Average Meetup Size by Event Type: What the Evidence Actually Supports

Search for the average size of a meetup and you will be handed a confident number within seconds. Chase that number back to its source and it usually dissolves. This piece does the chasing: what the peer-reviewed data actually says, which widely repeated figures have no traceable origin, and a method for sizing your own event that does not depend on anyone's industry average.

Quick answer

There is no authoritative statistic for the average size of a meetup, because no statistical agency measures it. The strongest published dataset covers 4,418 offline Wikipedia meetups held over roughly two decades and reports a mean of 8.71 attendees and a median of 7, with a standard deviation of 9.28 and a maximum of 206.

That describes one specific community. It is a reasonable anchor for grassroots interest-group meetups, and it is close to useless for a conference or a training day.

A more reliable way to size an event is to work out its four independent ceilings - floor area, facilitator attention, interaction, and demand - and plan for whichever is lowest. Every one of those four can be calculated from things you already know about your own event.

Why there is no official number

It is worth starting with an absence, because that absence explains most of the bad numbers in circulation.

Nobody official counts how many people attend meetups. The US Bureau of Labor Statistics tracks meeting, convention and event planners as an occupation - employment, wages, projected growth - and publishes nothing at all about how many people show up to the events they plan.6 The Events Industry Council's global economic study, the most ambitious attempt to size the sector, counts 1.6 billion business-event participants across more than 180 countries in its 2019 baseline year and puts direct spending at $1.15 trillion. It defines event categories carefully - conventions, conferences, congresses, trade shows, incentive events, corporate meetings - and then reports no participant counts by category and no event count at all.5 You cannot divide one by the other, because the denominator is not there.

So when an article states that the average conference draws 470 people and the average training workshop draws 760, the interesting question is where those digits came from. In researching this piece we tried to trace that specific set of figures - widely attributed to a hospitality consultancy's event-planner survey - through four separate convention-centre feasibility studies. None of them contained the table. What those studies do contain is demand forecasts for one proposed building in one city, and the values differ wildly between cities: one report projects 210 attendees for conferences and 75 for meetings, another 200 and 75, alongside 570 for conventions. Those are projections for a specific venue's business case, not measured industry averages, and they were never meant to travel.

The same applies to the frequently quoted claim that a typical Meetup.com event draws about 15 participants. We could not trace it to any peer-reviewed paper. The study it is usually attributed to does exist and is genuinely useful, but it reports something different, as the next section explains.

This matters practically, not just pedantically. An organiser who plans a twelve-person book club around a "470 average" will over-book a venue, over-order food and read the resulting turnout as a failure. Bad benchmarks do not just misinform - they change decisions.

The data that does hold up

Two published datasets survive scrutiny. They are narrow, and their narrowness is exactly what makes them usable: you know what population each one describes.

Offline Wikipedia meetups: 4,418 events over two decades

The single cleanest measurement of grassroots meetup attendance comes from research published in 2023 in the Journal of Computational Social Science, which assembled and analysed the offline meeting records of the German-language Wikipedia community from 2001 through March 2020.1

Across 4,418 meetups, attendance averaged 8.71 people, with a median of 7, a standard deviation of 9.28, a minimum of 1 and a maximum of 206. Roughly 77% of the gatherings were primarily social; the remaining 23% were work meetings.

What makes this dataset valuable is not that Wikipedia editors are representative of everyone - they are not. It is that the meetups were recorded by the participants themselves as they happened, over twenty years, in a community with no incentive to inflate the count. That is closer to a census of a well-defined population than to a survey of self-selected respondents.

Meetup.com RSVPs: volume without attendance

A 2016 study presented at the International AAAI Conference on Web and Social Media analysed Meetup.com activity in three US cities over roughly four months in 2015.2 The scale is substantial: New York alone contributed 11,884 groups, 495,581 events and 3,473,822 RSVPs, with comparable volumes in Chicago and San Francisco.

The paper does not report an average number of participants per event. Dividing RSVPs by events in the published figures yields roughly 6.4 to 7.7 RSVPs per event across the three cities - that arithmetic is ours, not the authors', and it counts RSVPs, which are intentions, not attendance. Read carefully, though, it points in the same direction as the Wikipedia data: the modal community event on a large public platform is a single-digit gathering, not a crowd.

What the evidence supports

• For grassroots, interest-based community meetups, single-digit to low-double-digit attendance is normal rather than disappointing.
• The one well-documented dataset reports a mean of 8.71 and a median of 7 across 4,418 meetups.
• No credible source establishes an average for "events" as a whole, because the category spans a book club and a trade show.
• Anything presented as a cross-industry average attendance figure should be treated as unsourced until you can open the underlying study.

