15 Proven Datadog Cost Optimization Strategies

Your Datadog bill is driven by more than just the number of hosts you monitor. Hidden costs often lurk in unexpected places, from an explosion of custom metrics to long-term log retention policies and overlooked data transfer fees. These subtle drivers can quickly turn a manageable expense into a significant financial drain. To truly get your spending under control, you need to look beyond the obvious. Effective Datadog cost optimization requires a deep understanding of how the platform's pricing model works and where the financial traps lie. This article will uncover these common culprits and give you actionable strategies to address them.

Key Takeaways

• Control your data at the source: The most significant savings come from reducing data volume before it ever reaches Datadog. Implement filtering and sampling to discard low-value logs and metrics at their origin, ensuring you only pay to ingest and analyze high-signal data.
• Eliminate monitoring waste: Your monitoring bill directly reflects your infrastructure's efficiency. Right-size your hosts and containers and regularly audit and remove unused custom metrics to stop paying to observe resources that provide no value.
• Make costs visible and predictable: Turn cost management into a shared responsibility by using tags to attribute spending to specific teams. Combine this with budget alerts to create an early warning system that prevents surprise bills and keeps your optimization efforts on track.

How Does Datadog Pricing Work?

Before you can cut costs, you need a clear picture of what you’re paying for. Datadog’s pricing model is complex because it’s not a single product but a suite of services, each with its own billing structure. The platform’s flexibility is one of its biggest strengths, but it’s also why expenses can quickly become unpredictable. At its core, Datadog uses a usage-based model, meaning your bill is a direct reflection of your activity—from the number of hosts you monitor to the volume of logs you ingest.

This structure puts you in control, but it also requires active management. Without a solid strategy, you can easily find yourself paying for noisy, low-value data or inefficient processes. Understanding the key components of your bill is the first step toward building a more cost-effective observability practice. Let's break down how Datadog calculates your costs and where to look for the charges that often catch teams by surprise.

Breaking Down the Core Costs

Datadog offers more than 18 distinct services, which are grouped into categories like infrastructure monitoring, APM, log management, and security. Each of these services comes with its own pricing metric. For example, infrastructure monitoring is typically priced per host, while APM might be priced per host or by the volume of data traced. This multi-faceted approach means you aren't just paying one flat fee; you're paying for a collection of services that add up.

Services like Logs and Synthetics often use volume-based pricing, where your costs are tied directly to data ingestion and the number of tests you run. This is a critical area to watch, as even a small misconfiguration in your log processing pipeline can lead to a massive increase in data volume and, consequently, a much higher bill.

Watch Out for Hidden Cost Factors

Some of the most significant expenses in Datadog aren’t always the most obvious. Log indexing and retention, for instance, can become major cost drivers if left unmanaged. When you send logs to Datadog, you pay to ingest them, but you also pay to index them for searching and to store them over time. Many teams set generous retention policies without realizing the long-term financial impact. As your data grows, these storage costs can quietly spiral out of control.

Mindfully managing your log retention policies is key to balancing observability with cost efficiency. You need to decide which logs are critical enough to keep hot for immediate analysis and which can be archived or discarded. Without this level of governance, you’re likely paying a premium to store data you’ll never use. This is why it's so important to find a solution that gives you fine-grained control over your data pipelines before they hit your monitoring platform.

How Usage-Based Billing Adds Up

The core of Datadog's model is that you pay for what you use. This sounds fair, but it also means costs can escalate rapidly as your environment grows. Monitoring more hosts, collecting more log data, creating custom metrics, and storing data for longer periods all contribute to a higher bill. For large-scale enterprises, especially those with dynamic, containerized environments, usage can fluctuate dramatically, making budget forecasting a serious challenge.

This model requires constant vigilance. An application update that generates more logs or a new team spinning up a cluster of servers can have an immediate impact on your monthly spend. While the flexibility is great for scaling up, it lacks a built-in mechanism for cost control. To truly manage your expenses, you need solutions that can help you filter, aggregate, and route your observability data intelligently, ensuring you only pay for the high-value insights you actually need.

What Drives High Costs in Datadog?

Datadog’s usage-based pricing is incredibly flexible, but that same flexibility can lead to surprisingly high bills if you’re not paying close attention. Costs can escalate for many reasons, but they usually trace back to a few common culprits. It’s not about a single setting you forgot to flip; it’s about how your entire observability strategy interacts with the pricing model. When you monitor more hosts, ingest more logs, track more metrics, and retain data for longer, your expenses naturally climb.

