KPI anomaly detection

Find meaningful departures from expected performance, investigate the likely drivers, and route the evidence to the responsible owner.

How the workflow runs

  1. 01

    Define expected behavior

    Record metric definitions, relevant segments, seasonality, targets, and the minimum change worth investigating.

  2. 02

    Detect the departure

    Compare current performance with expected ranges and run data-quality checks before opening an alert.

  3. 03

    Investigate the drivers

    Break the change down by segment, correlate relevant operational events, and test plausible explanations.

  4. 04

    Route the finding

    Send a sourced explanation to the metric owner and track the investigation through resolution.

Evidence in, action out.

The operating problem

Static thresholds generate noise and dashboards show the anomaly without explaining what changed underneath it.

How the workflow runs

Compare metrics with expected behavior, verify data quality, investigate the largest drivers, and route a sourced explanation.

Evidence and inputs

KPI time series, dimensions, targets, event calendars, operational data, and metric definitions.

Actions and handoffs

Detect anomalies, segment the change, test hypotheses, notify owners, and open an investigation workflow.

Operating outcome

Faster movement from alert to explanation, with fewer false alarms.

Controls for production

Minimum effect sizes, seasonality handling, owner-defined baselines, data-quality checks, and review before escalation.

Move from anomaly to explanation

Spinrun gives metric owners the context behind a material change, with the evidence and remaining uncertainty visible.

Company Brain

Pairs KPI definitions with agreed thresholds, investigation rules and owner notes for coworkers assessing anomalies.

Coworkers

Compare connected KPI series with agreed baselines, investigate threshold breaches and draft alerts with operating context.

Approvals

Metric owners examine data quality and anomaly explanations before escalation; each investigation run is recorded.

One intelligence layer from plan to delivery

16x

Faster time-to-insight

5

Fewer tools needed

80%

Avg. cost reduction

8 hrs

Saved per team per week

Put investigated alerts into existing workflows

Connect data platforms, BI tools, and incident or planning systems so findings reach the owner with supporting evidence.

Get a tailored demo

Connected sources

SlackConnected
SalesforceConnected
IntercomConnected
BigQueryConnected
StripeConnected

Turn AI potential into real business outcomes

Get in touch