Mean, median, and a very long tail

The Wikipedia dataset contains a detail worth pausing on, because it recurs in every attendance distribution we have seen: the standard deviation (9.28) is larger than the mean (8.71), and the maximum (206) is more than twenty times the mean.

That is the signature of a strongly right-skewed distribution. Most events are small; a handful are enormous; the average sits above the typical case and describes almost nobody.

A quick illustration. Take three real-ish events - 8 people, 12 people and 300 people. The mean is 106.7. No event in that set resembles 106.7 people. The median, 12, describes two of the three.

The practical rule that follows is simple. When you benchmark a recurring meetup, use the median of your own past events, not the mean. One unusually large launch night will drag a mean upward for a year and make every ordinary evening look like a decline.

The four ceilings

Here is the shift that makes event sizing tractable. Stop asking what the average is. Ask what stops your event from being bigger.

Every gathering has four independent limits. They are calculated from completely different inputs, they rarely land on the same number, and the one that matters is always the lowest.

Ceiling - What sets it - How to calculate it - Binds first for

Floor area - The room, and the building code that governs it - Usable square footage ÷ occupant load factor for your layout - Fitness classes, dinners, community halls

Facilitator attention - How many people one host or instructor can actually serve - Instructors × participants per instructor for the activity - Workshops, classes, hands-on training

Interaction - How much participation the format demands of each attendee - Available minutes ÷ minutes each person needs to speak or connect - Discussions, book clubs, small networking

Demand - How many people will actually turn up - Your own RSVP count × your own historical turnout rate - New events, off-peak dates, unproven formats

Ceiling 1: floor area

This is the only ceiling with the force of law behind it, and it is the one most organisers guess at.

Building codes do not express capacity as a single number per room. They express it as an occupant load factor: a required floor area per person that changes with what people are doing in the space. In the 2021 International Building Code, Table 1004.5 sets out the allowances used across most of the United States and adopted, sometimes with amendments, by individual states.3

Use of the space - Area per person - Net or gross - People in 1,200 sq ft

Assembly, standing space - 5 sq ft - Net - 240

Assembly, concentrated (chairs only, not fixed) - 7 sq ft - Net - 171

Assembly, unconcentrated (tables and chairs) - 15 sq ft - Net - 80

Educational, classroom area - 20 sq ft - Net - 60

Exercise rooms - 50 sq ft - Gross - 24

Business areas - 150 sq ft - Gross - 8

Three things in that table repay attention.

First, the same room has several different legal capacities depending on how you lay it out. A 1,200 square foot hall holds 240 people standing, 171 in rows of chairs, and 80 at tables. If you move from a talk to a seated dinner, you have cut your own capacity by more than half without moving a wall.

Second, net and gross are not interchangeable, and the difference is not small. Net floor area excludes the parts of the space nobody occupies - circulation, fixed equipment, walls. Gross does not. The 50 sq ft exercise-room figure and the 150 sq ft business-area figure are both gross; treating them as net will produce a number that is too generous.

Third, the code number is a maximum, not a target. It answers whether everyone could exit safely, not whether anyone will enjoy the evening. A room rated for 80 at tables is usually a pleasant room for 45 to 60 people who need to move around, get to a bar, or hear each other.

Laboratory and workshop spaces follow the same logic from a different code family: fire-safety guidance widely cited in science education assigns 50 net square feet per person to shops and laboratories, which is why a 1,250 square foot school lab is generally treated as a room for roughly 25 people rather than 40.4

How to use this in ten minutes

Ask the venue for the room's square footage and its posted occupant load. Divide the square footage by the factor matching your actual layout. If your answer and the posted load differ sharply, the posted load probably assumes a different layout than the one you are planning - ask which.

Ceiling 2: facilitator attention

Once an event requires someone to teach, coach, correct or supervise, the limiting resource stops being space and becomes attention. There is a real number here, and it is smaller than most organisers expect.

The American Red Cross publishes explicit ratios for its First Aid, CPR and AED training: 6 to 10 participants per instructor, with a hard maximum of 12, plus a minimum of one manikin and one AED training device for every two participants.7 That is unusually candid documentation of something most training bodies keep inside instructor manuals. Note the shape of it: an optimal band, then a ceiling roughly 20% above the band, then equipment ratios that constrain the room independently.

The transferable principle is that hands-on instruction has a servicing rate. If every participant needs eight minutes of individual attention during a three-hour workshop, one instructor can genuinely serve around 15 people once you subtract setup, demonstration and the inevitable overruns. Two instructors serve 30. No amount of enthusiasm changes that arithmetic; it only determines how tired everyone is at the end.

Worth being honest about the limits of the evidence here: outside safety-critical training, published participant caps are rare. We found no position stand from the major fitness certification bodies specifying a maximum class size per group-exercise instructor. In practice, group fitness capacity is governed by the floor-area ceiling above, not by a professional standard - which is itself a useful finding, and a reason to work the numbers yourself rather than look for a rule that does not exist. If you are sizing a team of hosts rather than instructors, our guide to event staffing ratios works through the same calculation from the organiser's side.