Understanding these cost drivers is the first step toward getting your spending under control. It’s about shifting from a "collect everything" mindset to a more intentional approach. By identifying where the waste is happening, you can make targeted changes that reduce your bill without sacrificing the visibility you need. Let’s look at the five most common areas where Datadog costs can spiral and what you can do about them.

Collecting Too Much (or the Wrong) Data

It’s tempting to send all your data to Datadog, but this is one of the fastest ways to inflate your bill. Since you’re charged for what you use, every non-essential log, metric, and trace contributes to your monthly costs. This includes noisy, low-value logs from development environments, redundant data from misconfigured agents, or verbose application logs that don't provide actionable insights. The key is to become more selective. Before data even leaves its source, you should have a clear strategy for what’s worth collecting and what can be filtered out. This proactive approach to log processing ensures you only pay to store and analyze the data that truly matters.

Using Resources Inefficiently

Your Datadog bill is often a direct reflection of your infrastructure's efficiency. Over-provisioned servers, idle containers, and poorly optimized Kubernetes clusters don’t just waste compute resources—they also drive up monitoring costs. Datadog’s pricing is tied to the number of hosts and containers you monitor, so an inefficient setup means you’re paying to observe resources you don’t even need. By focusing on rightsizing your infrastructure and eliminating waste, you can achieve a double win: lower cloud provider bills and a leaner, more affordable Datadog invoice. It all starts with assessing what you’re running and why.

Creating Too Many Custom Metrics

Custom metrics are a powerful tool for tracking business-specific KPIs, but they can also be a hidden source of high costs. Datadog plans include a certain number of custom metrics per host, and exceeding that limit results in overage fees that add up quickly. Often, teams create custom metrics for everything without evaluating their actual business value. A better approach is to be strategic. Reserve custom metrics for the data points that are truly critical for your applications and business outcomes. Regularly audit your custom metrics to remove any that are unused or redundant to keep your usage within your plan’s limits.

Mismanaging Your Logs

Log management is frequently the biggest line item on a Datadog bill. The sheer volume of logs generated by modern applications can be staggering, and ingesting it all without a plan is a recipe for runaway spending. The most effective way to control this is to filter and process logs at the source, before they ever reach Datadog’s ingestion pipeline. By excluding verbose debug logs, dropping duplicate entries, and summarizing events, you can dramatically reduce the volume of data you send. This not only cuts your ingestion and storage costs but also makes your logs more meaningful and easier to search.

Getting Infrastructure Sizing Wrong

The size and scale of your infrastructure have a direct impact on your monitoring expenses. Since Datadog's pricing is often based on the number of hosts or containers, any inefficiency in your environment gets passed along to your observability bill. For example, running many small, underutilized virtual machines instead of fewer, larger ones can increase your host count and your costs. Similarly, failing to properly rightsize your Kubernetes clusters leads to monitoring more nodes than necessary. Optimizing your infrastructure footprint is a foundational step in managing your Datadog spend effectively.

Start Here: Essential Ways to Cut Costs

Before you dive into complex optimizations, it’s best to tackle the fundamentals. These are the essential, high-impact strategies that will give you the most significant savings with the least amount of effort. Think of this as trimming the fat from your observability spending. By focusing on these core areas first, you can build a solid foundation for a more cost-effective Datadog implementation and see immediate results on your monthly bill. Each of these steps addresses a common driver of high costs and gives you a clear, actionable path to control it.

Right-Size Your Infrastructure

One of the most direct ways to lower your Datadog bill is to reduce the number of resources it monitors. Since Datadog’s pricing is often based on the number of hosts or containers, any inefficiency in your infrastructure translates directly into higher monitoring costs. Start by rightsizing your clusters. Are you overprovisioning CPUs or running more nodes than you actually need? By optimizing your resource allocation and ensuring your infrastructure is scaled appropriately for your workloads, you not only improve your application performance but also create an immediate reduction in your observability spend. This is a foundational step in any cloud cost management strategy.

