Review data points

The Events table in a project’s Data Inputs section lists your collected data one data point at a time. Each row is a single measured value with its own type, status, and alerts. Because each value gets its own row, a data point that arrives from a data feed or the API with no value still appears in the table, showing as Missing with a status of Incomplete. A data point that was never submitted has no row at all, and clearing a value on an existing event removes its row.

​

Reading the table

Source, Tracking ID, Feedstock, Asset and Created by are hidden until you turn them on. Use Columns to show or hide them. Your choice is remembered per project in the browser you set it in. You can also filter by event ID without showing it as a column. Substituted values all carry the same marker. Hover one to see how it was derived: the original value, the method, and either the rule that wrote it or the source it came from. A rule can revisit a correction it already made. The correction history in the Event Detail drawer then shows the value as re-derived by that rule, with the cause on the line beneath: a data point from an earlier method in the rule’s chain became available, the data point it used changed or was deleted, or the available data points changed.

​

Filtering and sorting

Start with the Event / data point type quick filter at the top to narrow to the event type or data point type you care about. Add column filters for anything more specific, and combine them freely. Filtering the Origin column offers three options. Substituted covers every value that currently holds a correction, including gaps that were empty and have since been filled. Reverted finds values that have been restored to their original, and Original finds the ones never touched. A value sits under exactly one of the three. To find the gaps that are still empty, filter the Value column with Is missing. Unlike the other Value comparisons, it needs no data point type filter first, so it works across the whole table. Filtering Status to Incomplete also returns them, along with values marked incomplete for other reasons. To see how much of the data you expected has arrived at all, rather than which individual values are missing, use the capture rate figure on the toolbar. See Data capture rate. Filter by AI takes a plain-language request instead of a filter stack. Ask for “anomalous data points from the last month”, “data points pending review”, or “data points from this month” and it builds the filter for you. Sorting and filtering by Value is switched off until the table holds a single numeric data point type, because comparing values from different types is meaningless. The Event / data point type quick filter does not switch it on, so use a column filter:

1. Open Filters and add a filter on the Data Point Type column.
2. Set its operator to is or is any of, and pick one numeric type.

The Value filter input and the Value column sort both come alive at that point, and both tell you what to add while they are switched off. If you add other filters alongside it, keep the filter panel on And.

​

Finding data points that need attention

Needs attention on the toolbar lists the values in the current view that are waiting on someone, with a count against each queue. Pick a queue and the table narrows to it; pick the same queue again to clear the filter. It runs no new checks of its own. When nothing in the current view qualifies, the button reads All clear. A count above 99 shows as 99+. Values carrying an alert show it in the Status column. Alerts are raised per data point, and by default a rule that reads several values can flag any of them. A rule can be set to flag only the values it is really about, which keeps triage on the value that breached. See Managing alerts for how to triage them, and Create a Data Rule for choosing which value an alert lands on.

​

Exporting

Click Export and pick CSV (.csv) or Excel (.xlsx). The export is flat, one row per data point, and it honours the filters you have applied, so you can scope the file by narrowing the table first. An Excel export is capped at 100,000 rows and a CSV at 1,000,000. Over the Excel cap the export fails and tells you the row count, so switch to CSV or narrow the filters. Over the CSV cap it reports a generic failure, so if a very large CSV fails, narrow the filters and try again. In the export, Status reads Alert when a rule or evidence alert is firing on the value, and Alert Reason gives each rule’s name and alert message. On accounts with evidence checks enabled, it includes failed evidence checks as well. When substitutions are enabled on your account, the export carries four more columns after Unit that describe each substituted value: Exports are prepared in the background. Stay on the page until the file is ready, because the download link is held by the page and is lost if you navigate away.

​

Deleting data points

Select rows and click Delete. Because data points live inside events, the dialog asks which you meant to remove and previews the effect before anything is deleted: The dialog counts the events and data points each choice would affect, and calls out cases where the choice is made for you. An event cannot exist without data points, so deleting the last value in an event removes the event as well. When both choices would delete exactly the same rows, the choice disappears. Values already used by a model run are locked, and they are skipped and reported as skipped.

​

Substituting a value

A substitution replaces a data point’s value without touching the original record, which stays visible with its full lineage. Substitutions are enabled per account, so ask your Mangrove account team to switch them on. A substitution is written from one of four places:

• Editing the value in the app, which records your edit as a manual override.
• A substitute rule, which corrects every data point matching its condition and keeps correcting new ones. See Data Rules Fundamentals.
• The API, one value at a time or in a batch. See Correct a data point value.
• A data feed transformation, applied as the value arrives. See Bulk import.

Hovering a substituted value tells you which of them wrote it. Every change to a value that already holds one appends to the same history, so the data point keeps what it was captured as alongside what replaced it. Editing a row to the value it already shows changes nothing and appends nothing. A rule, the API or a transformation filling a gap records a substitution like any other, and the value sorts under Substituted with the rest. Typing a value into an empty cell yourself records an initial measurement and starts no history. To undo a substitution, click the row to open the Event Detail drawer, click the data point’s name to expand its correction history, and click Restore to original. That puts back the original value, clearing every correction standing on it in one step, and the row’s Origin becomes Reverted. The history keeps each correction and records the restore as one entry from the value in force back to the original. A restore cannot itself be restored, though a later substitution on the same value can be. The rule that wrote the correction does not apply it again. A correction that a rule has proposed but nobody has decided yet shows Correction pending approval in the Status column, and the value stays as measured until someone approves it. See Approve corrections. Substituting a value does not recalculate anything downstream. A batch that was already generated keeps the value it was generated with. Data rules do re-evaluate against the substituted value, so a substitution can clear an alert or raise a new one.