What lowers it#
| Cause | Symptom | Lever |
|---|---|---|
| Exit type mismatched with the target | 403, CAPTCHA on first request | Move to residential or mobile |
| Per-IP rate too high | 429, blocks after N requests | Rotate more, slow down |
| Client fingerprint | Blocks regardless of exit | Fix the client |
| Pool churn | Occasional connection resets | Retry once; sticky sessions replace the exit |
| Target instability | Timeouts and 5xx from the site | Back off; not a proxy problem |
Measuring it honestly#
- Count only responses whose content is valid, not just status 200 (soft blocks hide there).
- Separate gateway errors from destination errors: the status codes tell you which side failed.
- Measure per target and per exit type; a global number hides everything useful.
How it works at ProxShift#
ProxShift does not publish a single success-rate figure because it would describe a target, not the network; instead the error codes distinguish gateway from destination failures, and retries cost only the bytes they move.