How to Document Neighbor Noise

The instinct is to catch the noise at its worst and show someone the number. It almost never works, and it is worth understanding why before spending three weeks doing it.

A single reading has no context. It cannot show how often, how long, at what hour, or whether the room is like that most nights. Whoever you take it to — a landlord, a managing agent, a council officer, eventually perhaps an adjudicator — is being asked to act on a pattern, and one number is not a pattern. The record is the evidence. The reading is just one entry in it.

Which thresholds actually apply

Not the occupational ones. This is the single most common mistake, and it weakens an otherwise good complaint.

Workplace limits sit around 85 to 90 dBA and are about hearing damage over a career. A flat that never gets anywhere near them can still be unlivable, and quoting them invites the reply that your readings are well within safe limits — which is true and beside the point. If you want to understand why they do not transfer, that is OSHA noise limits.

The figures written for homes are far lower. A 1974 EPA report, summarised by NIOSH, named 55 dBA outdoors and 45 dBA indoors as the levels above which noise begins to interfere with speech and become an annoyanceNIOSH. Those were never issued as a standard and no US agency enforces themNIOSH, but they are the right order of magnitude for the question, and they are a citable reference point rather than a number you made up.

For hearing, meanwhile, sound at or below 70 dBA is unlikely to cause hearing loss however long it lastsNIDCD. Most domestic noise complaints live entirely below that. Say so yourself, before someone says it to you: your case is about sleep, speech and quiet enjoyment, not about damage.

The rule that actually binds is local. Ordinances, tenancy agreements and statutory nuisance provisions vary by city and by country, they often specify a time of night rather than a decibel figure, and they are what an officer will apply. Find yours before you start, because it tells you what to record.

What makes a record credible

Consistency, boringness and volume of entries. In that order.

Same everything. Same spot in the room, same height, same device, same orientation, doors and windows in the same state. Most of the error in a phone reading is systematic, so it largely cancels between two readings taken the same way — and that means your differences are far more trustworthy than your absolute numbers. Vary the method and you throw that away.

Long enough to be a level rather than an event. Let the meter run for several minutes. Note where the number sits, not the highest thing it caught. A peak is a door slamming; a level is a party.

Both states. Record the quiet baseline as well as the noise. "62 at 11pm on Friday" means little; "38 on a normal night, 62 on Friday" is an argument.

Every night, not only the bad ones. A log with fourteen entries of which nine are quiet is more persuasive than a log with five entries all of which are loud, because the second one looks selected. It also happens to be true, which matters if anyone checks.

Dates and times, written at the time. Reconstructed logs read as reconstructed.

What to write down

For each entry: the date, the start and end time, where you were standing, what the level sat at, what the baseline was that night, what you could hear (bass, voices, footfall, machinery), and what you did about it. Keep it to one line each. A fortnight of one-line entries is a document; a paragraph per night is an essay nobody will read to the end.

Then, once, at the top: what device you used, what the reading is, and what it is not.

The part of that job a page can do for you is the timing. The overnight noise monitor records the level once a second while you sleep and hands back the list of times, which you can then annotate with the half nobody's microphone captures: what you could hear, and what you did about it.

Say what your numbers are

This is the part that decides whether the record survives being challenged, and it costs one sentence.

A browser or phone reading is an uncalibrated estimate. Sound apps have been found to scatter by roughly ±10 dB against a reference instrumentKardous & Shaw, and NIOSH's own position is that where a sound level meter is not available, apps can give a measure of area noise but may not comply with regulatory requirementsNIOSH.

So write it as what it is: "uncalibrated estimates from a phone, taken from the same position each night, not a compliance measurement." That sentence makes the rest of the document harder to dismiss, not easier — the alternative is presenting figures as though they were instrument readings and having someone else point out that they are not. Exactly how wide the band is, and why, is how accurate the reading is.

If the situation escalates to something formal, the reading someone will act on will come from an officer's own equipment. Your log's job is to get that officer to come at the right hour, which is a job it can do well.

How long to log before doing anything

Long enough that the pattern is visible, and not so long that you have spent a season on it. Two weeks is usually the right order: it covers both weekend nights twice, catches a normal working week, and is short enough that you will actually finish it.

There is one exception worth making. If the problem is genuinely intermittent — loud once a month — a two-week log will show nothing, and waiting for it to show something is how people end up with six months of entries and no case. In that situation log the incidents as they happen, keep recording quiet baselines in between, and let the record cover a longer span with fewer entries. The structure is the same; only the sampling changes.

Start the log before the first conversation with anyone, not after. A record that begins the day you complained looks like a record built for the complaint.

What not to do

Do not only record the worst night. It is the most tempting and the most damaging. A selected record invites the question of what you left out.

Do not round up, and do not quote a peak as the level. If the meter touched 71 and sat at 58, the number is 58. Anyone who checks will find the difference, and then everything else in the log is in question too.

Do not compare devices. Your phone against a neighbour's phone against a figure from a table of typical values is three unrelated numbers. Same device, every time.

Be careful about recording audio. What you may lawfully record of other people, and where, varies a great deal by jurisdiction — and a recording of voices is a different object, legally and socially, from a level log. The meter on this site never produces one: audio is read in the browser, turned into a number, and discarded frame by frame. Nothing is stored and nothing is uploaded.

Bringing it together

Two weeks of consistent entries, a stated baseline, an honest description of the equipment, and the local rule that actually applies is a strong document, and it can be assembled with nothing but a phone and a notebook.

If part of the problem is the sheer duration of the exposure rather than its level — a workshop next door running all day, for instance — the noise dose arithmetic is a way to describe that in a form someone else can check, with the caveat that it is a workplace model borrowed for a domestic situation rather than a rule about your flat.

The online decibel meter on this site is built for this pattern of use: it estimates the level in dBA, saves finished sessions on your device with the time and duration attached, and exports them so the series is the evidence rather than your memory of it.