Request an Actor
If you need structured data from a site that is not covered yet, describe the data rather than the scraper. What comes back first is a scope with its limits written down.
What to send
Four things, and the fourth is the one people forget:
- Two or three real URLs showing the pages the data lives on. Not the home page — the actual listing, profile or search result.
- The fields you need, named. “Everything” is a slower answer than a list of twelve, because it turns into a conversation about which of the forty available fields you were actually going to use.
- Volume and frequency. Ten thousand rows once is a different piece of software from five hundred rows every morning.
- What breaks if a field is empty on 40% of rows. This decides whether the job is possible at all, and it is almost never asked upfront.
Send it to hello@actorstack.dev.
What you get back
A scope: which fields are reachable on the public surface, which are not and why, the fill rates I would expect from a quick sample, and the cost per thousand rows. If a field you asked for is behind a login or too sparse to promise, that is in the scope rather than discovered after the invoice.
The four reasons a request gets declined
- It needs an account. These Actors do not sign in and do not run on borrowed session cookies. If the data is only visible to a logged-in user, I am not the right person.
- It needs the site's own rules broken. No CAPTCHA solving, no TLS or browser fingerprint spoofing, and nothing a target's robots.txt asks crawlers not to request.
- It is personal data I could not defend collecting. The clearest example is already public: private sellers' phone numbers are not returned by the coches.net Actor and there is no setting to change it.
- The data is not actually there. Sometimes a field exists on 15% of pages and the pipeline you are imagining needs it on all of them. Better to find that out in the scope.
How new Actors are priced
Published Actors are priced per delivered result, so you pay for rows rather than for compute time and a run that finds nothing costs essentially nothing. Custom work is quoted separately; if the target is likely to be useful to more than one person, publishing it publicly is usually cheaper for you than commissioning it privately.