Skip to content

Downtime calculator:
how bad were those 43 minutes?

Type what happened — we grade it against the nines, price it in your own numbers, and show where the minutes actually went. The other direction from our uptime calculator, which starts at the SLA.

over the last
The damage table

A month of downtime, graded

Same math as the SLA ladder, read backwards: start from the minutes you lost and see which targets survived them.

Downtime in one month → uptime grade

month = 30.42 days · 43,800 min
Downtime / monthFeels likeUptimeGradeStrictest target met
1 minute one bad deploy, caught fast 99.9977% four nines 99.99%
4m 22s the entire four-nines budget 99.99% four nines 99.99%
43m 48s the entire three-nines budget 99.9% three nines 99.9%
3h 39m one bad afternoon 99.5% two nines 99.5%
7h 18m a recurring problem 99% two nines 99%
24 hours an incident with a name 96.7123% one nine none of them
The method

Turning incidents into a downtime number

Three rules decide the number:

  • Sum, don’t average. Downtime for the window is the sum of every incident’s duration — three short outages are not “mostly up”.
  • Measure edge to edge. An incident starts at the first failed request, not the first alert — and ends when service is verified, not when the fix ships.
  • Decide what “down” means — in writing. Full outage, checkout broken, or just slow? Pick the definition before the incident, or the argument happens after it.

Then it’s one division: downtime ÷ window. The ledger on the right is June for a small shop — the calculator above is that last row, graded.

One month, fully logged

caldmont.com · June
Jun 4 · 03:12–03:24Host reboot after kernel patch — nobody noticed at 3 a.m. Still counts.12m 0s
Jun 18 · 14:41–14:45Deploy rollback. Short, but peak hours — the support inbox noticed.4m 0s
Jun 30 · 11:04–11:47The big one: a config change broke checkout at peak.43m 48s
June total59m 48s of 43,800 minutes — summed, not averaged.99.864%
Note the grade: three incidents and June already misses three nines — the monthly budget was 43m 48s in total.
MTTRMTTR = total downtime ÷ number of incidents
MTBFMTBF = total time up ÷ number of incidents
Availability from bothavailability = MTBF ÷ (MTBF + MTTR)
Cost of an outagecost ≈ hours down × revenue/h + people × hours × loaded rate
FAQ

Common downtime questions

Two directions share one formula. Start from an SLA percentage and you get the downtime it allows (that’s our uptime calculator). Start from the downtime you actually had, as this page does, and you get the uptime percentage it produced: (window − downtime) ÷ window × 100. We add the two numbers a spreadsheet won’t give you: what it cost, and where the minutes went.

Divide downtime by the window, times 100. A 43m 48s outage in a 30.42-day month: 2,628 s ÷ 2,628,000 s × 100 = 0.1% downtime — that is, 99.9% uptime. Two rules: sum every incident in the window (don’t average them away), and measure from first failed request to verified recovery.

Two to three nines is the normal band for small-to-mid production services — 43 minutes to 7 hours per month. Three nines (43m 48s/month) is the workhorse target: reachable with good hosting, fast rollbacks and someone on call. If your month runs past 7 hours, the damage table above shows the grade you’re actually delivering — and the timeline section of the result shows where to claw minutes back first.

Skip the headline per-minute figures from enterprise surveys — they average banks into bakeries. Your ceiling is your own arithmetic: revenue per hour × hours down, plus the people who dropped everything, plus anything contractual. The cost section of the result runs it live with your numbers. Two caveats. Some interrupted buyers return later, so lost revenue is a ceiling. And some costs never fit the spreadsheet at all: trust, SEO, the support backlog.

They count however you decided they count — before the incident. Common practice: a broken critical path (checkout, login) counts in full even if the homepage renders; degraded-but-working counts separately or not at all. Write the definition down and apply it the same way every time. SLA arguments are about this paragraph, not about the arithmetic.

MTTR is mean time to recovery: total downtime ÷ number of incidents. It is your best lever, because halving it halves your downtime without preventing a single incident. MTBF is mean time between failures, or how often things break. It lives in architecture and change discipline, not response speed. Together they give availability: MTBF ÷ (MTBF + MTTR). RTO is the MTTR you promised — how long an outage may last before the recovery plan has failed. Test it before an incident does.

Attack MTTR before MTBF — recovering faster is cheaper than failing less. And within MTTR, attack detection first: it’s the only phase a tool can shrink for you outright. In our worked example, detection took 11 minutes and the fix took 4. Thirty-second checks cap the silence before the first failed check at half a minute — they can’t shorten alert delivery or the time someone takes to pick up the phone, which is why detection shrinks by four minutes here, not eleven. Diagnosis shrinks with evidence (multi-location results, exact timestamps); the fix itself shrinks with rehearsed rollbacks — that part’s on you.

