Review and monitor

After an Event Stitching run, review coverage, exclusion reasons, identifier quality, and graph stability before using stitched events downstream.

Event Stitching processes events into two public attribution outcomes:

Outcome	Meaning
stitched	The event was attached to a dinmo_stitched_profile_id.
excluded	The event was processed but not attached to a profile. The reason explains why.

The main rate is:

stitch_rate = new_stitched_event_count / new_processed_event_count

What to review first

Start with these questions:

• Did the latest run complete successfully?

• Does the processed event count match the expected source volume?

• Is the stitch rate plausible for each model?

• Are excluded events mostly coverage issues or quality guardrails?

• Which identifiers have low coverage?

• Which identifiers have active conflicts or many conflicted values?

• Which audit entries explain recent graph changes?

Use Overview to review latest-run KPIs, trends, model breakdown, and identifier breakdown.

Overview

Use the Overview tab to validate the latest successful run and recent trends.

Review:

Metric	What to check
Events processed	Events selected and processed in the latest run with events.
Stitch rate	Share of processed events that were stitched.
Profiles	Active profiles with at least one active identifier after the run.
Known profiles	Active profiles with at least one active anchor identifier after the run.
Stitch and identification rate trend	Run-level trend over recent successful runs with events.
Known vs Anonymous	Split between profiles with an active anchor and profiles without one.

Use these metrics as a consistency check. For every run:

processed events = stitched events + excluded events

If this does not hold, treat it as a data or reporting issue before trusting downstream analysis.

Model breakdown

The model breakdown shows each selected event model.

Column	What to check
Model	Event model name.
Event processed	Events processed for that model in the latest run.
Stitched events	Events stitched from that model.
Stitch rate	Share of processed events from that model that were stitched.

If one model has a much lower stitch rate than the others, inspect its mappings first. Many issues come from mapping the wrong physical field to a logical identifier, or from source events missing identity fields.

Identifier breakdown

The identifier breakdown appears on both Overview and Quality.

Column	What to check
Identifier	Logical identifier name.
Anchor	Whether the identifier is configured as an anchor.
Coverage	Share of processed events carrying a usable value for this identifier.
Active values	Current active values available for matching.
Conflicted values	Values kept as evidence but removed from matching because they connected too many profiles.
Active conflicts	Conflicted values still active after the run. This should normally be zero.

High conflicted values on weak identifiers usually means the identifier is shared, corrupted, too long-lived, or too trusted.

Quality

Use the Quality section to understand why events were excluded and where the graph may be unsafe.

Use Quality to inspect health checks, identifier quality, exclusion reasons, and audit entries.

Health checks

Quality starts with four health checks:

Check	What it means
Last run	Latest project run status.
Stitch rate	Latest run stitch rate and stitched event count.
Excluded events	Events processed but not stitched in the latest run.
Active conflicts	Conflicted identifier values still active after the run. Expected value is zero.

Exclusion reasons

Reason	What it means	What to do
no_matchable_identifier	The event had no usable identifier after standardization, blocked values, and stitching lifetime checks.	Improve source mappings, add missing identifier fields, or accept that this source has low identity coverage.
conflicted_identifier_value	The event depended on a value quarantined because it connected too many profiles.	Inspect the value, tighten lifetime, add blocked values, or lower trust in that identifier.
limit_conflict	The event could not be stitched without violating a configured identifier policy.	Review identifier priority, anchor settings, and limits.

Interpret reasons by family:

Family	Reasons	Typical owner
Coverage	no_matchable_identifier	Data modeling, tracking, source mapping
Quality	conflicted_identifier_value	Tracking quality, identifier policy
Guardrail	limit_conflict	Identity configuration

High no_matchable_identifier usually means Event Stitching is working correctly but the source does not carry enough identifiers.

High conflicted_identifier_value usually means one or more weak identifiers are too broad, corrupted, shared, or long-lived.

Audit

Use audit entries when you need to explain why the graph changed. Audit helps distinguish normal stitching from guardrail behavior, demotions, and excluded events.

Use Identifier policy and guardrails to adjust priority, anchor, stitching lifetime, limits, and blocked values.

For SQL examples that reproduce these checks from the public output tables, see Investigate Event Stitching results.

Runs

Use run history to confirm the project is processing the expected windows.

Review:

• run status

• run type

• processed window

• events processed

• stitched events

• stitch rate

• excluded events

• failed runs and permission issues

Use Runs to confirm status, run type, processed events, stitched events, stitch rate, and excluded events.

If a run fails, check:

• output dataset or schema permissions

• whether source tables still exist

• whether mapped fields still exist

• whether model primary key, timestamp, and partition columns are still valid

• whether the output dataset can create or replace Event Stitching tables

Audit and warehouse validation

Use the UI for first-level monitoring, then query warehouse outputs when you need proof.

Need	Output
Join events to profiles	identity_event_profile_attribution
Inspect event profiles	identity_event_profiles
Explain identifiers on a profile	identity_event_profile_identifier_values
Resolve merged profile IDs	identity_event_profile_redirects
Explain graph changes	identity_event_profile_audit_events
Power summary dashboards	identity_event_profile_run_metrics

See Event Stitching output tables for the full table reference.

Tuning checklist

When coverage is too low:

• verify identifier mappings

• add missing event identifiers

• standardize email, phone, and IDs correctly

• check whether lifetimes are too short

• inspect no_matchable_identifier in Quality

• compare model-level stitch rates in Overview

When over-merge risk is too high:

• lower trust in weak identifiers by moving them to a later priority

• remove anchor from weak identifiers

• shorten stitching lifetime

• lower maximum profiles per value

• add known bad values to blocked values

• inspect conflicted value counts by identifier before loosening policy

When one source drifts:

• compare latest model metrics to previous runs

• inspect source schema changes

• validate the physical fields mapped to logical identifiers

• inspect event samples for missing or placeholder values

Do not loosen guardrails only to improve stitch rate. A high stitch rate is useful only when the resulting event profile graph remains explainable.