Resoniqa

Structural measurement from ordinary phones

Your building has a frequency. Your phone can hear it.

Left still and charging overnight, a phone records the building moving. Enough of those recordings, from enough phones, and the frequency of the building they sit in can be measured and followed.

One email, when the app reaches the App Store and Play. Nothing else, ever.

RecordingChannel HNZ, one of three50 Hz5 minute window

This is the scale of it: a building at rest moves by hundredths of a metre per second squared. Illustrative, not a measurement of any building.

What your phone does at night

Five minutes, while you are asleep and it is on charge.

The app waits until the phone is still, charging, and at the building you registered, then records for five minutes. That is the whole of it. It does not run during the day and it does not follow you around.

Anything recorded while the phone moved never leaves the device. Handling swamps a building’s vibration by about two orders of magnitude, so a segment with movement in it is not weak data, it is the wrong data.

Recording
5 minutesOnce a night, inside a stable overnight window
Sampled
50 HzAcceleration on three axes, with orientation and pressure alongside it
Uploaded
0.18 MBSmaller than a photograph, once a night

One logging test — all three components

XX.PHONE
Window
15:15:06 — 15:15:29 GMT+2
Rate
100 Hz
Duration
23.2 s
Samples
2,320

HNZZ

Peak
3.76
Mean
1.51

HNXX

Peak
3.64
Mean
-1.34

HNYY

Peak
3.54
Mean
9.48

Gate

  • Kept
  • Rejected
15:15:0615:15:1715:15:29
  • The dimmed sections at either end are the phone being handled. The gate on the device throws them away before anything is uploaded, because a phone in a hand is measuring the hand. The dashed block in the middle is different: samples the log lost, marked rather than filled in.
  • The mean of each component is mostly gravity. A phone at rest projects it differently onto each axis, which is why the three sit at three offsets.

What leaves your phone

Four checks, then five minutes. Nothing else.

You register your building once, when you install the app. From then on the phone decides for itself whether tonight counts.

Every night, before it records

  1. Still

    Face down and not moving

  2. Charging

    On power, so nothing is taken from your battery

  3. At your building

    The phone compares where it is against the address you registered

  4. Night window

    Inside the quiet hours, when a building is at rest

  5. It records

All four, or the night is skipped. A phone somewhere else is measuring somewhere else, and that reading would be worse than none.

Which sensors it reads

Accelerometer
3 axes · 50 Hz
The vibration itself, sampled fifty times a second on each axis
Gyroscope
3 axes · orientation
Which way the phone is lying, so its axes can be related to the building’s
Barometer
1 value · pressure
Roughly which floor it is on, from air pressure
Microphone, camera, contacts, photos
—
Never read, and never asked for. There is no permission prompt for any of them, because there is no code that would use one.

The boundary

Never leaves the phone

  • Where you are. The address check happens on the device; the answer is yes or no, and the position behind it is never sent
  • Anything recorded while the phone was moving

Leaves the phone

  • Once, at registration: the address of your building, so the measurement has somewhere to go
  • Each night: five minutes of acceleration, the phone's orientation, a pressure reading, and the time it started
  • A random identifier for the phone, and nothing attached to it

Nightly recording is being reviewed against Italian and EU data-protection law before the app ships. Whatever that review changes will change here first.

Your control

You decide when, and where.

Nothing starts without you. You choose the building, you grant each permission one at a time, and you can stop it for a night or for good.

There is no account to close, because there was never one to open. Your phone is one random identifier, generated on the device the day you install the app, and everything we hold about you hangs off it.

Pause any night
One tap and the app sits the night out. Nothing is recorded and nothing is sent.
Change or remove your building
The phone stops recording the moment you do. A building you no longer live in is a building you no longer measure.
Remove your phone from the record
The identifier goes, and every link between your phone and what it recorded goes with it. That cannot be undone or rebuilt. The recordings themselves stay with the building, with nothing left in them that points back to a phone.

A building’s history is the measurement. A frequency means something only against what the same building read last winter, so the recordings are kept under the building rather than under the phone that made them.

Sample device record

Held

Identifier
3f6c1a92-7b40-4e51-9c8d-2a5f0e6b71d4—
Building
MI-0221—
Recordings sent
184—

Last 60 nights. Short ticks are nights the checks did not pass.

First seen
2026-03-04—

Not held

Name
—
Email
—
Phone number
—
Your address
—
Where you have been
—

Kept with the building

MI-0221 · 184 recordings

Held as part of the building’s history, with nothing in them pointing back to the phone that recorded them.

The whole record. The dashes are not fields we left empty; there are no fields behind them.

What your five minutes becomes

One night is noise. A building’s worth of nights is a measurement.

