Prospecção

From the open CNPJ data that the Federal Revenue publishes every month: legal name, trade name, CNAE, size, capital, address, and registered phone and email of every establishment in the country. It's the same source as the paid prospecting tools. The difference is what happens after the search: here each company reaches the CRM with its fit to YOUR ideal customer profile, calculated from your finance data.

CNAE accepts the full code (6201501) or just the beginning (62 = the whole software division). Combine sector + state or sector + municipality: sector alone returns tens of thousands. The search shows up to 500 companies, sorted by share capital; when the total exceeds 10 thousand, the screen says "more than 10 thousand" instead of counting — refine the filter. "Active only" is always on: a closed company is not a lead.

Check the rows (or none, to import everything the search showed) and click Import. Each company becomes a lead with source PROSPECTING and the campaign you entered; the CNPJ, the sector and the size are already stored, and the fit to the ideal profile is calculated on the spot. Whoever already exists as a lead (same CNPJ) doesn't come in twice — the screen says how many already existed.

The contact that comes from the database is the company's registered contact (often the accountant's). Treat it as a starting point, not as the decision-maker's phone. And remember the data protection law (LGPD): the data is public and the purpose is legitimate (B2B prospecting), but a requested opt-out must be respected — the CRM cadence already takes care of that.

Loading the country's 60 million establishments at once weighs on the database and takes hours, so the data comes in state by state. At the top of the screen there's the line "States in the database": it's exactly what the search sees today. Searching for a state that isn't on that line returns zero companies — not because they don't exist, but because that state hasn't been loaded yet. Whoever administers the platform adds states in the parameter "Federal Revenue database: only these states"; the new state comes in on the next load without deleting those already inside, and until it arrives it shows as "awaiting load".

Adding a state stacks up; REMOVING a state is different. Since the list no longer describes what's inside, the load goes back to being a snapshot: it deletes everything and reloads only the remaining states. It's on purpose — a list that lies about coverage is worse than a smaller database. If the intention was just to stop using a state, leave it on the list: it doesn't get in the way of the search and costs nothing beyond the space it already takes up.

The panel at the top of the screen belongs to whoever administers the platform and is always visible: it shows the states the database covers, how many companies each one brought, those that were requested and haven't arrived yet, the stage of the load in progress and the reason for the last one that failed. The button next to it starts the load right away, without waiting for the monthly job — use it after adding a state to the list, or when you want the month's snapshot before the 10th. The load runs for hours in the background: you can leave the screen, and the panel shows the stage when you come back. While it runs, the old database keeps answering searches.

The structure exists, the load doesn't: it's about 60 million rows, and loading is the decision of whoever operates the installation (it takes hours and uses 20–30 GB). Whoever administers the platform sees the "Load now" button on this screen: it downloads the month's files straight from the Federal Revenue and reloads. To reload by itself every month, turn on the parameter "Federal Revenue database: reload by itself every month"; for a small or test installation, "load only the sample" and "only these states" reduce the database to a fraction. The load log appears at the top of this screen.

The new state comes in on the next run of the load — automatic (from the 10th of each month, if reload-by-itself is on) or through the "Load now" button. It takes hours, because the Federal Revenue publishes the files by index and never by state: adding a state rereads the same 13 GB filtering for the new one. That's why the cheap path is to request all the states you intend to prospect at once, instead of one per week. While the load runs, the old database keeps answering — nothing is deleted ahead of time.