Metadata vs pipeline output

When a spinner finishes a contact, CXGear stores two different kinds of information. Mixing them up is the most common reason Records looks “wrong” even when the run succeeded.

This page explains the difference in plain language: what went in, what came out, and what is deliberately hidden.

The short version

Kind of data	What it is	Where you see it
Metadata payload	The contact fields that entered the pipeline	Records → detail → Metadata (pipeline input)
Pipeline outcomes	Results the run produced (disposition, AI text, message ids)	Records list Outcome column and detail panel
Internal / debug fields	Log bookkeeping, model names, step plumbing	Pipeline journey only — not the main Records view

Think of metadata as the inputs on a work order, and outcomes as the stamp on the completed form.

What is the metadata payload?

Before a spinner runs, CXGear copies only the metadata fields you configured on the data table into each execution record.

That payload is the contact context every step can read — for example name, phone, account id, language, or amount due. Columns that are not metadata stay in the source table under Data tables, but they do not enter the pipeline.

Screenshot needed

Records detail panel showing Metadata (pipeline input) with only configured fields.

Where: Project → Records → click a row

Save as: src/assets/screenshots/33-records-metadata-payload.png

Why metadata is separate

Operations teams often import wide CSVs (twenty or more columns). Automation usually needs a handful. Metadata fields are that short list — the contract between your spreadsheet and every spinner step.

If a field is missing from metadata:

Templates and step parameters cannot see it
The Records contact line will not show it
Column pickers in the spinner builder will not list it

Configure metadata under Data tables → Metadata. See Metadata fields.

What are pipeline outcomes?

Outcomes are what the pipeline produced for that contact — the operational answer to “what happened?”

Typical outcomes include:

A disposition or tag (for example “Interested”, “No answer”)
An AI reply or summary text from an AI agent step
Channel references such as a WhatsApp message id
Other step results that matter for follow-up and reporting

Outcomes are written as the spinner runs. They are not a copy of your CSV. They are the result of work CXGear did.

Screenshot needed

Records list Outcome column and detail panel Outreach outcome section.

Where: Project → Records

Save as: src/assets/screenshots/34-records-outcomes-panel.png

Why outcomes matter more than full step output

At high volume, supervisors need:

Who was contacted (metadata)
What happened (outcome)
When (completion time)

They do not need every intermediate log line or internal key in the grid. That is why Records stays compact and outcome-focused.

What stays hidden (on purpose)

Spinners produce more data than operators should see every day. CXGear hides internal fields from the main Records outcome list, including things like:

Log step bookkeeping and template debug noise
Model names and provider plumbing
Intermediate keys used only between steps
Journey event payloads meant for debugging

Those details are still available when you open Pipeline journey from the record detail panel. Journey is for builders and support — not the daily ops screen.

How the pieces fit together

Data table row
    → metadata fields only
        → spinner steps run
            → outcomes written for ops
            → journey events stored for debug

You want to…	Look at…
Confirm which contact fields entered the run	Metadata (pipeline input)
See disposition / AI text / message ids	Outreach outcome
Browse the full import (all CSV columns)	Data tables → Show data
Debug a failed or surprising step	Pipeline journey

Success looks like

Metadata fields are set before serious runs
Records contact lines show the right identity fields
Outcome column answers “what happened?” without opening every row
Internal noise is absent from the list (and only appears in journey when needed)

Common problems

Problem	Likely cause	What to do
Contact line is empty or incomplete	Metadata not configured (or wrong fields)	Set metadata and run again
A CSV column never appears in steps	Column is not metadata	Add it under Metadata
Outcome looks thin but job completed	Step did not write an operational outcome	Check Tag / AI / channel steps; use journey
Old records still show old fields	Finished jobs keep their payload shape	New runs use the new metadata list
Too much technical clutter on Records	Expecting journey data in the ops view	Use journey only when debugging