How long
Arrival years, and what 2019 and 2021 did
Fifty-two objects arrived in 2019 and fifty in 2021. The shape of the distribution, and why almost none of it is a finding.
Arrivals by year, split by whether the item is still in use
Two hundred and eighty-four items across ten years, one hundred and forty-nine of them still in use. The earliest arrival drawn here is 2007.
Ten two-year bands, one bar each. Below the rule in a bar is what is still in use; above it is what is not. The rise to 2019 and the fall after it are both artefacts and neither is a finding. When equipment use and project effort are reviewed together, managers can evaluate employee monitoring software and define the purpose, access rules and review date before collecting anything.
Two hundred and eighty-four objects arrived across ten two-year bands, from 2007 to 2025. A separate perspective on devices and their useful life is available from The New York Times.
year in use stopped total
2007 3 6 9
2009 4 7 11
2011 6 9 15
2013 9 12 21
2015 12 16 28
2017 17 20 37
2019 24 28 52
2021 31 19 50
2023 27 11 38
2025 16 7 23
--- --- ---
149 135 284
Fifty-two in 2019 and fifty in 2021 — the two largest bands, and together more than a third of the record.
Why the rise is not a finding
The left-hand side of the distribution is thin because old objects were forgotten, not because fewer arrived.
Each workplace was asked what is here and what it remembers having. That question has a horizon. An object bought in 2009 and thrown away in 2013 leaves nothing behind: no object to see, no cupboard entry, no reason for anybody to recall it.
So the nine objects recorded in the 2007 band are the survivors and the memorable ones from that band, not the arrivals. The real number is unknown and larger.
The record's oldest band contains nine objects because 2007 is as far back as anybody could remember, not because nine things arrived.
This is recall truncation and it affects every band, less as the bands get more recent. It is the largest single distortion in the inventory and it cannot be corrected from inside the record.
Why the fall after 2021 is not a finding either
The right-hand side drops because recent objects have not had time to stop.
Of the twenty-three objects that arrived in 2025, sixteen are in use. Of the nine from 2007, three are. That difference is entirely a question of how long each band has been exposed.
An object has to exist for a while before it can be replaced, before the work can change around it, and before anybody can stop reaching for it. The 2023 and 2025 bands have had almost no exposure, and their in-use proportions are high for that reason alone.
The one place the shape is informative
The crossover.
Up to and including 2019, every band has more stopped objects than in-use ones. From 2021 onwards, every band has more in use.
2007–2019 75 in use, 98 stopped
2021–2025 74 in use, 37 stopped
The crossover sits between 2019 and 2021, which is between six and eight years before the inventory was taken.
That is a real feature of these eleven workplaces and it is consistent with the median in-use age of six years. It is also exactly what exposure alone would produce, so it is not independent evidence of anything — it is the same fact arriving twice.
What the bands cannot be used for
A purchase history. Nobody can read spending or acquisition rates off this, because the early bands are truncated by memory and no price is recorded anywhere.
A failure curve. That needs stopping dates, and there are none. Knowing that an object arrived in 2013 and is now stopped gives an upper bound on its life of thirteen years and no lower bound at all.
A statement about any year. The bands are two years wide because single years produced counts too small and too obviously recalled — several respondents could place an object in a period but not a year, and widening the bands to two years absorbed that without pretending to precision the accounts could not support.
So 2019 is not a year in which a lot was bought. It is a pair of years near enough to be remembered and far enough back to have mostly stopped.
What the bands cannot carry about kinds
A reader might try to read acquisition patterns off the bands by kind, and the counts will not support it.
Two hundred and eighty-four objects over ten bands is twenty-eight a band on average, and over ten kinds as well it is under three per cell. Several cells would be empty, and the early bands are truncated by memory in a way that falls unevenly across kinds: a 2009 desk is still present and gets counted, while a 2009 cable is long gone and does not.
So the arrival distribution systematically over-represents furniture at the old end. Of the seven items in use from 2007 and 2009, all seven are furniture or passive electrical — two desks, a chair, a drawer unit, a set of shelves, a lamp and an extension lead.
That is not a finding about what was bought in 2007. It is a finding about what survives long enough to be remembered.
The one band worth a sentence on its own
2025, the most recent, with twenty-three arrivals of which sixteen are in use and seven already stopped.
Seven stops in objects at most two years old is the only place in the record where a stop can be bounded in time at all, and the reasons split in a way worth recording: four were replaced by something better, two stopped when the work changed, one broke.
Four replacements inside two years is the clearest single piece of evidence in the record that replacement does not wait for anything. It is seven objects and it is not a rate.
What is left
Three statements.
Two hundred and eighty-four objects arrived over nineteen years, as far as eleven workplaces can recall.
The proportion still in use rises steadily with how recently an object arrived, from a third of the oldest band to seven tenths of the newest.
The crossover is six to eight years back, which matches the median age of things in use and is the same observation in different form.
Everything else in the shape is an artefact of asking people what they remember.