You can’t calculate what you didn’t notice — the 3 a.m. reboot in our June ledger only exists because something was watching. Independent monitoring gives you the inputs this calculator needs: exact start and end times, from outside your own infrastructure. That covers every incident, including the ones nobody was awake for. Uptimia checks from 171+ locations in 70+ countries — every 60 seconds on the Basic plan, every 30 from Professional up — and keeps that ledger for you.

99.9% — THREE NINES · PER MONTH

43m 48s down = 99.9%.

Over one month, 43m 48s of downtime leaves 99.9% uptime — three nines. Below: the same minutes against four common SLA targets, what they cost in your numbers, and where they went.

Your outage — edit anytime, the whole page follows
over the last
Get the exact minutes next time uptimia.com/downtime-calculator?down=2628s&window=month
1 · The grade

43m 48s, judged against four common targets

Budgets are per window and sum across incidents — this verdict assumes these were your only minutes lost.

Verdict against four common targets

same downtime, four contracts
99.99% per monthallowance 4m 22s — over by 39m 25sbreached
99.95% per monthallowance 21m 54s — over by 21m 54sbreached
99.9% per monthallowance 43m 48s — 0ms left overmet — barely
99.5% per monthallowance 3h 39m 0s — 2h 55m 12s left overmet

The formula, with your numbers

downtime percentage
1 · the windowmonth = 2,628,000 s
2 · the downtime2,628 s = 43m 48s → 0.1% of the window
3 · the uptime(2,628,000 − 2,628) ÷ 2,628,000 × 100 = 99.9%

Convention: 365-day year, month = year ÷ 12 (30.42 days) — same as our uptime calculator and the nines ladder.

2 · What it cost

Your 43m 48s€14,037 — priced with your numbers, not a survey

Three inputs decide the bill. Edit them here — the total, the rows and the nav above update as you type.

€/h at the time of the outage
× €95/h loaded, incident + cleanup
€ that bought a dead page
€14,037 €320/min of outage

The itemized bill

43m 48s = 0.73 h
Lost revenue€18,400/h × 0.73 h. A ceiling, not a forecast — some buyers return later, some finish their cart at a competitor.€13,432
Ads buying a dead pageCampaigns don’t pause because the site did. The budget kept spending through the outage.€50
The incident team4 people × 0.73 h at €95/h — doubled, because postmortem, cleanup and apologies take at least as long as the outage.€555
Hosting SLA creditWhatever your host’s contract pays back for the breach — usually a rounding error. The worked example shows why.your contract
Trust, SEO, support backlogReal, delayed, and unpriceable — which is why they’re listed, not estimated.unpriced
Total, before the unpricedThis is one outage. Sum every incident in the window for the real total.€14,037
Worked example — not your numbers

One 43m 48s outage, and the €4.90 credit

1 · the breachHost promises 99.95% monthly (24 × 7). June delivered 99.864% — breached.
2 · the tierContract remedy for 99.5–99.95%: 10% service credit on the monthly fee.
3 · the feeHosting is €49/month. 10% of that is the entire payout: −€4.90 against a ~€14,000 loss.
4 · the catchCredits are claimed, not paid: file within 30 days, with evidence from a source that isn’t the provider you’re claiming against — an independent monitor’s incident log is exactly that.

SLA credits price the provider’s risk, not your loss. The useful question isn’t “what does the SLA pay?” but “how fast do we find out?”

3 · Where the minutes went

11m 0s of your outage was likely pure detection lag

The typical incident splits 25% detection · 39% diagnosis · 9% fix · 27% recovery. Applied to your 43m 48s — detection is the slice a tool deletes outright.

Your 43m 48s, cut by phase

typical split, drawn to scale
Detection11m 0sNobody knows yet. The only phase a tool shrinks outright — 30-second checks cap the silence before the first failed check at half a minute.
Diagnosis17m 0s“What changed?” Shrinks with evidence: multi-location results and exact timestamps warm the trail.
The fix4m 0sUsually the shortest slice — if rollback is one rehearsed command. This part is on you.
Recovery11m 48sEnds when the service is verified from outside — a timestamp, not a feeling.

At your bill above, the detection slice alone ≈ €3,525. Monitoring can’t fix your deploy — it deletes this slice.

Worked example — a real replay

How one 43m 48s outage actually played out

11:04:00checkout breaks (config push 11:03) 11:08:40first failed check — 5-min interval 11:09:40confirmed from 2nd location · alert sent 11:15:00engineer engaged — detection done 11:32:00root cause: the 11:03 config change 11:36:00rollback deployed — fix done 11:47:48all checks green — incident closed

With 30-second checks the 11:08:40 line becomes 11:04:30: the silence before the first failed check drops from 4m 40s to 30 seconds, and every line below it moves up by those four minutes.

Free tools are just the start.
Uptimia keeps your sites healthy.

Uptime, SSL, domain expiry, page speed, transactions — monitored from 171+ locations worldwide. Free for 30 days.

30 days free no credit card cancel anytime free plan after trial
100,000+ websites monitored · GDPR-compliant