Ceiling 3: interaction

This is the ceiling organisers feel but rarely calculate, and it is where the psychological research earns its place.

Participation falls as the group grows

The classic demonstration is the 1979 study by Latané, Williams and Harkins, which asked people to shout and clap as loudly as they could, alone and in groups.8 Two-person groups produced only 71% of the sum of their individual capacity. Four-person groups managed 51%. Six-person groups managed 40%.

The design's clever part rules out the obvious objection. Participants who were blindfolded and merely believed others were shouting with them also dropped off - to 82% when they believed one other person was participating, and 74% when they believed five were. Some of the loss is coordination, but a large share is motivational: as the group grows, each individual's contribution feels less consequential, and effort quietly declines.

Shouting is not conversation, and the study should not be over-read. But every organiser has watched the effect in a room: the twelve-person discussion where four people carry it, and the thirty-person discussion where six do.

Satisfaction has an optimum, not a maximum

Hackman and Vidmar's 1970 study of group size and task type found that satisfaction with group size peaked at around four to five members - and, more strikingly, that pairs were the most satisfied of all, with dissatisfaction rising roughly linearly from three members up to seven.9 Bigger was not experienced as better at any point in the range they tested.

Professional practice lands in the same place

Ask a methodologist how many people belong in a focus group and you get a number, not a shrug. The US Centers for Disease Control and Prevention defines a focus group as a group interview of approximately six to twelve people.10 That range exists because it is the band in which a moderator can keep everyone talking and still finish inside ninety minutes. Any discussion-led meetup - a book club, a peer circle, a roundtable - is running the same format under a different name, and inherits the same ceiling.

Connection grows quadratically; an evening does not

Networking events invert the logic, and it is worth being precise about why. The number of possible pairs in a group of n people is n(n−1)/2. Ten attendees give 45 possible introductions; twenty give 190; forty give 780.

The catch is that the pool grows quadratically while the individual's evening does not grow at all. Someone with two hours and five minutes per conversation will meet a dozen people whether forty or four hundred attend. Past a certain size, extra attendees stop improving one guest's experience and start diluting the odds that the right two people find each other.

There is also a structural reason larger social gatherings feel different rather than simply bigger. Analysis of the sizes of human social groups across a wide range of societies found that they cluster at discrete scales rather than spreading evenly - fitted layers of roughly 4.6, 14.3, 42.6 and 132.5 people, each about three times the last.11 Those are the familiar "5, 15, 50, 150" figures in their original, unrounded form. An event that grows from 15 to 45 has not just added thirty people; it has crossed into a scale that behaves differently, and it usually needs a different design to work.

Ceiling 4: demand

The fourth ceiling is the number of people who will actually walk through the door, and it is the only one you cannot look up.

We looked hard for a credible public benchmark on RSVP-to-attendance conversion, and specifically for the widely repeated claim that free events lose around half their RSVPs. It does not appear to have a primary source. Every instance we could trace led to a marketing blog citing another marketing blog, with no methodology, sample or year attached. There is no academic literature filling the gap either, in part because the platforms that hold the data have not generally exposed attendance alongside RSVPs.

So the honest position is this: turnout rates vary enormously by price, audience, weather, day of week, lead time and how well people know the host, and the only turnout rate worth planning against is the one from your own past events. If you have no history yet, use a deliberately conservative assumption, label it as an assumption, and replace it after three events. Our guide to reducing event no-shows covers the mechanics of moving that rate once you can measure it.

The arithmetic once you have a rate is trivial:

Required RSVPs = target attendance ÷ your historical turnout rate.

At a 70% observed turnout rate, a target of 20 people needs about 29 confirmations (20 ÷ 0.70 = 28.6). At 85%, it needs 24. The difference between those two numbers is entirely a fact about your community, not about events in general.

Sizes by event type, and which ceiling binds

With the four ceilings in place, a size table becomes genuinely useful - not because the ranges are measured averages, but because each row tells you which constraint you should be calculating.

Event type - Planning range - Ceiling that binds first - Why, and what it is anchored to

Coffee or casual social meetup - 4–12 - Interaction - A single shared conversation. Satisfaction peaks around four to five people in group-size research.

Book club or discussion group - 6–15 - Interaction - Structurally a moderated group interview; the CDC's focus-group range of six to twelve is the practical band.

Hobby or interest meetup - 6–20 - Activity design - Depends entirely on whether the activity subdivides. Board games split into tables; a shared critique does not.

Language exchange - 8–24 - Activity design - Works as multiples of small pods, not as one group. Size in units of four to six.

