Writing · Data Quality · 12 October 2026 · 2 min read

Quality with a history: nightly scoring that remembers

A quality score on a dashboard is a snapshot. It tells you the estate is imperfect, which you knew. A useful quality system answers when a record started failing, what changed, and whether the fix held.

A data quality dashboard usually shows a number. Completeness, validity, a composite. The number tells you the estate is imperfect, which everyone already knew, and it changes a little each week for reasons nobody can name.

The questions that matter are different. When did this record start failing? What changed? Did the fix hold? A snapshot cannot answer any of them. A history can.

Scoring that remembers

TellParrot scores quality across six dimensions, per record, per source and per system, with trends, monitors and dimension breakdowns. The scoring runs nightly and it remembers: when a record first started failing a dimension, the platform keeps that moment, so a quality problem has a start date rather than a current state. Drift and watchdog alerts fire when the shape of a source changes, before the change reaches a report.

Gates on the way in

The cheapest quality problem is the one that never gets in. Before a load is committed, TellParrot shows a preview of what processing will do. Records that fail a gate are held for review, and a held record never reaches a report, a customer or a model. An ingest diagnostics engine explains mapping faults in plain terms rather than leaving a failed load to be reverse-engineered.

Repair, not rejection

Rejecting bad data protects the estate and abandons the record. TellParrot quarantines and remediates: transform and record fixes, CSV repair, and a path back into the estate once the record is right. AI agents queue the worst-quality records for steward review, so attention goes where the damage is.

Quality where you work

Scores are not confined to a quality page. Each golden record carries its own score, and the quality field is visible in the same place stewards make decisions. A person approving a merge sees the quality of what they are approving. That is where a score becomes a behaviour.

Why history changes the conversation

With a start date on every failure, quality stops being a mood and becomes a set of events with causes. That is the difference between a dashboard that is glanced at and a system that is worked.

See it on your data

Reading is the slow way to find out.

A 45-minute demo or a 14-day guided trial, run with our team on your own sources.