Manage Logs Smarter

Not all logs are created equal, but they all cost money to ingest, index, and store. The simplest and most effective way to cut log management costs is to stop sending low-value data to Datadog in the first place. Take a hard look at what you’re collecting. Are you ingesting noisy debug logs from a stable production application? Or duplicate data from multiple sources? By implementing filtering and processing rules at the source, you can dramatically reduce the volume of data sent to your platform. This approach ensures that your teams have the critical data they need for troubleshooting without paying to store information that provides little to no value. Focusing on efficient log processing is key to getting this right.

Optimize APM and Custom Metrics

Application Performance Monitoring (APM) and custom metrics are incredibly powerful, but they are also significant cost drivers. Each custom metric you create and every service you trace adds to your bill. It’s easy for these to multiply over time, especially in large teams where new metrics are created without retiring old ones. Schedule regular audits of your custom metrics to identify and remove any that are redundant or no longer used. For APM, be selective about which services you fully trace. You might not need 100% tracing on non-critical, internal applications. Prioritizing your most important services ensures you get deep visibility where it matters most without overspending.

Review Your Retention Periods

How long do you really need to keep your data? Datadog charges for data storage, so your retention policies have a direct impact on your costs. While some data may be subject to compliance requirements that dictate long retention periods, much of it is not. Assess your different data types and ask if they all need to be stored for the same duration. For example, you might keep high-level performance metrics for a year but only retain detailed trace data for 15 or 30 days. By mindfully managing your retention settings, you can balance your long-term analysis needs with the goal of reducing storage costs. This is a critical part of a strong data governance framework.

Control Your Data Ingestion

Beyond just logs, it’s important to have a strategy for controlling all data flowing into your observability platform. This means actively assessing the total volume of metrics, traces, and events you’re sending and looking for opportunities to reduce it without compromising visibility. Use Datadog’s built-in features, like tags and filters, to fine-tune what data is collected and indexed. You can also leverage sampling for high-volume data streams, like RUM (Real User Monitoring) sessions or traces, to capture representative data without collecting every single event. A proactive approach to managing data volume is essential for keeping costs predictable and under control.

Use Committed Use Discounts

If your organization’s usage is relatively stable and predictable, you shouldn’t be paying on-demand prices. Datadog, like many SaaS providers, offers significant discounts for customers who commit to a certain level of usage over a one- or multi-year term. Analyze your past consumption to forecast your baseline needs, then talk to your Datadog representative about annual contracts or other committed use plans. This single conversation can often lead to savings of 20% or more compared to a pay-as-you-go model. It’s a straightforward financial move that locks in a lower rate for the capacity you know you’ll need. You can find more details on their official pricing page.

Ready for More? Advanced Cost Controls

Once you’ve handled the low-hanging fruit, you can move on to more sophisticated strategies that require a bit more planning but deliver significant returns. These advanced controls are about building cost efficiency directly into your architecture and workflows. Instead of just reacting to high bills, you’ll be proactively managing the flow of data and creating systems of accountability that encourage smarter resource consumption across your organization. This is where you transition from simply using a platform to truly mastering its financial footprint.

Automate Scaling and Filtering

Manual clean-up is a good start, but automation is where you’ll find lasting savings. Instead of periodically reassessing your monitoring needs, you can build rules that automatically scale resources and filter data in real time. For example, you can configure your systems to automatically discard verbose, low-value logs from a development environment before they are ever sent to Datadog. By processing and filtering data closer to the source, you can make intelligent decisions about what’s worth paying to store and analyze. This approach ensures you’re not just managing costs but actively preventing them, all without compromising the critical observability you rely on.

Attribute Costs with Tags

To truly get a handle on your spending, you need to know exactly where it’s coming from. Datadog’s usage attribution lets you use tags to track which teams, services, or projects are generating the most monitoring costs. Implementing a consistent tagging strategy is key. This allows you to move beyond a single, monolithic monitoring bill and allocate costs accurately across different business units. When a team can see its direct impact on the bill, it creates a powerful incentive for them to optimize their own data streams. This level of visibility turns cost management into a shared responsibility and helps you have more productive, data-driven conversations about budgets.

Optimize Resource Allocation

Sometimes the best way to lower your Datadog bill is to look at the infrastructure it’s monitoring. Since Datadog’s pricing is often based on the number of hosts or containers, rightsizing your clusters can lead to direct savings. If your Kubernetes clusters are over-provisioned, you’re paying to monitor resources you don’t even need. By optimizing your underlying resource allocation, you reduce both your infrastructure costs and your monitoring expenses in one move. This holistic approach ensures your entire stack is running efficiently, which is a core principle of any effective cost management strategy.

