TalkWithStrangersTalkWithStrangers.club Conversation Report
The TalkWithStrangers.club Conversation Report publishes anonymised, aggregated data about how people talk to strangers on our platform. The first edition covers our first full quarter of operation and appears on this page when that quarter ends. Until then, this page sets out exactly what we will measure and how.
Try TalkWithStrangers.clubWhat We Will Publish Every Quarter
| Metric | Question it answers |
|---|---|
| Median chat length by mode | How long do text, voice and video chats last? |
| Top 25 interest tags | What do strangers bond over? |
| Peak hours by country | When does each country chat? |
| Share of chats ending in a friend request | How often does a chat become a connection? |
| Skip rate in the first 30 seconds, with vs without shared interests | Do interests make first impressions better? |
| Opening-message analysis | Do questions outlast "hi"? |
| Mode switching | How often do text chats move to voice or video? |
Methodology
- Anonymised: no message content is published; only counts and durations.
- Aggregated: no figure based on fewer than 1,000 chats.
- Period: each calendar quarter.
- Sample: all completed chats in the period, excluding bots and banned accounts.
- Review: figures checked by two team members before publication.
Why We Publish
Too many chat sites quote huge numbers with no source. We would rather publish smaller, real numbers with the method attached.
For Journalists
Request early access or raw aggregates at [email protected].
Metric Definitions
| Metric | Exact definition |
|---|---|
| Chat | A match where both people were connected for at least 5 seconds |
| Chat length | Time from connection to either person skipping or leaving |
| Mode | Text, voice or video at the start of the chat |
| Mode switch | A text chat that moved to voice or video by mutual agreement |
| Early skip | A skip within the first 30 seconds |
| Friend request rate | Share of chats where at least one friend request was sent |
| Mutual friend rate | Share of chats where both people accepted |
| Interest match | Both people shared at least one interest tag |
Research Questions
- Do chats with a shared interest last longer than random ones?
- Do opening questions lead to longer chats than "hi"?
- How often do text chats move to voice or video?
- When are people from different countries most active?
- Which interests are most common, and which lead to friend requests?
- How do chat lengths differ between text, voice and video?
Privacy and Ethics
- No message content, voice or video is analysed for this report; voice and video are never recorded.
- Opening-message analysis uses only an automated classification (for example, question vs greeting) without storing or publishing text.
- All figures are aggregated across at least 1,000 chats.
- No individual user can be identified from published data.
- Data is drawn from the systems that already run the service; nothing extra is collected for research.
Publication Schedule
| Edition | Covers | Published |
|---|---|---|
| First | First full calendar quarter of operation | Within 30 days of quarter end |
| Following editions | Each calendar quarter | Within 30 days of quarter end |
How We Check Figures
Two team members independently compute each published number from the same query, compare results, and resolve differences before publication. Methods changes are explained in each edition.
Data Requests
Researchers and journalists can request additional aggregated cuts of the data. We will not share individual-level data.
Why Publish Our Own Data
Most chat services publish no data at all, or publish large user numbers without definitions. We think a smaller set of carefully defined, verifiable figures is more useful to readers, researchers and journalists, and keeps us honest about how people actually use the service.
Hypotheses We Will Test
| Hypothesis | Why we think it might hold |
|---|---|
| Shared-interest chats last longer than random ones | A shared topic gives both people something to talk about |
| Question openers lead to longer chats than greetings | Questions give the other person something to answer |
| Late-night chats last longer than daytime chats | Fewer distractions and more relaxed users |
| Voice chats last longer than text chats | Real-time talk is harder to drop mid-sentence |
| Language-exchange chats have higher friend-request rates | People want regular practice partners |
We will publish results whether or not they support these hypotheses.
Limitations We Will Report
Our users are adults who chose to use a random chat service, so results won't represent everyone. Chat length is an imperfect measure of a good conversation. Seasonal effects, such as exam periods and holidays, can change patterns from quarter to quarter.
Citing the Report
Please cite as: TalkWithStrangers.club Conversation Report, quarter and year of the edition, talkwithstrangers.club/research/conversation-report, and link to the edition you used.
What You Can Use Today
Until the first edition is published, our evidence review summarises the strongest published research on conversations with strangers, and our loneliness statistics collects the latest global data. Both link every figure to its original source.
Questions We Want Your Help With
If you're a researcher interested in conversation, loneliness or online safety, tell us which aggregated measures would be most useful. We prioritise questions that can be answered without collecting any additional personal data.
Example of How a Result Will Be Presented
| Measure | Shared interest | No shared interest |
|---|---|---|
| Median chat length | Reported in minutes | Reported in minutes |
| Early skip rate | Reported as % | Reported as % |
| Friend request rate | Reported as % | Reported as % |
Each table will state the number of chats it is based on and the period covered.
Commitments
We will publish every quarter, including quarters where results are unremarkable, and we will correct any published figure if we find an error, noting the correction on the page.
Frequently Asked Questions
At the end of our first full quarter of operation.
No. Only anonymised counts and durations.
Contact [email protected] for aggregated cuts; individual-level data is never shared.
Every calendar quarter, within 30 days of quarter end.
The first edition needs a full quarter of operation so figures are meaningful.
Yes, as counts and durations only; calls are never recorded.
Yes, through [email protected].
Yes, and each correction will be noted on the page.
It uses our own platform data; we publish definitions and methods so others can judge it.
Two team members compute each number independently before publication.
It will cover all users, with country breakdowns where samples are large enough.