Set render_js=false explicitly whenever you scrape a static page, a PDF, or a file. The parameter defaults to true and each rendered request costs 5 credits, so a bulk job that never needed JavaScript burns budget for nothing. URL-encode the target URL strictly before calling the API: unencoded characters are the most common cause of failed requests and junk responses. Keep the error map handy: 401 is an invalid API key, 403 is a bad key or blocked account, 429 means you are over your rate limit, 500 is a ScrapingBee-side failure worth retrying. For ordinary targets use premium_proxy=true only when the target blocks datacenter IPs; it is not needed for every request.

Context: Web (proxlane ScrapingBee migration guide): documents two cost gotchas that hit migrators: render_js defaults to true, which bills 5 credits per request, and URL encoding is strict, reportedly the number one support issue for malformed requests.