When an Indicator Stops Being Trustworthy
Treating every indicator with suspicion produces the same result as reading none of them. The skill lies in telling when a figure can be trusted and when it needs a second source of confirmation.
What follows addresses the when: the conditions under which one indicator suffices, the conditions under which it does not, and what to do once doubt sets in. The rules are few; applying them is a matter of habit.
When Is One Enough?
BU BÖLÜMÜN ÖZETİ
- When the unit is fixed
- When the scope holds
- When no system sits in between
- When all three hold
Three conditions allow an indicator to be read on its own.
When the unit is fixed
Units, headcount and days are unaffected by inflation. These indicators compare directly across time.
When the scope holds
Where the group being measured stays constant, the ratio is reliable. Same customer base, same product group, same period definition.
When no system sits in between
Orders placed directly, appointments made by phone, sales in the shop — none of these has a layer between indicator and outcome.
When all three hold
One indicator suffices and looking for a second creates busywork. Pairing every indicator makes the exercise unsustainable.
When Is It Not Enough?
BU BÖLÜMÜN ÖZETİ
- When it is denominated in money
- When it is a ratio
- When a platform sits in between
- When it is unexpectedly good
Verification becomes necessary when any one of the three breaks.
When it is denominated in money
Revenue, basket value, wages and asset values. In a high-inflation environment none of these gets read alone.
When it is a ratio
Unemployment, conversion, satisfaction and return rates. The denominator can move and the improvement can come from there.
When a platform sits in between
Anything arriving through a search engine, marketplace or intermediary. When the layer changes, the indicator holds still while the outcome moves.
When it is unexpectedly good
An improvement you cannot explain usually comes from a change in measurement. Unexplained good news gets verified first.
What to Do When Doubt Sets In
BU BÖLÜMÜN ÖZETİ
- Check the pair first
- Then check the period
- Then question the measurement itself
- Intervene last
Four steps, applied in order.
Check the pair first
Units for revenue, denominator for ratio, clicks for ranking. Because the paired indicator is already in hand, this step takes minutes.
Then check the period
Was the comparison made against the previous period or the same period a year earlier? The wrong choice manufactures trends that are not there.
Then question the measurement itself
Did anything change in how this was measured? A new system, a new definition or a fault can have distorted the data.
Intervene last
A correction made before these three steps usually addresses a problem that does not exist while masking the one that does.
When Should You Wait?
BU BÖLÜMÜN ÖZETİ
- When it is one period’s movement
- When something changed externally
- When you are in a seasonal period
- When several things changed at once
Four situations where holding off is the right call.
When it is one period’s movement
A month’s or a quarter’s data can be volatility. Structural decisions wait for two or three periods in the same direction.
When something changed externally
Measurements taken immediately after a platform update, model change or regulatory shift can be unreliable. In one model transition, part of the early measurement came from a fault later corrected.
When you are in a seasonal period
Every business fluctuates at certain times of year. No interpretation without comparison against the same period last year.
When several things changed at once
A site change, a campaign and an external update overlapping makes the cause impossible to separate. If you cannot isolate a variable, you cannot attribute the outcome.
When Should You Look Immediately?
BU BÖLÜMÜN ÖZETİ
- When there is a signal on the cash side
- When a fault is possible
- When two channels give the same signal
- The test is simple
Three situations where waiting does not apply.
When there is a signal on the cash side
Collection periods lengthening or overdue receivables growing make waiting on a volatility assumption expensive.
When a fault is possible
A form not working, a price displaying incorrectly or a page failing to load is a malfunction rather than a trend. That gets fixed without waiting for verification.
When two channels give the same signal
Where sales data and customer feedback point the same way, verification has effectively already happened.
The test is simple
Wait where waiting produces a recoverable loss; look where it does not. The decision rests on reversibility rather than on the size of the problem.
A Solid Digital Foundation
BU BÖLÜMÜN ÖZETİ
- Set up period comparison correctly
- Log your changes
- Keep historical data
- Set it up once, use it forever
Verification depends on data being stored with the right period and the right breakdown.
Set up period comparison correctly
Comparing against the same period a year earlier rather than the previous one strips out seasonality. The Google Search Central documentation explains how period selection changes the result in performance data. The wrong period manufactures a trend that is not there.
Log your changes
Without a record of what you changed and when, you cannot separate the cause of a decline. A simple change log ends that problem.
Keep historical data
Verification requires a time series. Without data from earlier periods there is nowhere to look when doubt arises.
Set it up once, use it forever
A comparison window and a change log take an afternoon to establish and then sit waiting every time something looks odd. Building that habit runs through digital consulting; the guide to the period places it among the other corrections.
Frequently Asked Questions
Sık Sorulan Sorular
No. Indicators with a fixed unit, an unchanging scope and no system in between are reliable on their own.
Minutes where the paired indicator already exists. The lengthy part is building the pair, and that happens once.
A few weeks after an external change; in a seasonal period, comparison against last year’s data is enough.
The test is reversibility. Wait where the cost of waiting can be recovered; do not wait on the cash side or where a malfunction is possible.
Where you cannot explain it, yes. Unexplained improvement usually comes from a change in measurement.
Whichever figure you quote most often in meetings. Pair that one and the habit spreads to the rest by itself.
Source: Prepared from the shared verification pattern across the six developments covered in this set.