Skip to content
Bulk jobs

Runs long enough that nobody is watching them

Can I look up emails for thousands of people at once?

Point Cuvy at a list of any size and it shows the most the run can cost before it starts, then works through it on our side, so the tab can be closed. It stops when your balance is gone rather than billing past it, keeps everything found up to that point, and never charges twice for anyone if the run has to be started again.

Shown before it starts
Maximum cost and your balance
If the balance runs out
Pauses at that person
Kept when it pauses
Everything already found
Cost of running it again
Nothing already paid for

50 addresses free every month · no card

Before it starts

The number in the preview is a ceiling, not a price

A run is priced by the people it will attempt, and that is rarely the number of rows in the file.

Before a bulk run begins, Cuvy counts the rows it would actually try. Anyone the workspace already has a confirmed address for is left out — they were paid for once and will not be charged for again. So is anyone a recent run already missed: a person who came back empty is left alone for thirty days by default, which is what stops pressing Run twice in an afternoon from asking the same question twice. Re-running a list you have worked through before can quote a small fraction of its row count.

The number that survives all that is the most the run can spend, not what it will spend. It assumes every attempt finds somebody, and no search does. You are charged for the people who come back with an address; the rest cost nothing and come back in their own file, so the gaps in a list are visible rather than implied. The preview puts your balance next to the ceiling and says plainly whether the run is going to reach it.

How a run goes

Four steps, and none of them is watching it run

  1. Gather the rows first

    A list in the web app, a CSV you have imported, or whatever a saved search matches today. Assembling the set looks nobody up and spends nothing.

    Nothing spent yet
  2. Read the ceiling, then your balance

    The preview counts the people the run would attempt, shows the most that can cost, and puts your balance beside it. If the two will meet, you are told before it begins rather than halfway down.

    Priced before it starts
  3. Start it and go and do something else

    The run happens on our side. Signing out, losing the wifi or shutting the lid has no effect on a job already going, and the job card carries the count while it works.

    Nothing to keep open
  4. Take the file, or take part of it

    Rows are written as they finish rather than at the end, so a run you have not waited for is still exportable. Confirmed addresses come out in one file and single-source ones in another.

    Exportable mid-run
When it goes wrong

Nothing about a long run should arrive as a surprise

Four things that interrupt a bulk job, and what each of them costs.

The balance runs out
It pauses on the person it could not pay for, and asks. Everything before them is already yours; nobody after them is attempted. Top up or wait for the renewal, start it again, and the remainder of the list is still sitting there.
You close the browser
The run was never in the tab. Closing it, sleeping the machine or losing the connection changes nothing about a job in progress, and the count is waiting for you when you come back.
Something breaks on our side
The job stops and says it stopped, rather than reporting a finish it did not manage. Every person it had already worked through keeps the address it found for them, because rows are written one at a time.
You start the same run again
The second attempt skips everyone already found and everyone still inside the cooling-off window. A person your workspace has been charged for is free from then on, and the database enforces that rather than trusting the run to remember.
What comes out

Three numbers from three parts of a run

74%
of the people a run attempts come back with an address

Which is why the ceiling in the preview is nearly always higher than the invoice underneath it.

3.8s
of processing for 400 people

Several thousand rows is minutes of work rather than an afternoon spent keeping a tab awake.

1
credit per person, however many runs touch them

A retry, a re-run and a colleague asking the same question all land on the charge that was already made.

Questions

What people ask before starting a long run

Is there a limit on how many rows one job can have?

Not a fixed one. The limit you meet in practice is your balance: a job that would cost more than you have still starts, works down the rows, and pauses when the balance is gone rather than refusing at the door or billing past it. The preview says in advance whether that is going to happen.

Why is the credit estimate smaller than the number of rows?

Because a run does not attempt every row. Anyone the workspace already has a confirmed address for is skipped and cannot be charged for again, and anyone a recent run missed is left alone until the cooling-off window is up — thirty days unless you change it. The estimate counts what is left.

Does the run keep going if I close the tab or shut the laptop?

Yes. A bulk job is not something the browser is doing; the tab is only where you started it. You can sign out, switch machines or lose your connection, and the job card will have the same run on it when you come back.

What happens to a job that stops halfway through?

Everything it had already done is kept. Each person is written as they are finished rather than at the end of the run, so a job that stops has cost exactly what it found and no more. Start it again and it walks the same rows, skipping everyone it has already dealt with.

What if my plan renews in the middle of a long run?

A renewal landing mid-run simply raises the balance the run is drawing on, and it carries on. A run that had already paused does not restart itself when the credits arrive — nothing spends your money without you asking it to. Start it again and it takes up the rows it never reached.

How do I see what a finished job actually spent?

The job card keeps five counts: how many people it attempted, how many came back confirmed, how many came back single-source, how many it missed, and how many it skipped as already yours. The billing screen carries one line per credit spent, saying which of the two kinds of address it bought.

Do rows that already have an email address get charged?

No. If the address came in on your CSV, a lookup that agrees with it costs nothing — the credit buys finding an address, not confirming one you brought. If the address is one Cuvy found for you before, the row is skipped and never reaches the meter.