APEX FLOW WEB SCRAPER / REFERENCE LIBRARY
How to measure scraping quality
Measure field correctness, coverage, failure clarity and maintenance effort together. Speed alone can reward a collector that returns incomplete or incorrect records.
Create a reviewed reference set
Choose representative page types and manually verify the expected fields. Include static HTML, JavaScript pages, tables, feeds, changed layouts and explicit failures. Keep the reviewed dataset version with the benchmark.
Report useful denominators
Field precision asks how many returned values are correct. Recall asks how many expected values were found. Coverage asks how many eligible pages were collected. Separate access failures from parser errors and recipe failures.
Benchmark the whole job
Report cold and cached timings, p50 and p95 latency, transferred bytes, peak memory and total operating cost. Include versions, hardware, concurrency, render mode and sample size. Do not extrapolate one fast page into a general throughput claim.
Current evidence
Apex has 33 passing local tests covering extraction and important lifecycle and boundary behavior. A single server smoke run fetched example.com in about 1.55 seconds with no hosted scraping call. That is a connectivity check, not a comparative performance benchmark. A representative cross-product benchmark has not been completed.
← All reference articles · Current capabilities and limits →