← All articles

Monitoring SLA: A Practical Guide for Small SaaS Teams

Published · 11 min read

Key takeaways

  • An SLA is the contractual promise, an SLO is your internal target, and an SLI is the raw measurement feeding both.
  • 99.9% uptime over a rolling 30-day window allows about 43 minutes of downtime per month.
  • Alert on error budget burn rate rather than raw downtime to catch breaches before they happen.
  • Probe the paths customers pay for, like login, database reads, and webhooks, not just the homepage.
  • Realistic SLA monitoring spend at 1-10 person scale is $0-25/mo using tools like Uptime Kuma or StatusPulse.

Monitoring SLA means tracking whether your uptime and performance commitments to customers are actually being met, then reporting on it. This guide walks through the full process: defining SLOs, computing error budgets, picking probes, building a dashboard, automating reports, and handling breaches, using tools that fit a 1-10 person team instead of a $200/mo Atlassian Statuspage bill.

What does SLA monitoring mean (and how is it different from SLO monitoring)?

SLA monitoring means measuring whether you're meeting the uptime commitments written into customer contracts, and keeping the evidence to prove it. An SLA is the contract, an SLO is your internal target (say, 99.9% monthly uptime), and an SLI is the raw measurement that feeds both.

SLA vs SLO vs SLI in one paragraph

The SLI is what you measure (successful probes ÷ total probes), the SLO is the threshold you aim for internally, and the SLA is the contractual promise with credits attached if you miss it. Google's SRE book frames the gap between SLO and reality as your error budget, the downtime you're allowed to spend each month (SRE workbook, SLO chapter).

Why solo founders care: SLAs are contractual, SLOs are internal

Miss an SLO and you fix things quietly. Miss an SLA and you owe credits, and customers can churn over the paper trail. That's why your monitoring needs protocol-level probes and a public status page, not just a homepage check. StatusPulse starts at $5/mo (pricing) so the compliance evidence doesn't cost more than the tooling.

What does an SLA monitoring process look like step by step?

An SLA monitoring process runs five steps: pick the SLI, set the SLO target and window, compute the error budget, probe the real failure paths, and report on a fixed cadence. Skip any one of them and you get alerts that fire at 3 AM for nothing, or a status page that says "all systems operational" while customers cannot log in.

Step 1: Pick the SLI that matches your contract

Your SLI is the measurement your SLA promise rests on. For most B2B SaaS that is uptime of the API or dashboard, but latency or error rate can be the better fit if your product is a checkout flow or a webhook receiver. Pick the one a customer would actually cite in a refund request.

Step 2: Set the SLO target and window (99.9% over 30 days, not forever)

A target without a window is unmeasurable. 99.9% over a rolling 30 days allows about 43 minutes of downtime; over a year, it allows 8.7 hours. The window decides whether one bad afternoon breaches the contract or not.

Step 3: Compute the error budget and alert on burn rate

Subtract your SLO from 100% and you have the error budget: the downtime you are allowed to spend. Alert on burn rate, not on raw downtime. If you burn 10% of the budget in an hour, that pace breaches 99.9% in under a day, so page someone. If you burn 10% across a week, open a ticket instead. Burn-rate alerts are what stop false alarms, because they only fire when the pace of failure threatens the window, not every time a single probe hiccups.

Step 4: Probe the real failure paths, not just the homepage

A homepage check returns 200 while your database is on fire. Probe the paths that pay you: the login endpoint, a read against the database, the webhook receiver. This is where protocol-level probes beat URL checks, and it is the gap between a $5 tool and a $200 one that both show a green light. StatusPulse runs protocol-level probes for this reason; UptimeRobot and Pingdom-style URL checks are the shallower end of the same pool, and Uptime Kuma covers it if you self-host (Uptime Kuma on GitHub).

Step 5: Report on a fixed cadence

Pick a cadence, monthly is the norm, and publish uptime against the SLO for that window. A public status page plus a scheduled report covers most contracts. For more on building the dashboard behind that report, see the rest of this guide series at StatusPulse's guides.

Which SLA metrics should you track: uptime, error rate, or error budget burn?

Track all three, but each answers a different question: uptime satisfies the contract, error rate catches what uptime hides, and burn rate tells you when you're about to breach.

Uptime percentage: what 99.9% actually allows per month

99.9% is 43 minutes and 12 seconds of downtime per month. 99.95% is 21.6 minutes. Put your allowance in the SLA itself so customers know exactly what they're buying, then measure it with protocol-level probes rather than a homepage URL check. A homepage can return 200 while your WebSocket endpoint is dead; that's why StatusPulse probes at the protocol level, including WebSocket monitoring, instead of just pinging a URL.

