Weather, and Other Things People Blame
The explanations that get offered for a miss and never tested, why they survive, and the cheap test that settles most of them.
Weather, the news, the school holidays and "people are just busier at the moment" are the explanations that get offered when a forecast misses, and they are almost never tested. They survive because they are offered after the fact, about a specific period, and nobody checks whether the same condition occurred in periods that were normal.
The method in “Weather, and Other Things People Blame” is only as useful as the definitions and source records behind it. A team exploring consult the official workflow overview for how employees cheat time trackers can add timely work and project context, with transparent settings, limited access and a documented correction route around every consequential use.
The test is cheap and it resolves most of them in an afternoon.
For an independent reference relevant to “Weather, and Other Things People Blame”, consult ILO employment-policy resources. Compare the general material with the organisation’s actual horizon, data definitions and decision process.
The asymmetry that keeps them alive
An explanation is only ever offered for the periods it appears to explain. Bad weather in a high week is cited; the same bad weather in an average week is not.
- Bad weather in a high week is cited.
- The same weather in an average week is not.
- The same weather in a low week is never mentioned at all.
This is what makes the belief unfalsifiable in ordinary conversation and trivially falsifiable in the data.
Testing a claimed cause
Take the condition — rain over a threshold, a school holiday, whatever is claimed. Split the series into periods with and without it. Compare the distributions, controlling for weekday and season.
If the two distributions overlap almost entirely, the driver is not operating at the size claimed.
What usually happens to weather
In most indoor service operations it fails, or produces an effect of one or two per cent against daily noise of ten. In operations with a physical element — field work, walk-ins, transport — it frequently passes and is substantial.
The point is not that weather never matters. It is that it is an empirical question with a cheap answer.
School holidays
Usually real but smaller and differently shaped than people assume. The effect is often on the intraday profile — a flattening of the morning peak — rather than on the daily total.
Tested, it becomes a usable adjustment. Untested, it is an excuse available six weeks a year.
"Things are just busy"
The purest form. It describes the observation rather than explaining it, and it closes the conversation.
The useful follow-up is "busy with what" — which category, which channel, which customers. That question frequently finds a real driver that nobody had in the register.
The news
Occasionally enormous, usually irrelevant, and almost always invoked. Where an operation is genuinely exposed to news — financial services, utilities, anything with a regulator — it is worth treating as an event class with a recorded history.
Elsewhere it is a story applied retrospectively.
Why it matters that these get tested
Not for tidiness. Because every untested explanation accepted for a miss is a correction not made. The forecast missed for a reason, and "the weather" ends the search before the reason is found.
In practice the real cause is usually in the event calendar, or in a definition change, or in a driver whose size has moved.
The explanation that is really a defence
Sometimes the explanation is offered by the person whose forecast missed, which makes testing it socially awkward. That is worth recognising rather than ignoring.
The way through is to test the claim as a claim about the operation, not about the forecaster: "let us see how often that condition moves the number, because if it does we should be carrying it."
Testing turns a blame into a driver
The best outcome of the test is not disproof. It is that a vague belief becomes a measured driver with a size and a trigger, which the forecast can then use.
A tested weather effect of 6% on field visits is worth having. The untested belief was worth nothing to anybody.
Recording the failed tests
In the register, as unknown, with a date and the result. Otherwise the same explanation returns in eighteen months and is tested again, or worse, is not.
A line saying "tested March 2025 against four years of daily rainfall; no effect above 2%" ends the discussion permanently.
The one that is genuinely untestable
Competitor behaviour, mostly, because the data does not exist. This belongs in scenarios rather than in the register, and saying so is better than pretending it has been assessed.
The explanation that is correct
Sometimes the offered cause is real, and testing confirms it. That is the best outcome and it should be said plainly, because a register that only ever disproves things stops being used.
The point of the test is not scepticism. It is that the belief acquires a size, a trigger and a place in the forecast.
Who should run the test
Whoever offered the explanation, with the planner supplying the data. A test run by the planner alone arrives as a contradiction; one run together arrives as a finding.
It also tends to produce better tests, because the person who believes the driver usually knows exactly which days to look at.
What the exercise is worth
It removes two or three standing excuses, converts one or two beliefs into measured drivers, and changes the character of every subsequent forecast review — because once the untested explanations are gone, the question "why did this miss" has to be answered with something.