Time, ageing and data versioning

Equipment ages, observed lifetimes change, electricity mixes evolve and datasets are revised. Every assessment must identify its observation period, retained lifetimes, factor reference periods and method version.

Why time matters

Changing the electricity factor's year can change the result without any change in activity. Similarly, allocating manufacturing over three or five years changes the annual share. Comparability requires these choices to be explicit.

Four time dimensions

Dimension	Question	Example
Observation period	Which activity period is assessed?	Calendar year or quarter
Lifetime / allocation	How is embodied impact distributed?	Four-year nominal life; 5.2-year estimate
Data reference date	Which period do background data represent?	A specified electricity month/year or inventory vintage
Method version	Which formulas and conventions were used?	Actual method and factor dataset commit identifiers

Lifetimes

Definitions

Concept	Definition	Use
Age	Elapsed time since entry into the inventory	Input to a lifetime estimate; previous ownership recorded separately
Nominal lifetime $TL$	Organisational expected duration	Default where better evidence is unavailable
Observed / estimated lifetime $RL$	Total observed life for retired equipment; explicit estimate for equipment still in use	Allocation duration reflecting available evidence

Accounting schedule versus service LCA

A historical inventory can distribute manufacturing over four accounting periods and then close that entry. A service LCA instead allocates manufacturing over an expected or observed lifetime. Exceeding a nominal age does not automatically make manufacturing zero in that service LCA. These conventions remain distinct. A revised estimate retains its previous versions and specifies whether earlier results have been recalculated.

Proportional correction

For an embodied factor already annualised over a reference lifetime:

$$IF_{emb_i}=TIF_{emb_i}\times\frac{TL}{RL}$$

Both factors are in $U_i/\text{year}$ and lifetimes in years. The correction does not apply again to a total manufacturing inventory.

If $RL>TL$, the annual allocated share decreases; if $RL<TL$, it increases. Manufacturing already incurred is not physically reversed by this allocation change.

Inventory-based lifetime estimate

The historical convention below estimates an allocation duration. It is not a mean of fully observed lifetimes: equipment still operating has an unknown retirement date and is therefore censored. Complete cohorts or survival analysis are preferable for estimating statistical service life.

Population	Equipment state	Retained duration
A	Still present, older than nominal life	Current age
B	Retired before nominal life	Age at retirement
C	Still present, younger than nominal life	Nominal life1
D	Retired at or after nominal life	Age at retirement

$$RL=\frac{\sum_{k\in A}age_k+\sum_{k\in B}age_k+\sum_{k\in C}TL+\sum_{k\in D}age_k}{|A|+|B|+|C|+|D|}$$

Worked example

A fictitious four-phone inventory observed in 2024 uses a nominal lifetime of four years:

Purchase year	Status	Age, years	Population	Retained duration
2018	In stock	6	A	6
2019	In use	5	A	5
2021	Broken / retired	3	B	3
2022	In use	2	C	4

$$RL=(6+5+3+4)/4=4.5\text{ years}$$

The estimated allocation duration is 4.5 years. This does not prove the final lifespan of surviving devices or the quality of maintenance. Calendar time in stock differs from time delivering a service.

Refurbished equipment

Previous use may contribute to total lifetime, provided successive ownership periods and refurbishment are accounted for consistently.

Purchase year	Status	Age in organisation	Prior use	Population	Retained total life
2018	Stock	6	0	A	6
2019	In use	5	0	A	5
2021	Retired	3	0	B	3
2022	In use	2	0	C	4
2023	Stock	1	3	C	7

The estimate becomes $(6+5+3+4+7)/5=5$ years. The last device's seven years assume three previous years plus four additional expected years in the organisation. This is not an observed total life. If the nominal four years refer to total life rather than the second ownership period, adding them would be incorrect. The assessment includes refurbishment impacts and an explicit allocation between successive lives.

Assumed lifetime after refurbishment

First lifeSecond life

3 yearsknown

4 yearsexpected

↑Arrival in the organisation

7 years assumed in this scenario

An estimated lifetime, not seven years already observed. Refurbishment still needs to be inventoried.

Revising nominal lifetimes

A longer retained lifetime lowers the annual share but extends its allocation period. Changes must be justified and versioned; changing an assumption alone is not evidence of avoided production.

Versioning data

Why version data?

Electricity files may be refreshed with annual or monthly reference periods; equipment inventories and manufacturer reports also evolve. Identical activity calculated with different background versions can produce different results. Track those changes rather than treating them as operational improvements.²

Item	Traceability information	Location
Electricity factors	Dataset commit, reference month/year, geography	Assessment metadata
Embodied factors	Source, record identifier, date, boundary	Assessment metadata
Lifetimes	Nominal, observed/estimated, calculation date	Assumptions register
Method	Documentation / calculation version	Assessment report

Comparing versions

Assumptions and limitations

• Multi-year comparisons must distinguish changes in activity, electricity mix, model and data.
• Defaults can become obsolete faster than the methodological structure.
• Lifetime estimation needs complete, historical inventory data. A nominal fallback must remain explicit.
• Maintenance frequency does not prove freshness. Actual reference periods and any extrapolation from incomplete years determine the interpretation.

Footnotes

1. This assumes the equipment will reach the nominal duration. It is an organisational expectation, not an observed outcome. ↩

2. Versioned Digital4Better datasets, the central register and organisational hardware inventories. Digital4Better data catalogue, snapshot 77cae4a. ↩