Most extraction guides tell you what should work. This one tells you what did.
Every claim here carries a date and a sample size. When something fails, the page says so — including the cases no tool can fix.
Residentialn=150
100%
Datacentern=12
58%
No proxyn=12
0%
Where the request leaves from decides almost everything. Method and failure breakdown in the reliability benchmark.
Benchmarks
Re-run and re-dated. If a number here is stale, that is a bug.
3 Sept 2026How many company websites publish a contact address?16% of 400 domains yielded an email address. Measured 3 September 2026, with failures split: dead domains, sites refusing crawlers, sites publishing nothing.2 Sept 2026YouTube transcript API reliability: measured resultsMeasured success rates for YouTube transcript extraction across residential, datacenter and unproxied egress. 150 videos, 2 September 2026.
Playbooks
One business per page, costed at what the data actually charges. The input is pennies; the work nobody else mentions is in there too.
$0.003 per comment → $2,000-6,000 per reportSell audience research built from product commentsMine YouTube comments on a product category to find what buyers complain about and ask for, then sell the finding as a research report. Costed per comment.$0.01 per domain → $0.50-1.00 per verified rowSell B2B lead lists that separate role from personalBuild verified contact lists from a domain list, with role addresses split from personal ones, placeholders filtered, and the gaps stated. Costed per domain.$0.005 per transcript → $800-2,000 per monthCompetitor video intelligence as a monthly retainerTranscribe rival channels every month, report what they push and what they stopped saying, and sell it as a retainer. Costed per transcript.$0.005 per transcript → $19-49 per guideSell study guides built from public video coursesTurn public video courses into structured study notes you can sell or use as a lead magnet, without republishing the lecturer's words. Costed per transcript.$0.005 per transcript → $8,000-20,000 build, then $400-1,200 a monthBuild a niche answer assistant on video transcriptsTurn a field's video content into a cited question-answering assistant, and sell it as a build plus a subscription. Costed per transcript.$0.005 per transcript → $150-400 per episodePodcast clip service: sell short-form from a back catalogueTranscribe a podcast back catalogue, find the quotable moments by timestamp, and sell cut-ready clip specs. Costed at the real input price per episode.
Guides
Written to be useful to someone who never buys anything. That is also what makes them worth finding.
find company contact detailsFind company contact details from a domain nameHow to get an email, phone number and social profile for a company from its domain alone — what is findable, what is not, and why many sites return nothing.role emails websiteRole emails vs personal: why lead lists need both splitWhy info@ and jane.smith@ are different kinds of data, how to tell them apart from a domain, and how to filter the placeholder addresses real companies publish.youtube shorts transcriptYouTube Shorts transcripts: the format most tools skipHow to get captions from YouTube Shorts in bulk, why most transcript tools return nothing for them, and what Shorts text is actually good for.conference talk transcriptsConference talk transcripts: read a whole trackHow to turn a conference playlist into searchable text, find the talks worth watching, and cite quotes with timestamps that link to the moment.course transcriptsCourse transcripts: turn a video course into notesHow to get the transcript of every lesson in a video course, build revision notes you can search, and use an LLM to summarise without inventing content.lecture transcriptsLecture transcripts: archive a university channelHow to extract transcripts for every lecture on a university channel, organise them per video, use them for accessibility, and build a research corpus.youtube transcript api not workingYouTube transcript API not working: how to diagnose itWorks locally, fails in production? Empty response bodies? A symptom-by-symptom diagnosis of YouTube transcript API failures and what each one means.
Why this exists
We publish and maintain data-extraction tools on Apify. Running them at volume produces measurements — success rates by environment, failure modes, and how each of those changes over time.
Most of that data is not published anywhere, so this is where it goes. The tools are linked where relevant, but the pages are written to stand up without them.