Treat those cells as questions for Grafana or your account's billing page before you quote a client, not as zeros. | | Platform fee | Included series | Included logs/traces GB | Retention | Overage per 1,000 series and per GB | Seats | |---|---|---|---|---|---|---| | Free | None | Not confirmed | Not confirmed | Not confirmed | Not confirmed | Not confirmed | | Pro | Not confirmed | Not confirmed | Not confirmed | Not confirmed | Not confirmed | Not confirmed | | Enterprise | Set by contract | Set by contract | Set by contract | Set by contract | Set by contract | Set by contract | There's no public Enterprise figure to copy, because Grafana negotiates those contracts one by one, and any quote is the only number that counts. The overage column is the one to confirm first, because it prices everything you use beyond the included amounts. ### How usage is metered One metric carrying a client, environment and pod label becomes many series, because Grafana counts metric series as distinct label combinations. Adding a single high-variety label can multiply the count without anyone noticing a change on a dashboard. A verbose debug setting costs money whether or not anyone reads the output, because Grafana meters logs and traces by volume ingested. Retention is the third lever, and it applies after ingestion: you pay for data for as long as you keep it. Every client dashboard and alert rule sits on top of those three meters, and the automation that routes alerts onward adds no fourth: [Activepieces](https://www.activepieces.com) bills a run as 1 credit however many steps it has, so a ten-step routing flow costs what a two-step one does. The next section turns them into caps you can set before the first invoice arrives. ## The Tuesday Usage Alert at a 12-Person Team At 09:40 on a Tuesday, a 12-person team learns that one deploy, not the Pro plan it just bought, is now setting its Grafana Cloud bill. The team is illustrative, not a real customer, and it moved from Free to Pro the week before. ### Day 1: the realistic baseline bill On day 1 the invoice looks like the sticker: the Pro platform fee plus a small amount of usage. Series counts are modest, and the services write logs at their normal volume. Nobody opens the usage page, because nothing gives them a reason to. That baseline is the figure finance approved, so every later change is a drift from it. If the team never records it, it has nothing to compare the drift against. ### Day 9: the deploy that multiplied series On day 9 a deploy adds a label to a request metric, and the label carries a value like a user ID or a full request path. **Every distinct value creates its own active series**, so one metric becomes thousands.  Nothing breaks and dashboards still load, so the change goes unnoticed. Even if nobody queries those series, Grafana bills each active one, so the cost is real anyway. The same release also logs more per request, which raises the second meter. ### Day 12: the alert and the projected invoice By the time the bill moved, three meters had moved with it, and each has its own rate: - Active series, at $6.50 per 1,000: - Logs written, at $0.40 per GB - Logs processed, at $0.05 per GB The first is the one the deploy multiplied, so every extra thousand series adds $6.50 to the monthly line. The other two rise with traffic: each GB written costs $0.40, and each GB processed adds $0.05 on top, so you can't budget a fixed infrastructure cost. By day 12 the alert is projecting those rates across the rest of the month, and the projection's what lands on the team lead's desk. Grafana's already billed three days at the new series count.  The projected invoice is also well above the one finance approved, and nobody had agreed to that gap. Had a cap on active series been in place, it would have stopped the label on day 9, before the first unplanned dollar. The next step is to turn each meter into a limit set before the deploy, not after the alert. ## Cascading meters and hidden costs In a fixed order, each meter turned into an overage, and the earlier ones cost more than the later ones because they compound. The sequence below shows how one routine change cascaded into the invoice, so you know which limit to set first. 1. A new label on a metric multiplied the number of active series. 2. Grafana Cloud billed those series on usage, so the cost rose as soon as the series count did. 3. Log ingestion passed the volume included in the plan. 4. Someone extended retention beyond the default window, so Grafana stored and billed the extra logs for longer. 5. The team added seats and add-on products on top of an already inflated base. Steps one and two happened in a single deploy. Steps three to five were slower, which is why nobody noticed them until last. ### Active series and high cardinality costs An active series is one unique combination of a metric name and its label values that's currently reporting. Add a label that takes many values, such as a user ID or a request path, and every existing series splits into many. The deploy looks harmless in code review and shows up only on the invoice. On a shared stack this is also where one client's instrumentation quietly raises the bill for everyone, so tag series by owner before you need to argue about who pays. ### Grafana Cloud log volume and retention costs Ingestion and retention are separate charges. Volume is what you send in; retention is how long you keep it. A noisy service that logs debug output in production inflates the first, and a "keep everything longer, just in case" request inflates the second. The