Key Takeaways (BLUF)
Bottom Line Up Front (BLUF): With 5-minute checks, an outage can run for up to 300 seconds before the first failed check even happens. With 15-second checks, that window shrinks to 15 seconds. The difference between the two worst cases is 285 seconds, or 4 minutes and 45 seconds, during which real users hit errors and no alert exists. Add a confirmation rule such as "alert after 3 failures" and the gap grows by a full interval per extra failure, so a 5-minute monitor can take 15 minutes to page you in the worst case. Check interval is the single biggest lever on detection time, and the right value depends on what a minute of downtime costs your business.
What Happens During the 285-Second Gap
A monitor only knows your service is down when a check runs and fails. Between two checks it is blind. If your site breaks one second after a successful check on a 5-minute schedule, the next check is 299 seconds away.
That blind window is the gap this article is about. It is not a vendor claim. It follows directly from how scheduled checks work, and you can compute it yourself.
The worst case for a single-failure alert is one full check interval. For a 300-second interval that is 300 seconds. For a 15-second interval it is 15 seconds. The difference is 300 - 15 = 285 seconds.
- Worst case: the failure begins just after a passing check, so you wait one full interval for the next check.
- Best case: the failure begins just before a check, so detection is almost immediate.
- Average case: with failures landing at random times, the expected wait is half the interval.
The Detection-Time Formula
Two settings control how fast a scheduled monitor detects an outage: the check interval (I) and the number of consecutive failures required before alerting (N).
If a failure starts just after a passing check, check number 1 fails at roughly I seconds, and check number N fails at roughly N x I seconds. So:
Worst-case detection ≈ N x I. Average detection ≈ (N - 0.5) x I. Then add the check timeout and alert delivery time on top.
These numbers cover the scheduling component only. Real alerts also depend on request timeout, queueing and how fast your notification channel delivers.
15 Second Uptime Monitoring vs 5-Minute Checks: Detection Time Table
Applying the formula to common intervals gives the table below. Values are in seconds and cover the schedule component only. The plan column shows the fastest interval each WakeMetrics plan offers.
| Check interval | WakeMetrics plan with this interval | Avg detection (1 failure) | Worst case (1 failure) | Avg detection (3 failures) | Worst case (3 failures) |
|---|---|---|---|---|---|
| 5 minutes (300s) | Typical legacy free tiers | 150s | 300s | 750s | 900s |
| 3 minutes (180s) | Free | 90s | 180s | 450s | 540s |
| 1 minute (60s) | Starter | 30s | 60s | 150s | 180s |
| 30 seconds | Business | 15s | 30s | 75s | 90s |
| 15 seconds | Enterprise | 7.5s | 15s | 37.5s | 45s |
Why Confirmation Rules Multiply the Gap
Most teams add a confirmation rule to avoid false alarms: "only alert after 3 failed checks." That is a sensible instinct, because a single failed check can be a transient network event.
The cost is that every extra required failure adds one full interval to detection time. On a 5-minute monitor, a 3-failure rule means up to 15 minutes between the outage starting and a page.
This is the real tension in uptime monitoring. Short intervals detect fast but invite noise. Long confirmation windows reduce noise but delay alerts. You can see the false-positive side of this trade-off in our article on how multi-region confirmation reduces false alarms, and the single-probe weakness in why single-location uptime checks are dangerous.
The better answer is to shorten the interval and confirm quickly, rather than lengthening the interval to hide noise. A 15-second interval with 3 failures (45 seconds worst case) is still faster than a 5-minute interval with a single failure (300 seconds).
What the Gap Costs: An Illustrative Calculation
The financial effect depends entirely on your revenue per minute. The example below is illustrative, not a measurement. Replace the inputs with your own numbers.
Assume a service that earns $12,000 per hour in revenue, which is $200 per minute. The difference between a 5-minute worst case and a 15-second worst case is 285 seconds, or 4.75 minutes.
Extra undetected downtime per incident: 4.75 min x $200 = $950. If you have 12 incidents per year, that is 12 x $950 = $11,400 of avoidable exposure from detection delay alone, before counting churn, support load or SLA credits.
For a wider benchmark, Gartner has long cited an average cost of IT downtime of around $5,600 per minute (2014 estimate), and ITIC surveys report that most mid-size and large enterprises put hourly downtime above $300,000. Those figures vary a lot by industry, so treat them as ranges rather than targets.
Real Outages Show Why Minutes Matter
Public postmortems are the best evidence that detection time matters. Large providers detect incidents quickly, yet customers still depend on their own monitoring to know how it affects them.
- Fastly, June 8, 2021: a software bug triggered by a customer configuration change caused roughly 85% of the network to return errors. Fastly reported detecting the problem within a minute, and most of the network recovered in about 49 minutes. Customers with slow checks learned about it later than Fastly did.
- AWS us-east-1, October 20, 2025: a DNS resolution failure affecting the DynamoDB endpoint cascaded into many dependent services for several hours. Teams that only monitored their own application endpoint saw symptoms but not the cause.
- Cloudflare, November 18, 2025: an internal configuration issue produced widespread 5xx errors across sites that depend on its network for roughly three hours.
- The lesson: during a provider incident the first signal that matters is your own user-facing health check. A 5-minute interval can make you the last to know.
Which Check Interval Do You Actually Need?
