Skip to main content
Elevyr Research · Pre-Registered Study

When a family calls a senior living community at 9 PM on a Sunday, does anyone pick up?

Nobody has published an honest answer. So we are measuring it, and we fixed every rule in public before placing a single call.

Registered before any data existed

The full protocol, the sample rule, and the analysis plan were frozen and timestamped on the Open Science Framework on July 17, 2026, before the first call.

Read the registration, osf.io/8jx3t
The Gap This Fills

Everyone quotes a scary number. Nobody shows their work.

Every company selling call-answering software in senior living cites a statistic about missed calls. When we traced those numbers back to a primary source, they evaporated. There is no published, sourced measurement of how often a family reaches a live person after hours. The closest figure, Invoca's finding that 53% of senior-care callers reach a person, is measured across all hours on one vendor's own customers, so it cannot isolate the after-hours question at all.

We would rather measure it ourselves and hand you the receipts than repeat a number we cannot stand behind. That is the entire point of this study.

How It Works

One question, asked the same way every time.

A trained caller phones 100 randomly chosen Southern California communities at three set times, and records one thing: did a real, live human being answer?

The field
Licensed senior living communities (RCFE, 36 to 150 beds) across eight Southern California counties. 428 communities in all, every one with a listed phone number.
The sample
100 communities, drawn at random. The seed is a public California lottery number drawn after we registered, so no one could hand-pick which communities get called.
The calls
Each community is called three times in one week: Sunday 8 to 10 PM, Tuesday 8 to 10 PM, and Wednesday 10 AM to noon (the daytime comparison). 300 calls in all.
What we measure
A single yes or no: did a live human answer? A human answering service counts as a yes. A recording, a menu, or an AI voice counts as a no.
The comparison
The same communities, on the same phone lines, weekday morning versus Sunday night. The daytime call is not a nice-to-have. It is what makes the after-hours number mean anything.
Why You Can Trust the Number

Five decisions, made before we knew the answer.

This study is built to survive a skeptic reading it line by line. Here is what is locked, and why.

Fixed in public first

The rules were registered before call one.

The full protocol, definitions, and analysis are timestamped on a public research registry. We cannot quietly change what counts as a yes after seeing the results.

A lottery picks the sample

We could not cherry-pick the communities.

The 100 communities are chosen by a public lottery number that did not exist when we registered, so we could not have run the draw until it looked bad.

A blind caller

The person calling does not know it is ours.

The calls are placed by a contractor who is not told this is for Elevyr or which result we hope for. The outcome is a plain yes or no, not a judgment call.

No thumb on the scale

An answering service counts as answering.

The easiest way to make this look worse would be to score human answering services as failures. We do the opposite. A person is a person.

Published either way

We publish the result even if it hurts our pitch.

All three possible headlines are written in advance, including the one where communities answer just fine. The registration is what makes that a commitment, not a promise.

The Part Most Companies Leave Out

Our conflict of interest, said plainly.

Elevyr sells Sloane, an AI phone agent that answers calls for senior living communities around the clock. A poor after-hours number would help us sell it. We are not going to pretend that away.

That conflict is exactly why this study is run the hard way: registered in public before any call, sampled by a lottery draw, placed by a blind caller, and released with the row-level data so anyone can check our math. No community is ever named, and no community's own result is ever shown to it. You get the aggregate and the receipts, not a gotcha.

Read the frozen protocol and analysis plan on OSF
While the Study Runs

The calls happen this month. The findings publish within 30 days.

The results go public within a month of the last call, whichever way they land. In the meantime, you can read exactly how we are running it, down to the last definition.

No community is ever named · Aggregate results only · Published regardless of outcome