. It’s where to start when you want a full optimization picture. The service is free.

7. Quick comparison & picking the right tool

Cost Explorer vs CUR: want a quick, visual look → Cost Explorer; want the most detailed raw data, queryable with Athena / fed into BI → CUR.
Budgets vs Cost Anomaly Detection: want alerts on a threshold you set → Budgets; want automatic anomaly detection with no threshold → Anomaly Detection.
Budgets vs CloudWatch Billing Alarm: both alert, but Budgets is more flexible and modern; the Billing Alarm is the old way and its metric is only in us-east-1.
Trusted Advisor vs Compute Optimizer: Trusted Advisor surfaces broad waste (multiple pillars); Compute Optimizer specializes in rightsizing based on real metrics.
Savings Plans vs Reserved Instances: both discount up to ~72% for steady workloads, but Savings Plans commit to a $/hour rate and are therefore more flexible when you change instance types often.

Conclusion: four questions, one toolkit

Back to month-end with a bill 40% higher than expected. Now you have a map for each question:

“Where did the money go?” → Cost Explorer for a quick look, CUR for deep dissection.
“Why the sudden spike?” → Cost Anomaly Detection could have warned you ahead of time.
“How do I know in advance next time?” → AWS Budgets with thresholds and (if needed) Budget Actions to auto-stop.
“How do I pay less?” → Savings Plans / RIs / Spot for the pricing model, Trusted Advisor / Compute Optimizer / Cost Optimization Hub to clean up waste.

What to take into the exam room:

Cost management is a lifecycle: estimate → understand → alert → organize → optimize. Each stage has its own tools.
Analysis is you actively opening something to look; alerting is the system actively telling you. Don’t confuse Cost Explorer (analysis) with Budgets/Anomaly Detection (alerting).
CUR = most detailed, into S3, queryable with Athena. Cost Explorer = visual, interactive.
Budgets = a threshold you set; Anomaly Detection = ML that learns on its own, no threshold needed.
Consolidated Billing via Organizations brings volume discounts and RI/SP sharing — the reason multi-account almost always goes with Organizations.
Most of these tools are free; the notable exceptions: the Cost Explorer API ($0.01/request), the third budget onward (~$0.02/day), and the full Trusted Advisor (needs Business/Enterprise Support).