Matheus Oliveira

I build scrapers and data tools for public web data, and publish them here as Actors.

One habit runs through all of them: measure what the tool actually returns, then publish that number. Before publishing my first Actor I sampled 350 channels across 25 niches to find out how many of them list a reachable contact address. The answer was about a third, so that is the figure on the page, with its confidence interval, rather than a promise about "verified data in bulk".

• A measured hit rate, with sample size and confidence interval. When a number comes from a small sample, it is reported as the wide range it really is rather than rounded into a claim.
• Results you can audit. Output rows carry the source each value was read from, and the URL where that applies, so nothing has to be taken on trust.
• The weak cases named. Where something performs badly, the documentation says so. That costs me some sales and saves you a wasted run.
• Public data only. No login, no credentials handled on your behalf, and no working around protections that exist to stop automated collection.
• Priced to match. Cheap to run means cheap to buy, and no charging for results that were not delivered.

The Issues tab on the Actor is the fastest route. A run id and the input you used are usually all I need. Breakage after an upstream change gets priority.