What Counts as Evidence That a Public Building Repair Was Done

Spending public money creates an obligation that spending private money does not: at some
point, someone is entitled to ask you to demonstrate that the work happened, and a confident
recollection is not a demonstration.

The question usually arrives long after the fact and from someone who was not there — a new
manager, a committee, a resident who does not believe the repair was made, an internal review.
By then the crew has moved on and the only thing that can answer is what was recorded at the
time.

Four facts, every job

The minimum that answers almost every future question:

1. Before — the condition that prompted the work.
2. After — the same view, once finished.
3. Who — the individual or contractor who did it.
4. When — the date, ideally the time.

Four facts. If a process captures those reliably on every job, it is doing its work. If it
captures twelve fields on some jobs and nothing on others, it is not.

The same framing, twice

The single most common mistake is a close-up "after" photo that could be anywhere. A tight shot
of a clean pipe proves that a clean pipe exists somewhere in the world.

Train one habit: stand in the same place for both photographs. Include a fixed reference — a
door, a window, a sign, a corner — so the two images are obviously the same location.

This costs nothing and turns a pair of photos from decorative into evidential.

Capture location automatically where you can

Public estates have near-identical spaces by design. Photographs of a corridor in one building
look exactly like photographs of a corridor in another.

Where devices can record location alongside the image, let them, and keep it attached to the
record. It removes an entire category of dispute that is otherwise unresolvable, and it costs
the crew nothing because it happens without anyone doing anything.

Where that is not available, write the building and room into the record at the time. Not
afterwards — the whole point is that afterwards is when detail is lost.

Make the evidence a by-product, not a task

This is the part that determines whether any of it actually happens.

A crew that finishes a job, packs up, drives to the next site and is then expected to fill in a
form will do it for a fortnight and then stop. Not from unwillingness — because the form is
detached from the work, and detached tasks lose.

The processes that survive are the ones where producing the evidence is how you signal the job
is finished. Sending the photo is the completion. Nothing separate to remember, nothing to do
later, nothing that can be deferred to the end of a shift that never has a quiet moment.

Where crews already communicate by phone, the shortest path is usually to structure what they
already send rather than to introduce a second tool. That is the design principle behind
systems like Laksana, and it is worth applying whatever tooling you land on.

Record what was NOT done

An often-missed record: the job that was attended and could not be completed. Part unavailable,
access denied, further work needed, safety concern found.

These are frequently the most important entries in the log, because they are the ones that come
back. A job marked simply "not completed" with no reason will be re-raised, re-attended, and
fail again for the same cause nobody wrote down.

One sentence of reason converts a repeating failure into a queued decision.

Keep evidence attached to the asset

Filing by date answers "what did we do in March". Filing by asset answers "what have we ever
done to this boiler", which is the question that actually informs decisions about repair versus
replacement.

Wherever evidence is stored, make sure it can be retrieved by location or asset as well as by
time. If your storage cannot do that, you have an archive rather than a record.

Decide retention deliberately

Keep evidence long enough to cover the questions that realistically arise: a full budget cycle
at minimum, and considerably longer for anything structural or safety-related.

Set the period on purpose and write it down. The two failure modes are equally common —
discarding evidence before the question arrives, and accumulating everything forever until the
store is unsearchable and functionally empty.