Proxy cost per successful request: calculate the total

Define business success before calculating cost per successful request. Divide reconciled proxy, compute, recovery, and labor costs by distinct outputs that pass that definition.

Count duplicate retries once; report zero-output runs without a unit cost. The result measures operating cost, not outcome value or profit.

Define the unit the buyer accepts

Before comparing routes, choose one unit: a product record with required fields, a completed profile task, or a verified market observation. Document its validator and scope.

A syntactically valid response has not completed the job if it contains a challenge page, wrong market, stale item, or missing field.

HTTP 200 means the request succeeded at the HTTP protocol level. It does not validate a parsed record. A 429 signals rate limiting; follow the target’s instructions, including Retry-After when present, and do not use retries to ignore access controls (RFC 6585).

Use separate counters:

The output unit is not automatically the business outcome. A valid price record may still be too old for a repricing decision; a completed profile setup does not establish campaign performance. Track the downstream outcome separately when it matters.

Build a fully loaded run cost

For a fixed observation window, use:

cost per accepted output = total run cost / distinct accepted outputs

total run cost = proxy charge assigned to the run
               + compute and browser runtime
               + storage and delivery
               + operator recovery and support labor

Reconcile consumed traffic with the provider’s usage record. Assign the run’s proxy charge using current billing terms and a consistent finance rule. Check the current agreement; provider rules differ.

For prepaid packages, separate the run’s charge from remaining traffic. Record unused, valid balance as inventory instead of charging it to every run. For expired or unusable traffic, record the treatment and period.

Record diagnosis, recovery, data repair, and handoff hours. Multiply them by finance’s internal rate; without a defensible rate, report hours separately.

Keep planned fixed platform costs separate from per-run costs, or allocate them consistently across the same period.

If no output passes, the ratio is undefined. Report the cost, output count, and failure categories. A blank denominator is not zero cost, and a cost-per-output figure alone is not a measure of quality.

Compare routes on the same work

Use the same permitted target cohort. Hold the parser, acceptance rule, session requirements, geography, browser or HTTP client, concurrency, and retry policy constant.

Randomize run order where practical. Cover normal workload variation and report dates and sample size. A changed validator or target mix starts a new baseline.

Keep one row per attempt, then summarize by route and run:

Separate proxy failures from target throttling, login state, parser drift, stale inventory, and unsuitable acceptance rules. These can change output without a route change; do not bury them in one “success rate.”

For recurring crawls, Crawlee’s scaling guide covers queue and concurrency decisions that affect throughput. It does not establish which proxy route performs better.

Reproduce the arithmetic

The following invented values are a worksheet example, not observed provider prices or customer results. Assume both routes process the same permitted cohort, then substitute charges and time measured in a real test.

Under these assumptions, the second route costs more overall but less per accepted output, delivering 1,500 more accepted outputs. Per-client cost answers a different question.

Freshness, completeness, downstream use, and the buyer’s actual contribution determine whether that output justifies the run. The example establishes neither savings nor a preferred provider.

For media or agency work, use completed profile tasks, client handoffs, or accepted campaign checks if the buyer has defined completion. Profile count alone does not prove campaign value.

If finance has measured contribution per completed business outcome, estimate break-even separately:

break-even completed outcomes = incremental run cost / contribution per outcome

Do not use accepted records or proxy requests in that denominator unless they are the business outcome being valued. For sensitivity, recalculate with low, expected, and high accepted-output counts and recovery hours from the same workload. Label forecasts separately from observed runs.

The three-client row assumes equal allocation in this invented example. For unequal workloads, allocate shared spend by declared metered bytes, runtime, or operator hours; keep client-specific costs separate. Never divide equally when one client generates most retries or support.

Opportunity cost is a separate decision: what useful work would the same budget or operator capacity support elsewhere? Compare that alternative explicitly, without adding the same paid labor twice to the run-cost numerator.

Decide whether the workload is ready to scale

Before increasing concurrency, clients, profiles, or markets, answer four questions:

1. Is the result useful? The owner and operator agree on fields, freshness, duplicates, and recovery responsibilities.
2. Is the comparison fair? Target, geography, session, parser, and retry conditions are fixed and permitted.
3. Can the team absorb the work? Operator hours, queue delays, and recovery steps fit the intended workload.
4. Does the next stage make business sense? Compare expected accepted output and downstream value with total operating cost. Treat any value estimate as a forecast until observed.

If the route meets network requirements but validation fails, fix the data or workflow before buying traffic. Revisit the result definition if the buyer cannot use growing output. When labor dominates, changing routes may not solve the bottleneck.

Use the proxy bandwidth guide to forecast traffic before a run. Use this article to evaluate recorded accepted work afterward.

In a purchasing brief, list workload, required geography and session behavior, maximum spend, acceptance rule, and failed-attempt billing. Verify commercial terms against the current provider agreement.

For product-fit questions, contact ProxyLane support. Use test results to support the purchasing decision.

ProxyLaneTraffic never expires

Premium residential proxies
From $2/GB

Scale your workflows with the tools you already use. Rotating or sticky sessions up to 72 hours.

Try 350 MB for $1.95 View plans