Latency and error rate as SLIs for API-first products

If your product is an API, uptime alone undercounts incidents. An endpoint that answers in 8 seconds with 500s on 4% of calls is "up" to a URL checker and broken to a customer. Define SLIs as p95 latency under a threshold and 5xx rate under a percentage, measured per endpoint.

Error budget burn rate as your early-warning metric

Error budget = 1 minus your SLO. Burn rate is how fast you're spending it. A 2-hour outage burns your monthly 99.9% budget in one shot; slow degradation burns it invisibly. Alert on burn rate (e.g., 14.4x means you'll exhaust a 30-day budget in ~2 days) so you act before the breach, not after.

MetricCatchesMisses
Uptime %Total outagesPartial failures, slow responses
Error rate / latencyDegraded endpointsSilent SLA erosion over the month
Burn rateBreach trajectoryNothing alone; needs the other two

This is also why a $200/mo Statuspage bill is hard to justify: Atlassian's Statuspage pricing runs from tens to hundreds of dollars per month for page hosting, while the monitoring that generates these metrics is a separate product. StatusPulse bundles both from $5/mo, though it lacks the enterprise audit trails and multi-team features Statuspage sells at the top tier, so run your own checklist before switching.

How do you build an SLA monitoring dashboard your customers and leadership both trust?

Public layer: status page with rolling uptime windows

Your public dashboard shows what customers can verify: current status, per-component state, and rolling 90-day uptime per component. Rolling windows matter because a single "99.9%" lifetime number hides last month's outage. Show incident history with timelines so a customer disputing an SLA credit can see exactly when the outage started and ended. StatusPulse's public pages are driven by the same probe types that trigger alerts, so the page reflects what monitoring actually saw.

Internal layer: burn-rate alerts and per-component SLIs

Keep a second dashboard internal. Per component, track the SLI (successful probes / total probes over the SLA window) and the error budget burn rate. Alert on burn rate, not raw downtime: a fast burn (2% of budget in 1 hour) pages on-call; a slow burn (5% in 24 hours) opens a ticket. This catches degradation before the SLA is breached, while a monthly uptime report only tells you after the fact.

Honest limits: what a status page dashboard cannot prove

A status page is your own monitoring's word, not a neutral audit. If your probes run from one region and a customer's users hit from another, your 99.95% and their experience can differ. For real SLA disputes, export raw probe data with timestamps so both sides can recompute the number.

What are the best open source SLA monitoring tools in 2026?

Three open source options cover most SLA monitoring needs, but each leaves a gap you have to fill yourself.

Uptime Kuma: free probes, no built-in SLA reporting

Uptime Kuma gives you unlimited HTTP, TCP, and DNS probes on a $0 VPS, with a public status page included. It does not compute rolling 30-day uptime percentages or error budget burn, so you export the data and do the SLA math in a spreadsheet. Fine at 5 probes; tedious at 50.

Prometheus + Grafana: burn-rate alerts, you build the dashboard

Prometheus with the blackbox exporter measures probe latency and success rate, and multi-window burn-rate alerts are the standard pattern in the SRE workbook. Grafana renders the SLO dashboards. Expect a weekend to set up and a few hours a month to maintain: TLS cert renewals, alert routing, dashboard upkeep.

Open source vs paid: the maintenance cost nobody prices in

Running your own stack means you are on-call for the monitoring itself. Paid tools like Better Stack bundle monitoring, on-call, and status pages for a monthly fee in the tens of dollars per their pricing page, which buys back that maintenance time. StatusPulse sits between: protocol-level probes plus SLA reporting from $5/mo, though it lacks Prometheus's custom burn-rate alerting if you need exotic SLO math. Whichever route you pick, monitor the things customers actually pay for, like DNS resolution, not just your homepage.

How do you automate SLA monitoring and reporting without manual spreadsheets?

Manual SLA reporting breaks the moment you get busy. Your uptime monitor already records every probe result, so the fix is to pipe that data straight into scheduled reports and billing triggers instead of copying numbers into a spreadsheet each month.

Scheduled PDF/email uptime reports

Most monitoring tools, including Statuspage.io, UptimeRobot, and StatusPulse, can generate recurring uptime reports on a schedule you set: monthly, weekly, or per incident. Configure a monthly report per customer or per SLA tier, covering the uptime percentage, incident count, and total minutes of downtime in the window. Email it automatically so the customer sees the number before they ask for it. That single habit eliminates the end-of-month scramble and gives you a timestamped record if a customer disputes an SLA claim later.