two multiply, because each extra gigabyte now stays on the bill for longer. Cap what each service may send, and give every client a retention window they've agreed to pay for. ### Seats, add-on products and overage After a contractor's engagement ends, seats are the easiest line to leave running, though they're the smallest. Add-on products, such as extra capabilities switched on during the incident, bill separately from the base plan. Once a stack exceeds its included allowance, usage keeps billing instead of stopping, so the overage is the real price of the plan. Review seats and add-ons monthly, and treat any unexplained one as a cost with no owner. ## Grafana Cloud Against Datadog, New Relic and Elastic Of the four, **Grafana Cloud is the easiest to cap**, because its bill tracks series, log volume and retention most directly. The others meter on different units that a spike hits in different ways. The comparison below uses the same workload throughout, and it deliberately carries no dollar rates, because those belong on each vendor's own pricing page at the time you buy. ### Pricing model side by side Read the table for what each vendor actually counts; the takeaway is that only the first column of units is one you can cap from your own configuration. | | Pricing model | Free tier | Effect of the day-9 spike | Self-hosting or lock-in | |---|---|---|---|---| | Grafana Cloud | Usage on metric series, log volume and retention | Yes; check the vendor page for current limits | Lands on series and log lines, so a cap stops it | Open-source components can be run yourself, so exit is possible | | Datadog | Per host for infrastructure, plus per GB for logs | Limited; check the vendor page | Extra hosts and log GB both raise the bill | Proprietary SaaS; no self-hosted option | | New Relic | Per GB ingested, plus per user | Yes; check the vendor page | Ingest overage raises the bill, and seats stay fixed | Proprietary SaaS; no self-hosted option | | Elastic | Sized by deployment resources | Trial rather than a standing tier | Absorbed until capacity runs out, then you resize | Self-managed Elastic is available | Rates that a vendor doesn't publish openly, such as negotiated enterprise tiers, are left out on purpose. If an account manager quotes you one, get it in writing before you compare. ### What the day-9 spike costs on each On Grafana Cloud the spike shows up as extra series and log lines, and a cap on those two stops it before it reaches the invoice. On Datadog it lands twice, once if the spike adds hosts and again through log gigabytes, so a single incident can inflate two line items. On New Relic it lands on ingest alone, which is simpler to reason about but harder to cap without dropping data at the source. [Elastic](https://www.elastic.co/pricing/cloud-hosted) hides the spike inside provisioned capacity until the cluster fills, so the cost arrives later as a resize rather than as a surprise overage. Watch who absorbs the overage. On a shared account, **one noisy client's spike's billed to everyone**, so cap per client before comparing vendors at all. ## Per-client Grafana Cloud series and log caps One client's spike now lands on that client's line and not on the shared invoice, because each client has its own ceiling on series, log volume and retention. Five rules do the work, and each one stops a specific kind of overage before it's billed. ### Set Grafana Cloud usage limits at ingestion At the point of ingestion, active series are capped per client, with drop rules for labels we know to be unbounded. Cutting there is cheaper than cleaning up afterward, because a series that's never accepted is never billed. In a real case the size of the prize is visible. [Grafana Labs](https://grafana.com/success/sailpoint/) reports that SailPoint, an identity-security software vendor, started 2023 with 75 million active series. Cardinality cuts removed 25 million of them, leaving 50 million.  **That's a 33% reduction**, so a third of what the team was paying to store was labels nobody needed, such as per-request identifiers. Take from it that the drop came from trimming labels, not from monitoring fewer services. Logs get the same treatment. Noisy, low-value streams such as health checks are sampled, and retention is set per client and not left at the default. A client who wants ninety days of debug logs can have them, but on their own line, not hidden inside everyone else's rate. ### Grafana Cloud budget alerts and a named owner When usage crosses a set share of each client's monthly allowance, a spend alert fires. Early in the month it gives us time to act. On the last day it only explains the invoice. One named person owns the number every month. That person reviews usage by client, and any client running over their allowance gets a conversation about their rate before the next cycle, not a quiet subsidy from the others. Before any deploy that adds or changes a metric label, someone checks whether that label can take unbounded values, since one such label can multiply series count by itself. The Monday checklist: 1. Confirm every client has a series limit and a log retention setting. 2. Check last week's usage against each allowance and note who is over. 3. Review any new metric labels in this week's deploys. 4. Confirm the alert thresholds still match current allowances. 