Business

What is Success Rate?

Success rate is the share of requests through a proxy that return the content you asked for. Teams count clean 2xx responses carrying real page data as successes, and treat blocks, CAPTCHA pages, timeouts and connection resets as failures, including the ones that arrive with a 200 status. The figure varies enormously by destination, so one global number says little.

The hard part is classification, not counting. HTTP 403 and 429 are unambiguous refusals, while a 200 is not automatically a win, since challenge pages, consent walls and empty result sets all arrive with a success status. A scraper that inspects only the status code will report excellent numbers while collecting nothing usable. Below HTTP, count TCP resets, TLS handshake failures and timeouts as failures too. A 407 means the proxy rejected your credentials, which is a configuration fault rather than a target or network problem, so give it its own bucket.

Several things move the number, and IP type is only one of them. Rotation policy has a large effect, since rotating on every request breaks sites that expect a stable session, while holding one address through an aggressive crawl gets that address flagged. Header order and TLS fingerprint decide outcomes on destinations that fingerprint clients, independently of where the traffic comes from. Geographic consistency counts as well, because an account created in Germany that suddenly appears from a Brazilian mobile IP raises the risk score before the page even loads.

Quote two numbers instead of one. First-attempt success rate tells you how well the setup is tuned; success rate after retries tells you what the pipeline delivers. Vendors normally publish the second, which can look strong even when every third request needs two more attempts. Retries are not free under per-GB pricing, because a blocked response still carries a full TLS handshake and a block page, and those bytes are billed like any others. Track cost per successful record alongside the percentage.

Bucket the metric by destination domain, country and hour before drawing any conclusion from it. An aggregate across a dozen targets hides the single hardened site that is failing outright, which is usually the one you care about. Sample size deserves attention too, since a rate computed over a couple of hundred requests swings several points on noise alone. PROXIES.SX publishes 92% success on major platforms; like every vendor figure, it describes their test conditions, so measure against your own targets before committing volume.

Where you meet it

Success rate is the number your job runs on. It shows up when yesterday's crawl came back thin, or in a bug report saying the parser found no rows. The same number drives the decision about whether to change IP type or fix your request headers. Classify each failure at the moment it happens and store the reason with the record. A bare percentage tells you something is wrong without telling you which of those two repairs to make.

Common questions

What counts as a failed request?

Anything that does not return the content you asked for: explicit blocks such as 403 and 429, CAPTCHA or interstitial pages served with a 200, timeouts, TCP resets, gateway 502 and 504 responses, and truncated or empty payloads. Proxy authentication errors are normally tracked apart, since they indicate a configuration problem rather than a block.

Why is my success rate lower than a provider advertises?

Vendor figures come from their own targets, rotation settings, concurrency and request timing. On many sites, TLS fingerprint, header order, geographic matching and request rate move the outcome more than the IP source does. Compare providers by running your own code path against your own destinations at the same time of day.

What is a good success rate for mobile proxies?

It depends entirely on the destination. Lightly protected sites sit in the high nineties for most setups, while platforms with device fingerprinting and behavioural scoring run far lower for everyone. Benchmark against the same target, same code and same hour rather than against a general industry figure.

Related terms

Uptime
Uptime is the share of a measured period during which a service is reachable and answering correctly, written as a percentage such as 99.9%. Providers report it monthly or quarterly from their own probes. The complement, downtime, is easier to reason about, since 95% uptime allows roughly 36 hours of outage across a 30-day month.
Bot Detection
Bot detection is the practice of separating automated traffic from human traffic. Systems score each request against network signals, transport-layer fingerprints, browser characteristics and behaviour over time, then allow, challenge or block it. Vendors including Cloudflare, DataDome, Akamai and HUMAN sell this as a service running in front of the origin server.
Bandwidth Pricing
Bandwidth pricing charges for the data a proxy moves, metered per gigabyte, instead of charging per endpoint or per seat. Mobile and residential networks use it because the provider pays real carrier data costs. Rates usually step down as committed volume rises, and providers differ on what counts as billable data.
Port Pricing
Port pricing charges a fixed fee for each proxy endpoint - a port, a modem or a dedicated IP - usually monthly and regardless of how much data passes through it. Dedicated mobile proxy sellers use it because every port is a physical SIM and modem they keep powered and rotating.
SLA
A service level agreement is the contract text that states measurable service targets and what the provider owes when it misses them. It names the indicator, the target value, the measurement window and the remedy, normally service credits. A percentage published without a measurement method or a credit schedule is a marketing claim rather than a commitment.

Real 4G/5G mobile and residential IPs

PROXIES.SX runs carrier IPs in 100+ countries with HTTP and SOCKS5 on every endpoint. $4/GB down to $2.40/GB at volume, free endpoints and rotation, and your GB never expire.