Residential proxy buying guide: compare the traffic receipt and its limits

The 41-provider residential dataset cannot support one honest market price. It contains 203 rows: 63 monthly traffic plans, 23 per-GB tiers, 17 pay-as-you-go rows, 30 explicitly unrestricted-bandwidth products, 21 concurrency-priced plans, 11 configurable offers, 10 per-IP rows, and 28 quote-only rows. Forty-four rows carry unresolved conflicts, and only 45 conflict-free rows contain an exact traffic quantity and checkout total. Those 45 rows yield 208 package observations and an observed effective range of $0.27 to $10 per GB. The low is GeoNode’s Residential monthly subscription at 100,000 GB and a $27,000 checkout. The high is ProxyScrape’s Residential, non-expiring plan at one gigabyte and a $10 checkout. GeoNode residential pricing, retrieved 2026-07-25T15:42:45Z ProxyScrape residential pricing API, retrieved 2026-07-25T16:33:08.221Z

The rows can support exact receipts for named configurations, disclosure counts, and comparisons where the billing units match. They cannot prove one residential market rate, network performance, success rate, pool quality, or the cost of a field the provider did not publish. They also cannot turn a flat-price, per-IP, concurrency, configurable, or quote-only product into a per-GB offer by guessing traffic.

What the per-GB endpoints actually buy

The cheap endpoint is not a one-gigabyte offer. It requires 100,000 GB and a five-figure monthly receipt. The expensive endpoint buys exactly one gigabyte without an ongoing commitment. Calling either one “the residential price” removes the purchase size that creates the rate.

ProxyMesh publishes a lower sticker rate of $0.05 per GB for prepaid bandwidth and an order range of 100 to 100,000 GB. Its captured tier does not publish a checkout total, so it is excluded from the checkout-backed range. Multiplying the unit rate by the first quantity would create a receipt that is not recorded as a vendor total. ProxyMesh prepaid bandwidth terms, retrieved 2026-07-25T17:42:41.156Z

Expiry changes the value of the same quantity

ProxyScrape makes traffic life measurable within one provider. Its 100 GB, 30-day configuration costs $220 before applicable tax, while its 100 GB non-expiring configuration costs $295 before applicable tax. The $75 difference, or 34.1%, buys a different validity rule, not more traffic. Both rows publish account-wide pooling, unrestricted concurrent connections, rotating and sticky access, and a 120-minute sticky-session limit. ProxyScrape residential package variants, retrieved 2026-07-25T16:33:08.221Z

Across all residential rows, 50 publish a numeric expiry window, 42 explicitly state that traffic does not expire, and 111 leave expiry null. A monthly billing period is not evidence that traffic expires monthly. Rollover on the provider’s page is also a separate field: a buyer needs both the validity window and the carry-forward rule before valuing unused traffic. Null means not published, never a permanent balance.

Pooling decides whether one large tier fits

Pooling answers where purchased traffic can be spent. Account-wide pooling can let workloads share one balance. Per-plan pooling can strand unused traffic in one plan while another runs out. Cross-type pooling matters when residential and mobile, ISP, or datacenter traffic must draw from the same purchase.

Thirty-seven residential-labeled rows span more than one proxy type, but that label alone does not prove shared bandwidth. Only 15 rows from four providers also publish account-wide, cross-type pooling: Litport, NodeMaven, ProxyMesh, and SOAX. Litport’s captured residential availability has an unresolved conflict, and two ProxyMesh rows have a separate trial-availability conflict. Those caveats stop purchase claims, but they do not convert the published pooling field to a different state.

NodeMaven’s captured monthly and pay-as-you-go rows explicitly combine residential and mobile traffic in one account-wide balance. SOAX publishes unified credits across residential and mobile. ProxyMesh publishes account-wide service across four proxy types. These are different commercial systems, so “cross-type” still does not make their credits, GB, overage on the provider’s page, or expiry rules interchangeable. NodeMaven residential and mobile pricing, retrieved 2026-07-25T17:34:28Z SOAX unified proxy pricing, retrieved 2026-07-25T15:31:32Z ProxyMesh plan terms, retrieved 2026-07-25T17:42:41.156Z

Concurrency is capacity, not traffic

Concurrency determines how many simultaneous connections a plan permits. It does not say how many addresses the buyer receives or how much traffic those connections can consume. A one-gigabyte plan with unrestricted concurrency can still exhaust after one gigabyte. A plan with a positive thread ceiling can still publish unrestricted bandwidth.

The residential rows contain 75 explicit unrestricted-concurrency statements and 10 positive numerical ceilings. The other 118 rows leave concurrency null. Storm’s captured plans publish positive ceilings from 10 to 200 simultaneous connections, while ProxyMesh publishes 100 connections per IP on two rows. Those are not equal units: one is a plan-level thread ceiling and the other is attached to each IP. Storm residential connection limits, retrieved 2026-07-25T17:42:41.156Z ProxyMesh connection limits, retrieved 2026-07-25T17:42:41.156Z

Residential disclosure ledger

Counts below use all 203 residential rows and count a provider when at least one captured residential row publishes the field. A provider can therefore appear in the published column while other plans from that provider remain null. Legacy unknown pooling values are counted as not published.

Residential disclosure calculation, computed 2026-07-26T13:00:55.971Z from evidence retrieved 2026-07-25T15:31:32Z through 2026-07-26T02:15:19Z

A defensible residential buying sequence

1. Write down a realistic traffic range and when that traffic will be used. Choose a published tier that covers the upper end without treating its unit rate as a smaller offer.
2. Keep the GB quantity, checkout total, commitment, and tax treatment beside the per-GB result. If the checkout is not published, stop before ranking it as the cheapest receipt.
3. If workloads span proxy types or subusers, require an explicit pooling value and scope. A multi-type product label with null pooling does not prove a shared balance.
4. Reject expiry windows that are shorter than the usage schedule. Treat an explicit non-expiring statement, a numeric window, and null as three different states.
5. Set a concurrency requirement in the provider’s stated unit. Unrestricted connections, a plan-level thread cap, and connections per IP are not interchangeable.
6. Only then compare exact receipts. Price per GB is useful arithmetic after these decisions, not a substitute for them.