5. Record the owner's monthly review date. With these in place, we expect the next invoice to land close to the plan's sticker price. That's an estimate, not a result, and the first full month of usage will confirm or correct it. ## Fix Alert Routing With Activepieces Activepieces doesn't change what Grafana Cloud charges; it changes how quickly the right person learns that usage is drifting from the estimate, and its Free plan, at $0 with 1,000 credits a month, is enough to build and run the routing flow before anyone commits budget.  It's a workflow automation tool, and here it takes a Grafana alert webhook, the message Grafana sends when a rule fires, and turns it into work someone has to answer for. ### Grafana alert webhook to ticket and chat From a Grafana alert webhook to the person who reviews the result, the diagram shows the path. The webhook enters an Activepieces workflow, which produces a chat message, a ticket and a request to the named owner, using integrations drawn from the 746 counted in its MIT-licensed open source monorepo. The point is that the alert never lands in a shared channel where everyone assumes someone else will handle it. That matters for the same reason shared accounts hide costs. When a series or log-volume alert goes to a general channel, nobody owns it, and the overage gets absorbed by whoever holds the account. When the workflow asks the specific owner of the offending service or client to act, the cost has an address. The ticket leaves a record of who was asked and what they decided, which you can point to when a client questions a line on the invoice. The person at the end's deliberate. The workflow routes and records the response, and a human decides whether to drop a label, shorten retention or accept the spend. Automating that decision would just move the surprise somewhere less visible. The next question is whether you need this at all. ### When Grafana's own contact points are enough If one team owns the whole stack and the alert only needs to reach one place, Grafana's own contact points are enough. They send a notification to a chat tool or an email address directly, with no extra system to maintain. Add Activepieces when the response has more than one step, or when the right recipient depends on which service or client triggered the alert; every plan, including free, allows unlimited flows, so a separate flow per client adds no cost beyond its runs.  If you find yourself copying alerts by hand into a ticket queue, or asking in chat who owns a noisy metric, the workflow is worth building. If you aren't, leave it out. Neither option lowers a bill on its own. The caps on series, volume and retention do that. The routing only makes sure someone hears when one of them is about to be tested. ## Frequently asked questions In brief, the table gives five answers, as checked on the dates shown. Routing tells you a limit is close, but only these answers tell you whether the limit holds. | Question | Short answer | Grafana docs page checked | Date checked | |---|---|---|---| | Is the free tier enough for a small team? | Yes, for light use | Grafana Cloud pricing page | 2026-11-03 | | What happens past the free tier? | Limits apply or you upgrade | Grafana Cloud pricing page | 2026-11-03 | | Is there a hard spending cap? | Alerts, not a dollar ceiling | Grafana Cloud billing documentation | 2026-11-03 | | Is self-hosting cheaper? | Only if you count your own time as free | Grafana Cloud pricing page | 2026-11-03 | | Are users billed? | Seats beyond the included ones, on paid plans | Grafana Cloud pricing page | 2026-11-03 | Each answer below carries a condition, and the conditions are where the invoice moves. ### The free tier Is Grafana Cloud free for small teams? Yes, if the team stays inside the free tier's usage limits for series, logs and retention. A small team with a handful of services can run there indefinitely. The first busy service or verbose log stream ends that, so treat the free tier as a trial of your own habits. What happens when I exceed the Grafana Cloud free tier? Usage past the limits is restricted, or you move to a paid plan. Check the pricing page for which applies to which signal. Either way, the overage is set by volume, so one noisy client sharing your account can push everyone over. ### Spending caps Can I set a hard spending cap on Grafana Cloud? **I found no setting that stops billing at a fixed dollar amount.** What you get is usage visibility and alerting. The cap is therefore one you build: limit series at the source, drop low-value logs before ingestion, and keep retention short. Do that per client, or one client's growth lands on the shared invoice. ### Self-hosting and seats Is self-hosting Grafana cheaper than Grafana Cloud? Self-hosting removes the vendor line item and replaces it with servers, storage, upgrades and an on-call rota. It's cheaper only when someone's time costs nothing, which on a retainer it never does. Price the labor before you compare. Does Grafana Cloud bill for users? On paid plans, yes: seats beyond those included are charged. Contractors and client staff with logins count, so remove accounts when an engagement ends. ## Related reading - [Zapier vs Make for Agencies 2026 Comparison](https://www.activepieces.com/blog/zapier-vs-make-for-agencies-2026-comparison) - [Applied Epic AI Integration: A 2026 Guide for Agencies](https://www.activepieces.com/blog/applied-epic-ai-integration-a-2026-guide-for-agencies) - [CoreWeave Pricing 2026: GPU Cloud Costs Compared](https://www.activepieces.com/blog/coreweave-pricing-2026-gpu-cloud-costs-compared) ## References - [Grafana Labs](https://grafana.com/success/sailpoint/)On a shared account, the whole stack pays for one noisy client's series and logs, and the quiet clients quietly cross-subsidize it.