InnerWeb
Skip to main content

Comparing with the previous period in a safety report

A percentage change in a report looks like a ready-made conclusion, but it depends on what it is measured against and on the size of the sample. A report in InnerWeb compares the period with a window of the same length immediately preceding it. Below are practices for reading changes that protect against hasty decisions based on a single week.

4
independent content toggles, including comparison with the previous period
Source: reporting module documentation
-36.4%
change in the numerical example: 28 against 44
Source: sample calculation

How the comparison is calculated

The comparison is based on a window of the same length immediately preceding the report period, and the change is shown as a percentage change. For a weekly report the predecessor is the previous full week, for a monthly report the previous calendar month, and for a daily report the previous day.

The comparison is switched on with a separate toggle, alongside charts, details and analytics. You can therefore prepare a report with comparison for the director and a report without comparison for a hall supervisor who needs only a list of matters.

Practice 1: with small numbers, look at the numbers, not the percentage

A percentage change is misleading with small values. A rise from 2 to 4 permits is plus 100 percent, while a rise from 200 to 220 is plus 10 percent, even though the absolute increase is larger. Always read the absolute value that the report shows next to the change.

For functions with small numbers, for example alarms from lone workers in a small hall, the percentage should not be the basis for a decision. Wait for several periods until the picture stabilises.

Sample number of closed permits over four weeks

Sample number of closed permits over four weeksLine chart: 40, 44, 28 and 41 closed permits in consecutive weeks, with one week of deviation.2530354045permitsWeek 1Week 2Week 3Week 4Week 1: 40 permits40Week 2: 44 permits44Week 3: 28 permits28Week 4: 41 permits41
Source: Sample calculation: four weeks, illustrative values

Practice 2: one week is not a trend

A change between two consecutive weeks may be caused by the calendar: holidays, maintenance shutdowns, seasonality. Before you treat it as a signal, compare it with at least several earlier periods. The report compares with only one window, so judging a trend requires placing several reports side by side.

A simple method: keep an archive of weekly reports and once a month look at four consecutive values of the same indicator. Only a direction that holds over several periods justifies a change in how work is organised.

Numerical example

For example, the number of closed permits over four consecutive weeks is 40, 44, 28 and 41. The change in the third week is minus 36.4 percent, and in the fourth week plus 46.4 percent. Only the whole series shows that the third week was a deviation and not the start of a decline. This is a sample calculation.

Practice 3: make sure the periods are comparable

Windows are always complete and of the same length, but months differ in the number of days. The monthly report compares a month with the previous calendar month, so February may look weaker than January simply because it has fewer days. For such comparisons, check the number of working days as well.

In the weekly report this problem does not arise, because the window always covers seven days from Monday to Sunday. This is an argument for the weekly cadence when a clean comparison matters.

Practice 4: combine the comparison with clean data

A change calculated from inflated data is inflated twice: once in the current period and once in the previous one, but differently. The comparison therefore makes sense only when the data is cleaned of service accounts and test alarms, and when times are presented as a median.

If system tests ran in the previous period but not in the current one, the change will show a decline that is merely the effect of the tests ending. Keep notes in the report on when trials were carried out.

Practice 5: set the reaction threshold before reading the report

The best decisions are made when the reaction threshold is set in advance: for example, three consecutive periods of growth in a given indicator, or exceeding an absolute value. The report then becomes a test of the rule, not a pretext for interpretation.

InnerWeb provides data from its own area, so the threshold applies to indicators that the system measures. Compare data from other systems, such as time and attendance or ERP, separately.

Checklist

Video

Frequently asked questions

What does the report compare the current period with?

With a window of the same length directly preceding the report period. The change is shown as a percentage.

Can the comparison be switched off?

Yes. Comparison is one of four independent content toggles in the report, so it can be enabled only for selected recipients.

Is one report enough to assess a trend?

No. The report compares with only one window, so draw conclusions about a trend from several consecutive reports.

Let us read reports together with your team

We will show a report with comparison and discuss how to set reaction thresholds for hall supervisors and management. Book a short demo.

Let's talk