Insight Cortex
Catch the Trend Going Wrong — Before the Quarterly Review.
The Insight Cortex watches every metric over time and surfaces drift before it becomes a number on a board.
What it solves
Most service companies discover problems at the quarterly business review. By then the trend has been bad for three months. The conversation that should have happened in week 4 happens in week 13, and the corrective action that should have started in week 5 starts in week 14.
The mechanism is structural. Nobody watches every metric every week. The CEO has too much to do; the operations team is heads-down on the operation. The trend is visible in the data, but nobody is looking.
A team large enough to monitor every KPI continuously is a luxury most service companies can't afford. So they buy the quarterly review and pay the lateness cost.
How It Works
Once the metrics cadence is running and historical values are accumulating, the Insight Cortex watches every metric continuously.
Drift detection runs against each metric on its own scale: three consecutive periods trending the wrong way; a value falling outside two standard deviations of the trailing average; a cross-correlation between two metrics that historically moved together starting to diverge.
Drift events surface in the leader brief: "Avg cycle time has risen each of the last three weeks — from 22 to 24 to 27 days. Worth a look."
The Cortex does not over-alert. Drift signals are conservative; false positives undermine trust. The model is biased toward surfacing things that pattern-match real problems, not every wiggle.
Pattern detection compounds with the Bottleneck Discovery use case: when a metric drifts, the Cortex can drill into the operational data and propose where the drift might originate.
This is the operating equivalent of a great chief of staff watching the dashboard and flagging the things that need executive attention — without the chief of staff.
What You Get
- Continuous drift detection across every metric in the cadence
- Configurable drift criteria — number of periods, deviation thresholds, cross-metric correlations
- Drift surfacing in the daily and weekly leader brief
- Drill-down — see the trend, prior comparable periods, and operational data that may explain it
- Conservative alert design — fewer signals, higher signal quality
- Trend tracking — drift events catalog, including ones that resolved without action
In practice
A 30-person service company runs 18 metrics on weekly and monthly cadences. The Insight Cortex watches them continuously. In late June, the weekly leader brief flags: "First-call resolution has trended down 3 weeks in a row — from 84% to 81% to 78%. Two weeks ago this would have been within normal variance; three weeks running breaks the pattern." Leadership digs in immediately rather than waiting for the Q3 review. Root cause: a new product line launched 3 weeks earlier is producing calls the team doesn't have a clean playbook for. Within two weeks the playbook is built, the metric recovers. The cost of not catching it until the quarterly review would have been 13 weeks of poor first-call resolution and a Q4 review with bad numbers across the board.