September 2026 edition
The data behind
The Call Report.
A fixed observation window, explicit denominators, and findings you can inspect.
One window, ten analyses
This edition uses calls that started from August 10, 2026 at 00:00 UTC up to, but not including, September 7 at 00:00 UTC. All ten articles analyze this same historical window. Publication and social-sharing dates do not change the underlying observation period.
We analyzed aggregated Frontdesk call records and recorded caller speech. The published charts and downloads contain summary statistics, not recordings, transcripts, caller numbers, or named customer results.
How the sample was selected
The initial extraction contained 169,673 records. After the exclusions below, 147,885 records remained across 272 organizations. Organizations needed at least 100 retained calls in the window to enter the broad analysis.
| Selection step | Records |
|---|---|
| Initial in-window extraction | 169,673 |
| Duplicate records removed | 3 |
| Playground calls removed | 1,700 |
| Deleted calls removed | 2,245 |
| Observer / voicemail records removed | 997 |
| Records from organizations below the volume threshold removed | 16,843 |
| Final eligible sample | 147,885 |
The property-management subset was manually curated using available business descriptions. It contains 27,103 calls across 19 organizations, including residential rental management, student housing, single-family, and mixed management businesses. It should not be described as a strictly multifamily sample or as representative of the industry.
What the call records can tell us
Call timestamps support timing and recontact analysis. Recorded duration supports a distribution of logged call time. Recorded caller speech supports explicitly defined topic and word-count screens. System events can indicate that a transfer was logged.
These fields do not, by themselves, verify a tour booking, lease, completed repair, successful human pickup, satisfied caller, billed cost, or a cause of disconnection. We keep those distinctions in the article text and chart notes.
Topic classification and transcript limits
Topics are deterministic dictionary matches in caller speech, rather than inferred from the assistant’s script. We reviewed samples to refine the rules. Broad follow-up language produced false positives in the initial application-status screen, so the published measure uses a narrower application-specific phrase definition and excludes explicit job-application language.
Dictionary matching can miss indirect phrasing and can include references to older issues. Selected maintenance categories overlap. A tour mention may be a reschedule or a discussion of an existing appointment. The counts are mentions under those definitions, not verified business outcomes.
Time zones, repeat callers, and routing
Local-hour and weekday analyses use the agent’s configured time zone at extraction. We use weekday 9 a.m.–5 p.m. as a reference schedule, not the actual operating hours of every property. The UTC window creates partial local dates at its boundaries.
Repeat calls are grouped by organization and valid caller number. Shared numbers can represent multiple people; callers may be residents, prospects, vendors, or staff. Repeat intervals run from the start of one call to the start of the next observed call. The first observed call in a pair has no preceding in-window interval.
Forwarding policies influence which calls reach Frontdesk. The sample does not observe every call handled by an organization’s staff or other phone systems. Changes in routing could change the measured distribution without a change in underlying customer demand.
Concentration and sensitivity checks
The largest organization contributes 40.2% of property-management records. We therefore repeated selected measures with that organization removed. The remaining subset contains 16,200 calls.
| Measure | Full PM sample | Without largest operator |
|---|---|---|
| Tour-topic calls outside reference hours | 38.5% | 39.5% |
| Call share from numbers with 5+ calls | 42.1% | 34.8% |
| Repeat calls within 10 minutes | 44.4% | 44.4% |
| Monday volume premium over Friday | 35.6% | 23.9% |
The greeting analysis uses one common cohort of 15 organizations with at least 50 calls in each greeting-length group. It compares pooled rates with equal-organization means to show how the choice of weighting changes the summary. This is a descriptive comparison, not a randomized experiment.
How to interpret the precision
Displayed percentages are generally rounded to one decimal place. Small differences in sums can come from rounding. The reported values summarize the selected records; decimal precision should not be interpreted as precision for the wider property-management market.
We do not treat calls as independent random samples of the industry, report population-wide confidence intervals, or infer causal effects from these aggregates. Calls are clustered within businesses and caller histories. Each article specifies its own denominator, exclusions, and limitations.
Downloads and attribution
Every article includes a PNG, editable SVG, and a CSV of the displayed numbers. The aggregate calculation file provides the underlying summary values. Raw customer records are not distributed.
Charts may be shared with attribution to Frontdesk Insights and a link to the relevant article. Preserve the observation window, source, and sample notes when using the image.