Faster is not always necessary. Match the interval to the cost of downtime and the criticality of the service. A practical guide:
| Service type | Suggested interval | Reasoning |
|---|---|---|
| Personal blog, hobby project | 3–5 minutes | Low impact. Cost of delay is small. |
| Marketing site, documentation | 1 minute | Reputation matters, direct revenue loss is limited. |
| SaaS app, customer dashboard | 30 seconds | Users notice quickly. Support tickets begin within minutes. |
| Payments, checkout, public APIs | 15 seconds | Each minute has a direct revenue and SLA cost. |
| Internal tools | 1–3 minutes | Impact is limited to staff. Alerts can be relaxed. |
Calculate Your Own Detection Window
Use this small Python script to compute average and worst-case schedule detection time for any interval and confirmation rule. It uses the formula from earlier in this article.
# detection_window.py
# Schedule-only detection time for a polled uptime monitor.
# Does not include request timeout or alert delivery latency.
def detection_window(interval_s: float, failures_required: int = 1):
worst = failures_required * interval_s
average = (failures_required - 0.5) * interval_s
return average, worst
scenarios = [
("5 min, 1 failure", 300, 1),
("5 min, 3 failures", 300, 3),
("1 min, 3 failures", 60, 3),
("30 s, 3 failures", 30, 3),
("15 s, 3 failures", 15, 3),
]
for name, interval, n in scenarios:
avg, worst = detection_window(interval, n)
print(f"{name:20} avg={avg:7.1f}s worst={worst:7.1f}s")
# Cost of the gap between two worst cases (example inputs)
revenue_per_minute = 200 # replace with your own number
gap_seconds = 300 - 15 # 285
cost = gap_seconds / 60 * revenue_per_minute
print(f"Gap = {gap_seconds}s -> ~$" + f"{cost:,.0f} per incident")Common Mistakes When Choosing an Interval
- Judging only the interval and ignoring confirmation. A 1-minute interval with 5 required failures is slower than a 30-second interval with 3.
- Alerting on the first failure from a single location. Fast checks from one probe produce more noise. Confirm from more than one location before paging a human.
- Setting a long timeout. A 30-second request timeout delays failure detection by up to 30 seconds on every check.
- Monitoring only the homepage. A cached homepage can return 200 while the API behind it is failing. Check a path that exercises the real dependency.
- Ignoring alert delivery. A fast detection followed by an email that lands 5 minutes later erases the benefit. Use a channel with fast delivery such as Slack, Telegram or a webhook.
Further Reading in the WakeMetrics SRE Series
These articles cover the other parts of detection speed and alert quality:
- How to Calculate Mean Time to Detect (MTTD) and Slash It by 70% — measuring detection speed end to end: /blog/how-to-reduce-mean-time-to-detect-mttd
- Why Single-Location Uptime Checks Are Dangerous — why one probe cannot tell a global outage from a local network problem: /blog/why-single-location-uptime-checks-are-dangerous
- How 3-Stage Consensus Eliminates False Positive Downtime Alerts — confirming failures before paging: /blog/how-3-stage-consensus-eliminates-false-alarms
- The Math Behind Alert Fatigue — what false alarms cost on-call teams: /blog/cost-of-alert-fatigue-sre
- Compare tools: observability and uptime monitoring alternatives at /alternatives
References
Google SRE Book, Monitoring Distributed Systems: https://sre.google/sre-book/monitoring-distributed-systems/
Fastly, Summary of June 8 outage: https://www.fastly.com/blog/summary-of-june-8-outage
AWS, Summary of the Amazon DynamoDB service disruption in the Northern Virginia (us-east-1) Region: https://aws.amazon.com/message/101925/
Cloudflare, Cloudflare outage on November 18, 2025: https://blog.cloudflare.com/18-november-2025-outage/
Frequently Asked Questions
Is 15-second uptime monitoring worth it compared with 5-minute checks?
It depends on the cost of downtime. A 5-minute interval can leave an outage undetected for up to 300 seconds, while a 15-second interval limits that to 15 seconds, a difference of 285 seconds. For payment flows, public APIs and customer-facing SaaS where each minute costs real revenue or SLA credits, the faster interval usually pays for itself. For a personal blog or an internal tool, 3 to 5 minutes is normally enough.
How long does it take an uptime monitor to detect an outage?
Roughly the check interval multiplied by the number of failures required before alerting, plus request timeout and notification delivery. With a 5-minute interval and a single-failure rule, expect up to 5 minutes. With a 3-failure rule, up to 15 minutes. With a 15-second interval and the same 3-failure rule, up to 45 seconds. Average detection time is about half an interval shorter than the worst case for the first failure.
Do shorter check intervals cause more false alarms?
They can, if every single failed check pages someone. A short interval samples the network more often, so transient events such as a brief routing change are more likely to be observed. The fix is not a longer interval. It is requiring confirmation, ideally from more than one location, before sending an alert, so short blips are filtered without slowing real detection much.
What check interval does WakeMetrics offer?
WakeMetrics plans offer 3-minute checks on Free, 1-minute on Starter, 30-second on Business and 15-second on Enterprise. You can compare plans on the pricing page and start a 14-day free trial of the Business plan at /signup?plan=business.
Conclusion
The gap between 5-minute and 15-second monitoring is not marketing. It is arithmetic: 300 seconds of blindness against 15, a difference of 285 seconds, multiplied further by every confirmation check you require. Pick an interval from your own cost of downtime, confirm failures quickly instead of slowly, and measure your real detection window. To test faster checks on your own services, start a 14-day free trial at /signup?plan=business.