Implement Data Sampling

You don’t always need to send 100% of your data to get the insights you need. Intelligent sampling allows you to reduce data volume while preserving observability. This goes beyond basic filtering; it’s about strategically deciding what to keep. For instance, you might ingest every single error log but only sample 10% of successful transaction logs. The most effective way to do this is at the source, using tools that can apply these rules before the data ever leaves your environment. This proactive approach to log processing can dramatically lower the volume of data you send, process, and pay for in Datadog.

Use Private Links for Security and Savings

Data transfer fees, or egress costs, are an often-overlooked part of the cloud bill. If you’re on AWS, you can use a service like PrivateLink to send your monitoring data to Datadog over a private connection instead of the public internet. This simple architectural change can cut your data transfer costs significantly—sometimes by as much as 80%. It’s a huge win for your finance team, but it’s also a big plus for your security team. Keeping your sensitive operational data off the public internet strengthens your overall security and governance posture, making this a powerful two-for-one optimization.

How to Know If Your Changes Are Working

Making changes to your Datadog configuration is one thing; knowing if they’re actually saving you money without hurting performance is another. You can’t just trim your data ingestion and hope for the best. To make smart, sustainable optimizations, you need a clear feedback loop. This means moving from guesswork to a data-driven approach where you can directly measure the impact of every adjustment.

The goal is to create a system of checks and balances. This system will help you confirm that your cost-saving strategies are effective and ensure you haven’t accidentally sacrificed critical visibility. By setting up the right tracking and alerts, you can confidently answer the question, "Are we saving money?" while also protecting your system's reliability. It’s about building a continuous improvement cycle where you can tweak, measure, and repeat until you find the right balance for your organization. This approach turns cost management from a reactive fire drill into a proactive, strategic discipline.

Set Up Usage Attribution

If you don’t know where your costs are coming from, you can’t effectively control them. That’s where usage attribution comes in. Datadog allows you to track which teams or services generate the most monitoring costs by using tags. By tagging resources by team, project, or application, you can get a granular view of your spending. This isn't just about pointing fingers; it's about creating accountability and empowering individual teams to manage their own consumption. When a team can see its direct impact on the bill, they’re more likely to be mindful about the data they send. This simple step is foundational for creating a culture of cost-consciousness across your engineering organization.

Configure Budget Alerts

No one likes a surprise bill at the end of the month. Setting up budget alerts is a crucial step for keeping your Datadog costs in check and avoiding those unpleasant conversations with finance. These alerts act as an early warning system, notifying you when your spending approaches preset limits. This allows you to be proactive rather than reactive. For example, you can set an alert when you hit 75% of your monthly budget, giving you time to investigate spikes and make adjustments before you go over. Think of it as a safety net that ensures your cost optimization efforts stay on track and that you maintain control over your monitoring expenses.

Define Your Key Performance Indicators (KPIs)

Before you start cutting data, you need to define what success looks like for your applications. Cost reduction is important, but not at the expense of performance or reliability. Identify the key performance indicators (KPIs) that matter most for your business—things like P99 latency, error rates, and uptime. These metrics will be your North Star. As you implement cost-saving changes, keep a close eye on these KPIs. If you reduce log ingestion and your error rates start to climb, you may have cut too much. This practice ensures you’re making informed trade-offs and not sacrificing the health of your systems for the sake of savings.

Analyze the Impact of Your Changes

Once you’ve made an optimization—whether it’s filtering logs, adjusting retention, or sampling traces—you need to verify its impact. Don't just assume it worked. Regularly analyzing the impact of your changes is essential for understanding which strategies are effective and which aren't. Set aside time each week or month to review your Datadog usage dashboards. Did that new filter actually reduce log volume as expected? Did consolidating custom metrics lower your bill? This review process creates a feedback loop that helps you refine your approach over time, doubling down on what works and correcting course when something doesn't deliver the expected results.

Monitor Costs Continuously

Cost optimization isn't a one-and-done project; it's an ongoing process. Your applications, infrastructure, and teams are constantly evolving, and your monitoring strategy needs to adapt along with them. Make cost monitoring a regular part of your operational rhythm, just like performance monitoring. By keeping a close and continuous eye on your expenses, you can spot trends, catch anomalies early, and ensure your resources are always focused on essential monitoring activities. This sustained attention is what separates organizations that temporarily cut costs from those that build a long-term, cost-efficient observability practice.

