IP Reputation

IP reputation is the composite score that third-party intelligence services attach to an address based on its history and characteristics. IPQualityScore, MaxMind, Spamhaus and Cloudflare's own internal dataset all maintain versions of this, and defenders query them at the edge before a request reaches an application. IPQualityScore's proxy detection API returns `proxy`, `vpn`, `tor`, `bot_status` and `recent_abuse` flags alongside a numeric `fraud_score`, plus the ASN and a connection type such as residential or mobile (documentation read 2026-07-27). That field list is a fair map of what a defender is actually weighing.

Some inputs are yours to control. Request volume, how concentrated your traffic is on one target, whether you hammer a single exit, and whether the pool you rent is shared with tenants doing card testing or credential stuffing. Others are not. You inherit whatever the previous lessee of that address did, and on residential supply you inherit whatever the household behind it does with its own connection. There is no reset button and no appeal process. A poisoned exit is simply a bad draw.

On rotating residential pools, reputation is a shared resource. Your traffic and every other customer's traffic accumulate against the same finite set of addresses, so the exit you get handed at 3pm carries the afternoon's aggregate behavior from strangers. That is the structural argument for curated or filtered supply, and vendors sell against it: NodeMaven's site advertises "30M+ real filtered IPs" and a "95%+ clean IP rate", claiming every IP is checked before assignment (read 2026-07-27). Treat that as a vendor claim, not a measurement. The number to trust is the one your own crawl produces.

Which points at the practical build. Put a pre-flight reputation check inside the crawl rather than in a spreadsheet: on session start, resolve the exit IP, score it against whichever intelligence API you already pay for, and discard the exit before it touches the target if the score is bad. Cache results so you are not paying per lookup on repeat draws, and log the discard rate per vendor. That discard rate becomes the comparison metric across suppliers, and it is worth far more at renewal than any clean-rate figure on a pricing page. The premium for filtered supply is worth paying exactly when your own discard rate says the cheap pool is burning targets.