From CSV to Answer in 10 Minutes

~10 minutes · your CSV, a landed table, a first result

This is the quick version. You bring one CSV, and by the end you’ll have imported it, landed it as a table, run a single transform, and read a real result off your own data.

It’s the high-velocity companion to Load, Transform, and Publish Data — that one takes about an hour and builds a full multi-step pipeline with joins and a publish. Come here first for the win; go there when you want the depth.

Prerequisites

• A PlaidCloud workspace (start a free trial if you don’t have one).
• A project to work in — create one from the Projects tab if you don’t have one yet.
• One CSV file of your own. Anything with a header row and a numeric column works — sales, expenses, a list of orders.

Step 1: Create a Workflow

A workflow is the pipeline that imports and transforms your data.

1. Open your project and switch to the Workflows tab.
2. Click New Workflow in the toolbar.
3. Give it a name like “CSV Quick Start” and leave Workflow Type on its default, Standard Serial.
4. Click Create, then double-click the new workflow to open the Workflow Explorer.

Step 2: Import Your CSV

1. Add a new step and choose an Import → CSV step (or whichever import step matches your file’s format).
2. In the step configuration:
   • Source file — point at your CSV: upload it, or pick it from a connected document account.
   • Target table — name it my_data.
   • Delimiter, quote character, header row — adjust only if your file is non-standard; the defaults handle a typical comma-separated file with a header row.
3. Run the step.

Your CSV now lands as a table in the project. A table holds the physical row-and-column data; that’s what everything downstream reads from.

The source is never touched. Each step reads from a source and writes to a target, leaving the original alone. Your CSV import wrote my_data; the transform in the next step reads it and writes somewhere new.

Step 3: Check the Landed Table

Switch to the Tables tab and click into my_data. Confirm the column count, a few sample rows, and that the data types look right — a numeric column should read as a number, not text. If a type is off, you can correct it with the column mappers in the import step and run it again.

Step 4: Run One Transform

Now turn the raw table into an answer. This example filters to the rows that matter and keeps only the columns you care about — adjust the specifics to your file.

1. Add a Tables → Table Lookup step (or any table transform that lets you filter and select columns).
2. Configure:
   • Source table — my_data
   • Target table — my_result
   • Filter conditions — a condition that trims the rows, e.g. amount > 0
   • Columns to keep — the subset you want to read, plus any computed column such as revenue = quantity * price
3. Run the step.

For the full set of functions you can use in a computed column — string operations, date math, conditional logic — see the Expressions reference.

What you should see. After the step runs, my_result holds fewer rows than my_data — the filter trimmed them — and your computed column (such as revenue) is present and populated on every remaining row. If the row count didn’t drop or the computed column is empty, recheck the filter condition and the expression before moving on.

Step 5: Read the Result

Open the Tables tab and click into my_result. Fewer rows than my_data, only the columns you kept, and your computed column filled in — that’s your first answer, straight off your own data.

You can also run both steps at once: back in the workflow, click Run (or Run All from the Actions menu) to execute the whole thing in order. Each step shows a status icon as it goes — Completed means it finished cleanly. See Run a Workflow for what each status means and how to resume a stopped run.

Hit an error? The two usual culprits are a data type that imported as text when you needed a number, and a filter condition that references a column name that isn’t in the table. Check the table from Step 3, then re-run. Managing step errors covers the rest.

Where to Go Next

• Load, Transform, and Publish Data — the hour-long version: import two files, join them, compute columns, and publish the result for downstream consumers.
• Build a Dashboard — turn a table like my_result into a chart others can read.