What a Target Does
A target is a desired outcome stated as a number. Revenue of 500,000 by year end. Staff retention above 85%. Delivery completion rate of 95%. Targets are set by management — based on ambition, historical averages, donor requirements, or competitive benchmarks. They represent what the organisation wants to achieve.
Targets produce a specific type of organisational behaviour: target-chasing. Staff and managers focus on the number rather than the process that produces the number. When the number approaches the target from below, escalation is triggered. When it exceeds the target, it is celebrated. The process behind the number is largely invisible — only its output is assessed.
This produces two failure modes. First: the target is met while the process is deteriorating — because the output can stay above the target for months while the process drifts toward collapse. Second: the target is missed while the process is performing normally — because normal variation in any process will occasionally produce readings below any fixed threshold, regardless of whether the process is genuinely in trouble.
A target converts an observation into a performance score. A calibrated threshold converts an observation into a signal. The score tells you whether you met a desired number. The signal tells you whether something has genuinely changed in the process producing the number. These are not the same information.
What a Calibrated Threshold Does
A calibrated threshold is derived from the process's own history — specifically, from the natural variation that the process produces when it is functioning normally. Every process has a range of natural variation: some weeks will be higher, some lower, within a band that reflects the normal operation of that process. The calibrated threshold identifies the boundary of that band.
An observation that falls within the natural range is common cause variation — noise. It does not require a response. An observation that falls outside the natural range is a special cause signal — something has genuinely changed in the process. It requires investigation, not just reporting.
This distinction — between common cause variation and special cause variation — is what W. Edwards Deming established in 1982 as the mathematical foundation of process monitoring. MetriqOne integrates it directly into the PI layer through the calibration log and the moving average. Every PI threshold in the stack is a calibrated natural process limit — not a management target.
Back-End Verification Matrix
| Situation | Target response | Calibrated threshold response |
|---|---|---|
| Reading drops slightly below target | Escalation triggered. Manager intervenes. Staff feel scrutinised for normal variation. | Reading assessed against moving average. If within natural range, no intervention. If outside, investigation begins. |
| Reading stays above target while process drifts | Green signal. No response. Crisis builds invisibly for months. | Moving average trend detected. Amber signal triggered. Investigation before the threshold is crossed. |
| Reading drops far below target | Red signal. Emergency response. Usually too late to prevent damage. | Red signal triggered weeks earlier by natural process limit crossing. Response deployed when damage is still preventable. |
| Reading exceeds target significantly | Celebration. Possibly new, higher target set. | Reading assessed. If within natural range, normal variation. If outside upper limit, special cause — investigate what changed positively, and whether it can be sustained. |
How MetriqOne Sets Calibrated Thresholds
The calibration process in MetriqOne uses a moving average — a rolling calculation of the indicator's recent readings — and the natural process limits calculated from that average's variation over time. For a new deployment without historical data, the initial threshold is set using one of three proxy calibration methods: the Analogous Process Method, the Expert Estimate Method, or the Operational Minimum Method.
The threshold is declared before the first observation is taken. It is entered in the calibration log with a name, a date, and a documented basis. It cannot be revised after the fact to make a result look better — the calibration log is an immutable record. The first mandatory revision date is set at deployment: three, four, or six observation cycles depending on the proxy method used.
This discipline — declaring the threshold before observing the data — is what makes the calibration trustworthy. A threshold set after seeing the data is not a calibration. It is a post-hoc rationalisation.
A micro-enterprise tracked weekly revenue. Management set a target of 15,000 per week based on what they needed to cover costs and generate a small surplus. For twelve weeks, revenue ranged between 12,500 and 17,800. The target was met seven times and missed five times. Five times, staff felt they had underperformed. Seven times, they felt they had succeeded. The actual variation was entirely within the normal range for a business of this type and size.
When MetriqOne calibration was applied to the same twelve weeks of data, the natural process limits were calculated: lower limit 11,200, upper limit 18,900. Every reading in those twelve weeks was within the natural range. Not one of them was a signal. Not one required management intervention. The variation was noise — the normal operation of the business's revenue process.
In week fourteen, revenue dropped to 9,800 — below the lower natural process limit. This was a genuine signal. Something had changed. Investigation revealed that a key supplier had delayed delivery, causing the enterprise to miss two client orders. The response was targeted and immediate. The intervention addressed the actual cause rather than the number.
Under the target system, week fourteen would have triggered the same response as the five "missed target" weeks — escalation and pressure on staff. Under calibration, week fourteen triggered a different response to every previous week, because it was genuinely different. That is what a signal does. That is what a target cannot do.
Takeaway
Targets produce target-chasing. Calibration produces signal detection. The first focuses the organisation on a number. The second focuses it on whether the process behind the number has genuinely changed.
Common cause variation is not a signal. Responding to it as if it were is tampering — and tampering makes processes less stable, not more. The calibrated threshold identifies what is noise and what is signal.
The threshold must be declared before the first observation. A threshold set after seeing data is a rationalisation. The calibration log records the threshold with a name and a date — before any observation is taken.
Amber is the most valuable signal in the system. It detects drift before the threshold is crossed. Targets have no amber — only green and red. Calibrated thresholds give you time to respond before the failure arrives.