SchemaPulse
PricingContact

Intelligence · Recommendations

From showing information to explaining what requires attention.

A dashboard tells you a number moved. A recommendation tells you what that means, how confident it is, what it would cost you to ignore, and where to look — with the evidence attached so you can judge it rather than trust it.

Meet the assistantMeet the assistant

Anatomy

Six elements, always in the same order

Backup policy has never produced a verified restore

reporting-mssql-1 · policy nightly-full-30d

Elevated
Finding
The policy has completed 94 consecutive runs. None has been followed by a successful verification, so the restore path has never been exercised.
Evidence
94 runs since 2026-06-02 · 0 verification records · last restore run never · storage target s3://backups-prod
Impact
One instance carrying 2.1 TB. Recovery time is unproven and the recoverable window is unconfirmed.
Action
Run a verification restore into an isolated target. SchemaPulse can schedule this in the existing maintenance window.
Confidence
High — derived from run records, not inference
Why now
Retention shortened to 30 days on 2026-08-28, reducing the recoverable window while verification remained absent.
Fig. 1This recommendation demonstrates insight that monitoring-only platforms typically cannot derive without backup-verification and recovery context.
Finding
What was observedMetrics, query analytics, topology, backup history, inventory
Evidence
How we know — linked and inspectableThe specific metric series, queries, events or runs
Impact
Which instances, how severe, what is at riskFleet scope and the severity model
Action
What to do about itEngine-specific operational guidance
Confidence
How certain, and what would raise itSignal strength and corroboration
Why now
What changed to trigger thisDeltas against baseline, deployments, topology change

Evidence is the load-bearing element. A recommendation that cites its evidence is auditable. One that does not is a guess with a confidence score attached, and database teams have learned to ignore those.

Coverage

What SchemaPulse recommends on

Query & schema

inefficient or unused indexes, slow-query regressions, plan degradation

Cluster & replication

replication issues, unstable elections, topology anomalies

Resource & capacity

saturation trends, capacity concerns

Resilience

backup policies without successful verification, recovery risks

Configuration & estate

configuration problems, infrastructure drift

Recommendations are drawn across the whole platform, not only query statistics — which is why resilience findings like the one above are possible at all.

The loop

Detect → Understand → Recommend → Act

  1. 01

    Detect

    A threshold crosses, a trend inflects, a run completes without verification.

  2. 02

    Understand

    Correlate it with topology, deployments and history to establish what it is part of.

  3. 03

    Recommend

    State the finding with its evidence, impact, suggested action and confidence.

  4. 04

    Act

    Where SchemaPulse can carry out the fix — schedule a verification restore, open a maintenance window — it offers the action.

Recommendations explain and advise before they automate.Where an action can be carried out inside SchemaPulse it is offered, never taken unsupervised. In database operations, restraint is a feature: the first bad automated action costs more trust than a hundred good ones earn.

Where to look next