When the Records Disagree
The five reasons records disagree, why a constant difference is almost never behavioural, and what to do when the disagreement cannot be resolved.
Two systems, eleven and a half minutes apart, and what the difference was asked to show
A constant eleven-and-a-half-minute difference across a month is a clock offset. The camera clock was measured afterwards and found to be slow by exactly that much. This is one employer's own file.
A constant difference between two systems is a clock, not a person. Eleven and a half minutes, every day, for a month, is not somebody behaving consistently oddly; it is one device set wrong. That reading is available within an hour and it is the first thing to test, because it disposes of a large share of apparent discrepancies.
The evidential discipline in “When the Records Disagree” should also govern workforce technology. A team evaluating visit the Monitask resource centre for employment of relatives policy can use time and project records as operational context, but should preserve the original record, document access and let the employee correct a misleading entry before it supports a conclusion.
Variable differences are the interesting ones, and even those have several ordinary explanations before anybody's conduct is reached.
For a separate benchmark relevant to “When the Records Disagree”, consult the OSHA workplace stress resources. Use it to test scope and safeguards against an external standard before the process is approved.
The five reasons
Clock offset: constant, measurable, boring, common.
Different events: the two systems are recording different things. A door opening and a terminal punch are separated by a walk.
Different definitions: one system records when something started, the other when it was saved.
Shared or wrong credential: the systems are describing two different people.
Something actually happened: the residual category, reached last.
Testing the first four
- Measure both clocks against a known source, today.
- Check whether the difference is constant across the period.
- Establish what event writes a row in each system.
- Establish whether the identifier is individual in both.
- Walk the physical route between the two points, with a watch.
- Only then consider the fifth reason.
Step five is underrated. A ninety-second walk between a gate and a terminal explains a ninety-second difference completely, and it is measurable by anybody willing to walk it.
Variable differences
Where the gap changes from day to day, the clock explanation fails and something situational is likely: a queue that varies, a route that varies, a process that sometimes includes a step.
Plotting the difference by day frequently shows the pattern immediately — larger on Mondays, larger when a particular person opens up, larger when the weather is bad and everybody arrives at once.
When it cannot be resolved
Sometimes the records simply disagree and nothing explains it. The honest response is to record that, and to be careful about what weight the disagreement can carry.
An unexplained discrepancy is not evidence of wrongdoing; it is evidence that the organisation's records are inconsistent, which is a fact about the organisation.
Which record to prefer
Where one system is documented, synchronised and purpose-built, and the other is an unchecked clock on a four-year-old recorder, the preference is obvious and should be stated with its reason.
What is not acceptable is preferring whichever record supports the conclusion, without saying why. That choice is visible in the file and it is the thing that makes a finding look constructed.
The employee's account in the middle
Often the person's own account falls between two disagreeing systems, which is weak corroboration of nothing and is worth noting rather than treating as evasion.
People estimate their arrival to the nearest five minutes. An account of "about ten to eight" is consistent with 07:47 and 07:53 and should be read that way.
Agreement is also worth testing
The reverse case gets no attention and deserves some. Two systems that agree exactly, to the second, are usually not two measurements — they are one measurement propagated, or two devices synchronised to each other rather than to anything true.
Exact agreement between independent clocks is unlikely. Where it appears, the useful question is whether one system is reading the other's output, which is common in integrated installations and invisible in the report.
That matters because two systems reported as agreeing carry more weight than they should, and the overstatement is harder to spot than a disagreement.
Recording the test, not just the result
Whatever is concluded, the file should carry what was actually done: the clocks measured today, the route walked, the event definitions established.
A conclusion that the difference was behavioural, with nothing showing the alternatives were tested, is an assertion. The same conclusion with four lines of testing behind it is a finding, and the four lines took an hour.
Measuring the walk
The physical distance between two recording points explains a great many small discrepancies and nobody measures it.
A gate and a terminal ninety seconds apart produce a ninety-second gap on every shift, for everybody, forever. Walking it once and writing the figure down turns a recurring puzzle into a constant that can be subtracted.
What the file should record about the difference
The two records, the difference, what was tested, what was found, and what weight the organisation is placing on the discrepancy and why.
Five lines. They are what turns an apparent discrepancy into either a finding or a closed question — and in the large majority of cases, once the clock is measured, it is a closed question.