Brainforge Knowledge

data-quality-assessment-template

3 min read

[Client] — Data quality assessment — [Period / domain]

About this document (Brainforge)

Internal conventions for how this file works in the repo. Optional: strip or export without this section when sharing a client-only artifact.

Titling and filename

Use [Client] — Data quality assessment — [Period or domain] for the document title. Examples: LMNT — Data quality assessment — Q2 2026 · Acme — Data quality assessment — Wholesale (April 2026).

Filename: {client}-dq-assessment-{period}.md under knowledge/clients/{client}/resources/.

When to use this template

Use this template when:

  • producing a periodic data quality health report (monthly, quarterly)
  • tracking freshness, completeness, and accuracy trends over time
  • surfacing data decay before it causes a fire drill

Do not use this template when:

  • investigating a specific data accuracy issue (use the Data Findings Memo)
  • documenting a production incident (use the RCA Memo)
  • profiling a new data source for the first time (use the Discovery Memo)

Document metadata

Status: [Draft / In review / Published] Assessment period: [MMM YYYY – MMM YYYY] Warehouse: [Snowflake / BigQuery / other] — Account/region: [details] Sources assessed: [source 1], [source 2], ... Previous assessment: [date or link] (omit if first) Prepared by: Brainforge Last updated: [YYYY-MM-DD]

ArtifactLink / pathNotes
Discovery Memo(s)[path to A1 memo]Baseline source catalog and SLAs
Data Platform Documentation[Google Sheet link]Source catalog, metric definitions
Previous DQ Assessment[link if exists]Trend comparison
Known open issues[Linear URL or list]Items still unresolved from last period

1. Overall health

One-line summary of the assessment period. Give the overall health status as a color, then a sentence explaining it. A director should read this and know whether to worry.

Health: [🟢 Green / 🟡 Yellow / 🔴 Red]

[2–3 sentences. What is the overall state of data health this period? What improved, what declined, and what is the one thing leadership should know?]

1.1 Period-over-period comparison

MetricThis periodPrevious periodTrend
Sources meeting SLA[N / total][N / total][improving / stable / declining]
Sources with errors[N][N][improving / stable / declining]
Days with ingestion delays[N][N][improving / stable / declining]

2. Freshness

Per-source report card on whether data arrived within its SLA.

2.1 Freshness by source

SourceSLAActual latencyMet SLA this period?Missed daysNotes
[source][e.g., ≤ 24h][p50/p95 latency][Yes / Mostly / No][N][e.g., 3 weekend delays in March]
[source][e.g., ≤ 6h][p50/p95 latency][Yes / Mostly / No][N][...]

2.2 Freshness incidents

  • [Date]: [Source] was [X]h late due to [cause]. Resolved.
  • [Date]: [Source] missed refresh entirely. Impact: [Y].

3. Completeness

Assessment of null rates, missing date ranges, and row count anomalies.

3.1 Completeness by source

SourceCritical columnNull rate this periodNull rate previousNotes
[source][column][%][%][e.g., increased nulls caused by schema drift in March]
[source][column][%][%][...]
SourceExpected rows/periodActual this period% ExpectedNotes
[source][~N][N][%][e.g., Black Friday volume spike accounted for]
[source][~N][N][%][e.g., connector outage Mar 10–12]

4. Accuracy

Spot-check results and known discrepancies. This section is lighter than a full Data Findings investigation — it flags issues for deeper triage.

4.1 Spot-check results

CheckMetricExpectedActualMatch?Date checked
[e.g., Revenue vs source][Total revenue][$X][$Y][Yes / Variance: X%][YYYY-MM-DD]
[e.g., Order count parity][Order volume][N][N][Yes / No][YYYY-MM-DD]

4.2 Anomalies flagged for investigation

  • [Issue] — [brief description, linked to relevant Linear ticket or Findings Memo]

Ordered by priority. Each action should be specific and actionable by a named owner.

  1. [Action] — [Why this, why now, who should own it]
  2. [Action] — [...]
  3. [Action] — [...]

Appendix — Pre-handoff QA checklist

  • Freshness SLA table covers every source assessed
  • Completeness null rates are queried (not estimated)
  • Row count trends compare current vs expected with explanation for variance
  • Accuracy spot-checks include date run
  • Periodic open items from last assessment are carried forward if unresolved
  • Overall health color is justified in the narrative

Graph View