Skip to content

Build and Validate

Build & Validate is the first phase of the Data Reconciliation recipe: "Configure the reconciliation check and validate it." Its six steps take you from choosing two assets to an active Data Diff check whose settings the platform has accepted. This page describes each step, what its Next button requires, and what AgentQ does there. For the mechanics shared by the whole recipe, see How It Works.

Select Assets

Choose the asset you want to investigate on the left, and the reference asset to compare it against on the right.

The step shows two columns. Left, the Asset to inspect, is where mismatches are reported. Right, the Reference asset, is treated as correct. In each column, pick a Datastore and then the asset:

  • Existing Table picks a table that already exists in the datastore.
  • Computed Table builds a new table from a SQL query, and Computed Join combines two assets into one joined view. Both appear only when you can create computed assets on the datastore. A computed asset created here is selected automatically once it is ready.

For file datastores the choices read Existing File and Computed File. Either side can come from any datastore you can access, including the same datastore as the other side.

Once both assets are chosen, the Schema overlap panel shows how their fields line up: how many fields are shared, how many share a name but not a type, and how many exist on one side only. Each shared field is listed with its type; a type mismatch is marked with both types and excluded from the steps that follow. When no field name appears on both assets, the panel reads No fields in common.

Next requires both assets and at least one shared field with a compatible type.

Row Identifiers

Pick the shared field(s) that identify the same row on both sides.

A row identifier is how a row on one side finds its counterpart on the other. Rows sharing the same identifier pair up and are compared field by field, and a row whose identifier has no counterpart surfaces as missing. Add identifiers with Add identifier from the shared fields that qualify: an identifier needs 100% completeness on the left asset, and a text identifier must also be fully unique. When no shared field qualifies, the step says so. With several identifiers, rows match when all identifiers match.

When AgentQ is available, the step asks for a suggestion as soon as it opens ("Asking AI for a row identifier suggestion…"), weighing uniqueness and completeness on both sides. Suggested fields are marked with a star, and Apply all adds them.

Next requires at least one row identifier.

Compare Fields

Choose which shared fields must agree, and which ride along for context.

Compare Fields are the shared, type-compatible fields whose values must agree between the two sides; a disagreement becomes a mismatch to resolve. The row identifiers appear first, locked, since they always ride along. Use the tiles, Select all, or Clear to choose the fields to compare.

Passthrough lists the remaining fields of the left asset, including ones the right side does not have. Passthrough fields are never compared; they ride along so the results carry their context. A field cannot be both compared and passed through.

When AgentQ is available, the step asks for suggestions as it opens ("Asking AI for compare column suggestions…"), splitting the shared columns into comparisons, context, and tolerances. Apply all selects the compare and passthrough fields and pre-fills the tolerances on the next step.

Next requires at least one compare field.

Tolerances

Decide how close is "equal", and which differences count.

Differences to detect offers three toggles, all on by default:

Change type What it detects
Added Rows only the left asset has.
Removed Rows the left asset is missing.
Changed Rows whose compared values differ.

Comparators set tolerances by field type, not by field. Only the types present among your compare fields are shown, and each card lists the fields it covers:

Type Options
Numeric Exact: values must match exactly. Absolute: values within a fixed threshold count as equal. Relative: values within a percentage of each other count as equal. Both tolerances take a value.
Date / Time Exact: timestamps must match exactly. Within: timestamps within a duration of each other count as equal, with a value and a unit of seconds, minutes, or hours. The default is 5 minutes.
String Ignore whitespace treats values that differ only in spacing as equal.

The How It Works panel reacts live here, showing how the numeric tolerance and the selected change types affect the worked example.

Next requires at least one change type.

Review

Confirm your configuration. Continuing creates a draft check that you then validate and activate.

Review shows the configuration as read-only cards: Left, Right, Row Identifiers, Compare Fields with the passthrough list, and Tolerances & Scope with the change types and tolerances (or "exact equality" when none is set). Every card is clickable and takes you back to its step.

While Review is open, the recipe checks whether an active Data Diff check already covers the same fields on the left asset. If one does, a Conflict with an existing check banner explains: "An active data diff check already covers these fields on this table. Load it into the recipe, or change the configuration." Load existing check discards the configuration you were building and loads the existing check into the recipe instead.

Clicking Next creates the check as a Draft on the left asset, with the default description Data Reconciliation check on followed by the two asset names, and moves to Validate. Next is unavailable while a conflict stands.

Validate

Confirm the settings are accepted, then activate the check.

The step opens on the Validate view: "Saving the check confirms the identifiers, compared fields, and tolerances are accepted, then activates it so the full scan can run." Clicking Validate saves your latest settings to the check and sets it to Active. There is no sampled dry run in this recipe: validation is the platform accepting the configuration, including the reference asset, which you need at least the Reporter team permission to read.

The Review Results view confirms Reconciliation check validated: your row identifiers, compared fields, and tolerances were accepted, and the check is now active. Continue to the full scan to compute the actual differences. When the save is refused because it would conflict with another active check, the conflict banner appears and sends you back to Review; other failures offer Retry.

You can change any setting and click Validate again. Editing the configuration after validation invalidates it. Once validated, Scan & Reconcile moves you to the second phase. See Scan and Reconcile.

See Also