A Practical Guide to Implementation

Knowing where your costs come from is one thing; actually reducing them is another. The good news is that you can make a significant impact with a few targeted changes. This guide breaks down the practical steps you can take to get your Datadog spending under control, starting today. Think of these as the foundational habits for a more cost-effective observability practice. By focusing on how you query, collect, and manage your data and infrastructure, you can build a more efficient and sustainable system from the ground up.

Optimize Your Queries

One of the most direct ways to cut costs is to be more selective about the data you send to Datadog in the first place. It’s easy to fall into the trap of collecting everything, but this approach often leads to noisy, low-value data that inflates your bill. Instead, focus on what’s truly necessary for your observability needs. Before data leaves its source, take the time to refine your queries to filter out redundant or irrelevant information. By intelligently processing your logs closer to the source, you ensure that only high-signal data makes it to your monitoring platform, reducing both ingestion and storage costs.

Fine-Tune Data Collection

Filtering logs at the source is one of the simplest and most effective ways to reduce data volume. Rather than paying to ingest and process unnecessary data within Datadog, you can use tools to exclude it before it ever leaves your environment. This could mean dropping verbose debug logs from production applications or filtering out routine health check pings that don’t provide critical insights. Implementing this kind of pre-processing gives you granular control over your data pipeline. Modern data processing solutions allow you to enforce these rules consistently across your entire infrastructure, from the cloud to the edge.

Analyze Resource Usage

You can't optimize what you can't measure. Datadog’s usage attribution features, which often use tags, are essential for understanding where your monitoring budget is going. By tagging resources by team, project, or service, you can create a clear picture of which parts of your organization are generating the most monitoring data and, therefore, the highest costs. This data allows you to have informed conversations with team leads about their usage and identify specific areas for optimization. It shifts the responsibility from a central platform team to the service owners, creating a culture of cost awareness and offering a path to significant cost savings.

Follow Sizing Best Practices

Your infrastructure footprint has a direct impact on your monitoring bill. Since Datadog’s pricing is often based on the number of hosts or containers being monitored, rightsizing your environment is a powerful cost-control lever. Over-provisioned Kubernetes clusters or idle virtual machines don’t just waste compute resources; they also drive up your monitoring expenses. Regularly review and adjust your resource allocation to match actual demand. Efficiently managing your fleet of hosts and containers is a key part of any cost optimization strategy, ensuring you only pay for what you truly need.

Common Mistakes to Avoid

As you start to get your Datadog spend under control, it’s just as important to know what not to do. Many teams, with the best intentions, fall into common traps that can undermine their efforts or even create new problems down the line. When you’re under pressure to reduce costs, it’s easy to make a reactive decision that feels right in the moment but hurts you later.

Think of cost optimization as a strategic initiative, not just a budget-cutting exercise. The goal is to eliminate waste while preserving the critical visibility you need to run your systems effectively. A hasty move can lead to monitoring gaps, which can turn a minor issue into a major outage. By understanding the most common pitfalls, you can make deliberate, informed changes that lead to sustainable savings without compromising performance or reliability. Let’s walk through some of the biggest mistakes we see teams make so you can steer clear of them.

The Risk of Cutting Too Much

When you see a massive bill, the first instinct might be to cut everything that isn’t essential. But what’s truly essential? Aggressively slashing your monitoring can leave your engineering teams flying blind. The short-term savings you gain can be quickly wiped out by the cost of a single prolonged outage that you couldn't see coming. While building your own monitoring stack with open-source tools might seem like a cheaper alternative, the engineering time and expertise required can often cost more than a SaaS service in the long run. The key is to be surgical. Focus on reducing noisy, low-value data, not eliminating your ability to see what’s happening in your environment.

Choosing the Wrong Commitment Level

Datadog, like many SaaS providers, offers discounts for annual commitments. It’s tempting to lock in the highest discount tier by promising a high level of usage, but this can easily backfire. You can almost always increase your commitment during the year, but you can never decrease it. If you overestimate your needs, you’ll be stuck paying for capacity you aren’t using. A much safer approach is to start with a conservative commitment. As you get a better handle on your actual usage and growth, you can work with your Datadog representative to increase your commitment level. This gives you the flexibility to adapt without getting locked into an expensive mistake.