Your recording is pooled with every other phone in the same building. Noise falls with the square root of how many there are, which is why a block with thirty contributing phones reads more clearly than one with three.

What comes out is a single number, the building’s natural frequency, corrected for temperature, humidity, occupancy and season, drawn with the uncertainty that is left after the correction.

Hz4.505.00
Sep 25Dec 25Apr 26Jul 26

One building, 48 weeks. Grey is what the phones reported. Teal is what it means.

  • Raw reading
  • Corrected
  • Residual band

How it works

Six stages, from a phone on a bedside table to a queue for an engineer.

  1. 01

    Capture

    Acceleration on three axes at 50 Hz for five minutes, with the orientation needed to relate them to the building.

  2. 02

    Gate

    Still, charging, at the registered building, inside the night window. All four, or the phone records nothing.

  3. 03

    Associate

    The phone already knows its building from registration, and refuses to record anywhere else. A segment arrives attributed, not guessed at.

  4. 04

    Aggregate

    Nightly records for one building are pooled. Noise falls with the square root of how many there are.

  5. 05

    Correct

    Temperature, humidity, occupancy and season are removed. The raw value and the correction are both kept.

  6. 06

    Score

    The corrected reading is compared against the building’s own history, and the inspection queue is ordered.

Illustrative. Nothing drawn here is a measurement of any building.

The problem

A building’s frequency moves with the weather.

Temperature, humidity and how many people are inside all shift the number, by 4 to 13 percent, with nothing wrong. Real earthquake damage shifted one instrumented building by 21 to 24 percent.

Between those two figures sits every reading that matters. No published method separates them, and that is exactly the problem Resoniqa is built to solve. Without the correction there is no continuous score, only an alarm that fires when the weather turns.

Change in natural frequencyPercent
Environmental variation, no damage present4 to 13 percent
Documented earthquake damage21 to 24 percent

Nothing published reads this

0102030

What you see about your building

The same record your council or your insurer sees.

Not a grade and not a verdict. The frequency we measured, how far it can be relied on, how many phones contributed to it, and whether it has moved enough to be worth an engineer’s time.

A building with too few phones reads no data, and says so. That is a result, and it is shown rather than hidden, because a blank is easily mistaken for nothing having changed.

An inspect state means the reading has moved far enough to be worth an engineer looking. It is a position in a queue, not a finding about the building.

Sample record

Via Ripamonti 118, Milano

MI-0221 · 12 storeys

Corrected frequency

4.82Hz±0.094

Frequency trend, 24 weeks

Corrected weekly readings. The step is the move that put this building in the queue; the figure above carries the uncertainty left after correction.

Where this reading sits

  1. Stable

    The reading sits where this building’s own history says it should.

  2. Watch

    It has moved, but not far enough to spend an engineer on.

  3. Inspect

    It has moved far enough to be worth an engineer’s time.

    This building

Behind the reading

Confidence
High
Contributing phones
19
Last measured
3 days ago

Who else this is for

Somebody pays for this, and it is not you.

They buy the portfolio view: every building they are responsible for, ordered by what changed.

Insurers pricing the catastrophe cover Italian property owners are now required to hold. Municipalities and civil protection deciding where a fixed inspection budget goes. Portfolio owners, and the engineering firms who do the inspecting.

What each of them takes from it

Insurers and reinsurers

One building, on one day

A building-level input on the day a policy is written, not on the day of the last earthquake.
Municipalities and civil protection

Every building, every week

Continuous screening across a city, so a fixed inspection budget goes to the buildings that moved.
Portfolio owners and engineers

Ordered by what moved

A standing queue of what to look at first, refreshed weekly and re-run after any felt earthquake.

The app is free to run, and stays that way. A resident is not the customer, and nothing about the reading changes depending on who is paying for the view of it.

The boundary

What Resoniqa does not do.

Three shortcuts that would make the output look stronger, and are not taken. A screening layer that overstates itself is worse than no screening layer at all.

No data is a result

Not this

Dropped from the list, so the list looks complete

This

Kept in place, drawn as absence

A building with no contributing devices returns no data. It is never reported as no change, and never dropped from a list.

The correction stays visible

Not this

4.82

One number, with the model behind it hidden

This

The raw reading, the corrected one, and the band left over

The raw value, the correction applied, and the uncertainty that remains are all shown. A single number with an invisible model behind it cannot be audited.

It orders a queue

Not this

PASSED

A verdict on the building

This

An order of work for an engineer

Resoniqa prioritises inspections for a structural engineer. It does not certify a building and it does not replace an inspection.

The app

Get the app when it lands.

It is not out yet. Leave an email and you will hear once, on the day it is.

One email, when the app reaches the App Store and Play. Nothing else, ever.