Researcher

Ask for new columns, and the Researcher reads every row of your table to fill them in: a loss reason for each opportunity, say, or the company named in each note.

Who can see this: anyone who can open the project. Running it adds a step, so you need to be able to edit the project.

1. Preview table: the first 30 rows, with the new columns filled in.
2. Run chip: processes the rest of the rows.
3. Cancel: stops before anything else runs.

When to use it

Reach for it when the answer is in each row’s text but no formula can find it. It’s good for:

• sorting rows into categories you name
• judging sentiment
• pulling out names, places or other details
• writing a short summary of each row

If you don’t know the categories yet and want Querri to find them, use Categorize instead.

It works only from the data in each row, and doesn’t search the web.

Ask for it

Ask in project chat, and say exactly which columns you want and what goes in them. For example:

For each opportunity, add a loss_reason column: Price, Timing, Competitor, No Budget or Other.

You can ask for several columns at once, which is quicker than one at a time. Unless you say otherwise, values come back in title case.

Preview, then run

Nothing runs on the whole table until you say so.

1. First you see the first 30 rows in chat: “Here’s a preview of the researcher results (30 rows)”.
2. Then you pick what to do next with a chip:
   • Filter to rows with {column} first (Recommended) appears when the column you’re reading is often empty. Filtering first saves time and credits.
   • Run for all {n} rows processes the whole table, up to your plan’s limit.
   • Run for 1,000 rows (free tier) or Run for 100,000 rows (max) appears when the table is bigger than the limit.
   • Upgrade for more rows appears on the free plan when the table is over 1,000 rows.
   • Cancel stops here.
3. After that, a Researcher step does the work in the background. Your table fills in as rows finish, so a large table takes a while.

Since it runs in the background, wait for the step to finish before asking for anything that uses the new columns.

What you get

Your table comes back with the new columns added. If a row fails three times, its new columns read ERROR.

In the Data Flow panel, it’s a step labeled Researcher, like any other.

Limits

	Free plan	Paid plans
Rows per run	1,000	100,000

Rows past the limit are left empty.

Credits

Asking in project chat uses a credit for the question, as usual. On top of that, each row processed counts at 0.001 credits by default, and the fractions add up through the month. Your plan may differ; see How credits are counted on Monitor usage.

• Preview rows aren’t counted.
• Rows that end in ERROR, and rows already sent when you cancel, are counted.
• If the project runs again, on a schedule or a dashboard refresh, this step runs again and its rows are counted again.

Tips

• Drop rows with nothing to read before you run it.
• Name the allowed values. “Price, Timing, Competitor, No Budget or Other” gives consistent answers; “why did we lose” doesn’t.
• If the first 30 rows look wrong, cancel and reword the request instead of running thousands of rows.

Related

• Categorize: let Querri find the categories
• Credits
• Chat
• Projects