SDK sample rate calculator

Given event volume and a monthly budget, the SDK sample rate that fits — and the unique bugs you would silently lose.

Per-event error tracking pricing forces a tradeoff: drop your tracesSampleRate low enough to fit your budget, or watch the bill explode on a bad deploy. This Sentry sample rate calculator (also works for Bugsnag, Rollbar, and Datadog Error Tracking) shows you the sample rate that fits your error tracking budget — and estimates how many unique bugs sampling would silently drop before they reach your inbox.

Accepts 1,000,000, 1m, or 250k.
$
Most teams see 100–500 unique fingerprints per million events; we use this to estimate how many distinct bugs sampling would silently miss.
Recommended sample rate
0.0800

8.00% of events captured. About 6591 out of every 10,000 unique bugs would be sampled out before reaching you (≈ 659 of your 1000 estimated unique bugs).

Cost breakdown
At full capture (1.0)
$2,500/mo
At recommended sample rate
$200/mo (within budget)
Spike day: bad deploy, 10× volume for 6 hours
Month total at 1.0
$2,688
Month total at recommended rate
$215

Even with sampling, the spike pushes you $15 over budget — this is the trap with per-event pricing.

Compared to flat-rate

GlitchReplay's Pro plan is $49/mo for unlimited events. At your volume you'd save $151/mo and capture 100% — with flat-rate pricing you'd never need to sample.

The math behind this
max_events_in_budget = budget / price_per_event
sample_rate          = min(1, max_events_in_budget / monthly_events)

events_in_millions   = monthly_events / 1,000,000
total_unique_bugs    = unique_bugs_per_million * events_in_millions

# probability a unique bug (~5 occurrences) is missed
# entirely at sample rate s — every occurrence must be dropped:
P(missed | s)        = (1 - s)^5

unique_bugs_lost     = total_unique_bugs * P(missed | sample_rate)

# spike day: 10x volume for 6 of 720 hours in the month
extra_events         = (monthly_events / 720) * 6 * 9
spike_total_full     = (monthly_events + extra_events) * price_per_event
spike_total_sampled  = spike_total_full * sample_rate

The 5-occurrence assumption is conservative. Long-tail bugs that hit once or twice a month are dropped at much higher rates: at sample rate 0.1 a single-occurrence bug has a 90% chance of being missed entirely.

Per-event prices are approximate list prices and vary by tier, region, and contract. Use this for ballpark planning, not procurement.

Want this on every error automatically?

GlitchReplay does this on every event you capture. Sentry-SDK compatible, flat-rate pricing, session replay included — built on Cloudflare so a bad deploy will never blow up your bill.