Set up a rolling data series

A data series can be built from zero — exactly the way you would use a lab or process-tracking Excel sheet, only in an auditable, downloadable form that rolls across shifts. This article covers how you create a series and configure its columns. Shifts then record into the finished series — see Record into a rolling data series.

Before you start

• Layout-management permission: layout:manage — the same permission you use to edit shift-log elements.
• The paid add-in is enabled. If not, the settings page says so (the capability exists, it is just not ordered).
• Where: on the Management → Data series page.

Creating a new series. Backdating defaults to 1 day and the edit window to 15 minutes — both adjustable.

Steps

1. Open Management → Data series and click New data series.
2. Give the series a header: Label, Plant (or “Whole site” for a site-level series, e.g. a central lab), and optionally a Description. The Document ID and Version are controlled-document metadata (e.g. FRM-011/002 · v8) — a label only; they do not protect the history.
3. Set the behaviour:
   • Backdating (days): how far back a sampling time may be set (default 1 day). 0 means backdating is not allowed.
   • Edit window (minutes): how long your own row can be corrected in place (default 15 minutes); after that a new version with a reason is required.
4. Create the series. Then add its columns — these define the two-level header on screen and the order of the export.
5. For each column set the label, the type (number, text, select, date…), and where it makes sense: the unit or instrument tag, the min/max limit, or a calculated field. The group (the top header row) ties related columns together.

Column order is a contract

The export and the table follow the column definition, not the order of the stored values — so two downloads always produce the same column order. Moving a column therefore moves both the display and the export.

A limit only where a single bound is true

Add a min/max limit only where a single global bound is true. If a parameter’s legitimate range differs per sampling point or product grade, a shared min/max would raise a false alarm for half the series — in that case it is better to have no signal than a lying one.