Forgetting to Manage Custom Metrics

Custom metrics are incredibly useful for tracking application-specific data points, but they are also a primary driver of unexpected costs. Each Datadog plan includes a certain number of custom metrics per host, and exceeding that limit gets expensive quickly. The mistake many teams make is creating them without a clear governance strategy. Over time, you can accumulate hundreds or thousands of redundant, unused, or poorly tagged metrics that inflate your bill. Make it a regular practice to audit your custom metrics. Only create them for hosts that truly need them, and establish a process for reviewing and removing metrics that are no longer providing value.

Making Data Retention Mistakes

How long should you keep your logs? The answer directly impacts your storage costs. Many organizations default to a one-size-fits-all retention policy, either keeping data for too long "just in case" or not long enough to meet compliance requirements. Both are costly mistakes. A better approach is to create tiered retention policies based on the data's purpose. For example, security and audit logs may need to be stored for a year or more to meet regulatory standards, while debug logs from a development environment might only be needed for a few days. By mindfully managing your log retention, you can strike the right balance between meeting governance needs and keeping storage costs in check.

Overlooking Egress Charges

Egress charges are one of the most frequently overlooked costs associated with cloud monitoring. These are the fees your cloud provider charges you to move data out of their network and into a third-party service like Datadog. If you’re sending a high volume of logs and metrics, these data transfer fees can add up to a significant, and often surprising, expense on your monthly cloud bill. Fortunately, there’s a straightforward fix. If you’re on AWS, for example, you can use a service like AWS PrivateLink to send data to Datadog over a private connection. This simple change can reduce your data transfer costs by as much as 80% while also improving security.

Related Articles

• Log Processing Solutions - Streamlined Data Management
• Distributed Systems: Weighing the Advantages & Disadvantages
• Financial Services Solutions
• Healthcare and Life Sciences Solutions
• Manufacturing Solutions

Frequently Asked Questions

What's the first thing I should do to lower my Datadog bill? Start by looking at what you’re monitoring and what you’re sending. The two biggest cost drivers are often an oversized infrastructure and an unfiltered stream of logs. First, check if you are monitoring idle or over-provisioned hosts and containers. Rightsizing your infrastructure means you pay to monitor fewer resources. At the same time, put a filter on the data you send. By processing logs at the source and dropping noisy, low-value information before it ever reaches Datadog, you can make an immediate and significant dent in your ingestion costs.

How can I reduce log volume without creating blind spots for my engineers? The goal is to be strategic, not just to cut data indiscriminately. You can achieve major cost reductions without sacrificing visibility by focusing on the value of your data. For example, you likely don't need verbose debug logs from a stable production application, but you absolutely need every error log. The key is to implement rules that filter out the noise close to the source. This ensures your teams still have all the critical information they need for troubleshooting while you stop paying to ingest and store data that doesn't help them solve problems.

My teams don't seem to care about monitoring costs. How can I change that? Accountability starts with visibility. If your monitoring bill is just a single, large number, it’s impossible for any individual team to see their impact. The most effective way to create a culture of cost-consciousness is to implement a consistent tagging strategy. By using tags to attribute usage to specific teams, projects, or services, you can show each group exactly what their part of the bill is. When a team can directly see how their choices affect costs, they are naturally motivated to be more efficient with the data they send.

What are custom metrics, and why do they cause costs to spike? Think of custom metrics as specialized counters you create to track things unique to your business, like the number of items added to a shopping cart or files processed by an application. They are incredibly powerful but can become a major source of unexpected costs. Your Datadog plan includes a certain number of custom metrics per host. If you exceed that limit, you start paying overage fees that add up quickly. Teams often create them without a clear strategy, leading to a buildup of unused or redundant metrics that quietly inflate your bill.

Is signing an annual contract with Datadog a good idea for saving money? It can be, but you need to be careful. Committing to a certain level of usage for a year or more can secure you a significant discount compared to paying as you go. This is a great move if your consumption is stable and predictable. The risk is overcommitting. You can almost always increase your commitment level during the contract term, but you can never decrease it. If you overestimate your needs, you'll end up paying for capacity you don't use. It's often best to start with a conservative commitment and adjust it upward as you get a clearer picture of your long-term usage.