Run a process
A process can be run in two ways: manually, when you import the files and trigger processing immediately, or automatically, when the platform loads and processes the data on its own—by schedule, from another process, or through an external API trigger. This article covers both methods.
Complete flow: you can only run a complete data preparation flow, that is, one with no errors and at least one data source step and one output step.
Manual execution
The Run button opens the process execution screen, where you import the files that will go through the flow for a specific reference date. At the end, the platform loads and processes the files and takes you to the results view.
On the import screen, you need to:
set a reference date, the date on which the process results are stored;
select the files to be processed by choosing them on your computer or dragging them to the upload area;
associate the files with the flow's data sources, manually or by identification rule, if there are no automatic identification rules.
Dates with results: on the reference date calendar, days that already have results are marked with a green indicator. You can browse the calendar and view results for different dates.
Import type
Still on the execution screen, you choose how to import the data. The options appear when the selected reference date already has an import and vary according to the analysis type.
Replace data is available in Turbo AI and Standard analyses and is the default option. It deletes all imported data and previous analyses for the date and replaces them with the newly imported data.
Update execution is available in Turbo AI analysis and applies to Table Analysis and Reconciliation Analysis steps. It lets you update the data from an execution without deleting the entire history, using a row identifier when configured. With the identifier, the platform compares the existing data with the data from the new import:
records with the same key and the same data remain unchanged and are not duplicated;
records with a new key are added as new records;
records with the same key but different data have the existing record updated;
old records whose key does not appear in the new import are not changed and remain in the history.
Without a row identifier, the platform cannot compare records between executions. In this case, all records from the new import are added incrementally, which can lead to duplicates because the old data is preserved.
Increment data is available in Standard analysis. It keeps previously imported data and adds the new data. Incremental behavior applies to data input, not to analyses: if you import data and an analysis has already been performed in Analytics or Recon, the data is deleted from the analysis to receive the new import.
In short, when comparing Update execution and Replace data:
Objective: update data while keeping the history, versus recreating the execution data.
With a row identifier: updating keeps the history, updates records submitted previously, adds new ones, and preserves records that were not resubmitted; replacing removes all previous data and keeps only the new data.
Without a row identifier: updating incrementally adds the data and inserts everything from the new import; replacing deletes everything and inserts only the new data.
Records not resubmitted: remain in the history when updating and are removed when replacing.
Recommended use: updating is intended for partial or recurring loads; replacing is intended for complete reprocessing.
Always replace: API imports and the Integrate Process step always replace their data, without incrementing.
Process the results
With the files selected and associated with the correct sources for the reference date, click Run Process. The platform processes the results and displays the analysis screen.
When the reference date comes from the file
In Turbo AI processes with the reference date configured in the source (see the Data flow article), each distinct date in the file generates its own load, and the execution differs in four ways:
No date on the execution screen: the calendar is replaced by a notice indicating that the date comes from a source column.
Import type for the entire execution: the window always asks you to choose between Update execution and Replace execution, and the choice applies to all dates in the file.
Current folder only: manual execution considers only the folder you are currently in.
Tracking in Import Management: after you click Run Process, the platform opens Import Management filtered to this execution, and the dates appear as they are imported. At the end, a notice summarizes the rejected dates and ignored rows, when applicable.
To redo a specific date, run it again with a file containing the rows for that date: only the dates present in the file are processed.
Date limit per execution: by default, each execution accepts up to 31 dates, and the limit may be different in your environment. If the file contains more dates than the limit, no date is executed, and the execution is marked Not executed, with the number found shown in the message. Split the file and run it in parts.
Automated execution
In addition to manual execution, you can automate data imports. Automation can be triggered in three ways—by schedule, from another process, or through an external API trigger. The settings are under Settings (gear icon) → Automation. Under Automation type, you can choose:
None: does not apply any automation rule.
Scheduled process: defines the automation calendar that controls the reference dates, the loading execution time (with the option to add more times), and the offset relative to the period of the calendar.
Sequence from processes: triggers this process from the execution of another process. You select the desired process and the offset relative to the reference calendar period.
Sequence from Automation Flow: triggers the execution from an existing flow in Automation Flows under General Process Settings. You enter the trigger code—which also allows the execution to be triggered by an external trigger, such as an API call—and the offset relative to the reference calendar period.
How the offset works: the Scheduled process imports the results on the reference date calculated by the automation calendar and the offset. For example, when running the process on 03/01 with the Calendar Days calendar, it imports the data available on 03/01; with an offset of "-2", it imports the data from two days earlier, that is, 01/01.
In a process with the reference date configured in the source, automatic execution also gets the dates from the file, and each date generates a load. With Consider all folders selected, the platform generates a load for each folder and each date.
Related articles
Introduction to Processes
Create a process
View and analyze