To make the report trustworthy, compute it from the same probe data that drives your alerts and status page, not from a hand-copied spreadsheet. Include the measurement window, the probe locations, and any excluded maintenance windows, so a customer can recompute the number if they want to. Set the report to send on a fixed calendar day, the first business day of the month works well, and log the send date alongside the report itself.

Feeding SLA data into billing for service credits

If your contract promises credits for missed SLAs, automate the trigger: when a reporting window closes below the committed uptime, generate a credit line item and push it to your billing system (Stripe, Chargebee, or whatever you invoice with). Even a simple webhook or a script that reads the report API and writes a credit note beats a human remembering to do it. The key is that the same probe data drives the status page, the report, and the credit, so there is one source of truth and no room for disagreement about what happened.

How do you turn SLA breach data into incident post-mortems and customer credits?

A breach triggers three artifacts: a post-mortem within 48 hours, a credit calculation, and a customer-facing email that states both. Skip the post-mortem and the same outage recurs; skip the credit and the SLA becomes marketing copy.

Breach post-mortem template for a 3-person team

Keep it to five fields: timeline (detection to resolution, with timestamps from your probes), customer impact (which endpoints, how many requests), root cause, what worked, and one concrete fix with an owner and date. Detection time matters most: if your monitors only check the homepage, you learn about API failures from customers rather than from your own dashboards, and that lag is measured in minutes of unbudgeted downtime. Protocol-level probes on your critical paths shorten that window.

Calculating and communicating service credits without a legal team

Write the formula into your ToS once: credit = monthly fee × (downtime minutes ÷ total minutes in month) × 2, capped at 50%. Compute it from probe data, apply it automatically to the next invoice, and say so in the incident email. Even a free monitoring tier like UptimeRobot's gives you the uptime percentages to back the math.

What does SLA monitoring cost when you drop the $200/mo bill?

At 1-10 person scale, realistic SLA monitoring spend is $0-25/mo. Atlassian Statuspage runs $29-199/mo, Better Stack charges $25-85/mo, and UptimeRobot starts around $10/mo for basic checks. StatusPulse covers the same two jobs (public status page plus monitoring) from $5/mo, with a free tier of 5 probes and 1 page.

Comparing Statuspage, Better Stack, UptimeRobot, and StatusPulse on SLA-relevant features

What matters for SLA work: protocol-level probes (not just homepage pings), incident history you can export, and a public page customers trust during a 3 AM outage. StatusPulse does protocol-level checks rather than URL-only pings, which cuts false alarms on things like SMTP or database ports. UptimeRobot is cheapest but stays at URL-level checks on lower tiers. Statuspage and Better Stack are stronger on enterprise incident workflows you probably don't need yet.

Where StatusPulse falls short for SLA work

Honestly: no PagerDuty integration yet, fewer integrations overall than Better Stack, and no built-in SLA credit calculation. If you need deep on-call routing or automated breach reports, Statuspage or Better Stack still win. If you need a credible status page and real probes under $20/mo, StatusPulse does the job.

FAQ

Is 99.9% uptime the same as 43.8 minutes of downtime per month?

Close, but the exact number depends on the window length. A 30-day month has 43,200 minutes, so 0.1% is 43 minutes 12 seconds; a 30.4-day average month gives about 43.8 minutes. Put the exact allowance in your SLA so the downtime budget is unambiguous to both sides.

Do I need a formal SLA contract before I start SLA monitoring?

No. Start by setting an internal SLO, like 99.9% uptime over a rolling 30 days, and measure against it now. When a customer later asks for a contractual SLA, you already have the probe history and reports to back it. Monitoring first also tells you whether the promise you're about to sign is realistic.

Can Uptime Kuma calculate SLA uptime percentages automatically?

Not really. Uptime Kuma handles unlimited HTTP, TCP, and DNS probes and includes a public status page, but it doesn't compute rolling 30-day uptime percentages or error budget burn. You export the raw data and do the SLA math in a spreadsheet. Workable at 5 probes, tedious at 50.

How many probes do I need to measure SLA uptime accurately?

One probe per critical path is the minimum: login, a database read, the API endpoint your customers cite in refund requests. Add probes from a second region if customers hit your product from somewhere your monitoring doesn't run, since your uptime number and their experience can otherwise diverge during a dispute.

What happens legally if I miss my published SLA?

Whatever your contract says, which is usually service credits, not litigation. Write the formula into your ToS once, for example monthly fee times downtime share times two, capped at 50%, compute it from probe data, and apply it to the next invoice automatically. A documented, self-applied credit process is what prevents disputes from escalating.

Monitor it, then tell your customers

10 probes, 3 status pages, forever. No credit card.