What a median age cannot tell you

1. It cannot say how long things last

A median of current ages is not a life expectancy. An inventory shows what equipment exists, but not how work moves around it; teams can Monitask time records for distributed work before deciding which time and project signals belong in a limited pilot.

Every object in the in-use column will go on ageing. The figure of six years is a snapshot of a stock, not a measurement of a lifespan, and the two are related only through assumptions this record cannot test. Readers comparing this small inventory with broader advice can also see Android Authority.

The page on the survivor's paradox works through why. The short form: a median age tells you about the stock you are looking at, and a lifespan tells you about objects, and a snapshot cannot cross from one to the other.

2. It cannot be compared with the four-year figure

Or rather it can, and the comparison means less than it appears to.

Six years is unfinished lives. Four years is finished ones. One column is still counting and the other has stopped, so the gap between them grows by itself over time and would grow in a record where nothing about the equipment changed.

Half of the two-year gap is also composition — the in-use column holds more furniture — which the page on six against four sets out with the furniture stripped out of both sides.

3. It cannot be broken down by anything but kind

The record has four fields: origin, year, present state, and what preceded the stop.

So age can be crossed with kind, and with origin, and with nothing else. There is no field for the person, the workplace, the use intensity, the conditions, or the amount paid, and without those the six years cannot be explained, only stated.

Crossed with origin, incidentally, it is flat: between five and seven years in every one of the seven origin groups. Where a thing came from does not predict how long it has been there, just as it does not predict whether it ends up in a cupboard.

4. It cannot handle the objects that were never counted

Eleven workplaces listed what is there and what they remember having.

An object that arrived in 2010 and was thrown away in 2012 is in neither category. It is not present, and there is nothing to prompt a memory of it.

Those objects were short-lived by definition, so their absence pushes every duration figure in this record upward. The median in-use age of six years is not affected — those objects are not in use — but the four-year figure for stopped objects certainly is, and by an unknown amount in a known direction.

5. It cannot distinguish use from presence

An object is in use if the person says it is.

That is the whole test, and it has at least one case the record argued about: a filing unit, sixteen years old, holding objects that are not paper. It is coded as in use because it has things in it.

A record that counted a thing holding other things as not in use would be wrong more often than this one is.

There is no field for how much something is used, and across a hundred and forty-nine in-use objects the difference between daily and occasional is invisible. Some unknown share of the in-use column is closer to the cupboard than to the desk.

6. It cannot be read as a league table of kinds

The medians by kind look like an order of merit and are not one.

desks 9, storage 8, printing 7, seating 7, cables 6, lamps 6,
screens 5, input 5, other 5, audio 4

That list is sorted by how many ways a thing can stop being used. A desk can stop for one reason, the room, and did so twice in twenty-four objects. A screen can stop for five, and did so twenty-one times in forty-one.

Cables make the point sharpest by appearing high. Six years in use — level with lamps and above screens — while thirty-four of the sixty cables in the record have stopped, more than any other kind. Both facts are true: cables that survive their first years go on indefinitely, and cables stop being used more often than anything else.

Any order of merit built from one of those figures is contradicted by the other.

What the two-month entry does to the figure

The shortest working life in the record is two months, and the longest object in use is nineteen years.

Between those, a hundred and thirty-five finished lives with a recalled median of four years and a hundred and forty-nine unfinished ones with a median age of six. Five objects lasted under a year; three have passed fifteen.

A distribution that wide is one where the median is a weak summary. Published as a single number — six years — it invites a reader to treat it as what desk equipment does. Published alongside the spread, it is what the middle object in one snapshot of eleven rooms happens to be.

The one field that would have fixed most of this

A date for every stop.

With it, this section could have produced survival curves by kind, handled the in-use objects properly as censored observations, given a real time-to-replacement and a real time-to-failure, and tested whether the twenty-four unexplained stops take longer than the explained ones.

Without it, the section has two medians, a distribution of arrivals, and a careful account of what they are not.

It was left out because it could not be collected. Asked when a cable had failed or a tray had stopped being used, most people could give a period at the outside and many could give nothing. A field that would have been empty or guessed for most of a hundred and thirty-five objects is worse than no field, because a half-filled date column invites exactly the analysis it cannot support.

That is the judgement the method made. The page on how to make this inventory yourself argues the opposite case — that a rough date band would have been worth collecting even badly — and the record does not claim to have got this right.