Hands-on workshop or class - 8–24 - Facilitator attention - Red Cross training uses 6–10 per instructor, capped at 12. Two facilitators roughly double it.

Fitness or activity class - 8–25 - Floor area - Exercise rooms carry a 50 sq ft gross occupant load factor; equipment counts constrain it further.

Networking mixer - 15–40 - Interaction - More attendees means more possible pairs, but one guest's capacity is fixed. Above roughly 40, structure is doing the work.

Professional or industry meetup - 20–60 - Floor area - Talk plus mingling. Seated at 15 sq ft net, standing at 5 - the layout sets the number.

Community gathering - 40–150 - Floor area - Shared experience with optional interaction, so it scales until the room or the programme stops it.

Conference - 150+ - Programme design - Built to scale by separating speakers from audience. Group-size limits are handled inside sessions, not at the door.

These are planning targets derived from the constraints above, not statistical averages, and we would rather label them plainly than dress them up as findings. Use the range as a starting point and the ceiling calculation as the decision.

Three worked examples

The following are hypothetical, constructed to show the arithmetic rather than to report real events.

A monthly book club in a café's back room

The room is 400 net square feet with tables and chairs: 400 ÷ 15 = 26 people by code. Facilitator ceiling: not applicable, since discussion is peer-led. Interaction ceiling: a 90-minute discussion in which everyone should speak at least three times, at roughly 90 seconds a turn, supports about 15 people with room for the conversation to breathe. Demand ceiling: 18 RSVPs at an observed 78% turnout gives 14.

Target: 14. The room could legally hold 26 and would ruin the discussion at that size.

A pottery workshop with two instructors

Room: 1,200 net square feet at tables, giving 80. Facilitator ceiling: two instructors at a hands-on ratio of 12 each, giving 24. Wheel count: 16 - an equipment ceiling that the four-ceiling model treats as part of floor area. Interaction ceiling: every participant needs individual feedback twice in three hours, which one pair of instructors can deliver for about 15 people. Demand: 22 RSVPs at 70% gives 15.

Target: 15, with 16 wheels giving one spare. This is the case shown in the chart above, and the room's capacity of 80 never once entered the decision.

A quarterly professional networking mixer

Room: 2,000 net square feet, standing with a bar, giving 400 by code - a number that is legally correct and practically irrelevant. Facilitator: two hosts running introductions. Interaction: the organiser's goal is that each attendee makes at least three useful connections, which structured introductions can deliver reliably up to about 40 people and only with breakout structure beyond that. Demand: this is the fourth run, and the past three produced turnout of 64%, 71% and 69%, so 68% is a defensible planning rate. To reach 40 people: 40 ÷ 0.68 ≈ 59 RSVPs.

Target: 40 attendees, 59 RSVPs, with an explicit decision to redesign the format rather than simply admit more people if demand exceeds that. If it regularly does, a fair waitlist for an oversubscribed event is a better answer than quietly raising the cap.

Format beats headcount

There is a genuinely useful finding buried in education research, and it applies well beyond classrooms.

Tennessee's Project STAR is one of the largest randomised experiments ever run in education: roughly 6,500 pupils across around 79 schools were randomly assigned, from kindergarten through third grade, to small classes of 13 to 17 pupils or regular classes of 22 to 25.12 Cutting class size by about a third produced measurable gains - effect sizes of roughly 0.23 in reading and 0.27 in mathematics on standardised tests, with smaller carryover effects persisting into later grades. Real, but modest for a 35% reduction in group size.

Set that next to a widely cited synthesis of education research in which "reducing class size" carries an effect size of about 0.21 while "small group learning" - restructuring how people work within a class - carries about 0.47.13 That synthesis has been criticised in the peer-reviewed literature and should be treated as contested rather than settled. Taken together with STAR, though, the two point the same way: changing how a group is organised does more than changing how many people are in it.

Which is the practical answer to the most common growth problem an organiser faces. Suppose a 15-person networking meetup becomes popular and 50 people want in. Admitting all 50 turns a working event into a crowded one. Capping it at 15 turns away 35 people who wanted to come. The third option is to keep everyone and change the shape: a ten-minute welcome, five groups of ten with a discussion prompt, one rotation, then open mingling. The event is now 50 people running five parallel fifteen-person formats, which is a structure the research above suggests will work - not because it is smaller, but because it is organised differently.

Growth almost always requires more structure. If you are approaching that transition, our walkthrough on co-organising a meetup group covers splitting the hosting load that structure demands, and the event run-of-show timeline is where that structure gets written down.

Building your own benchmark

After three or four events, your own history is more informative than any published figure - and it is the only benchmark that accounts for your city, your audience and your Tuesday-versus-Thursday problem.

Track four columns and nothing else. The table below is illustrative, showing the shape of a log rather than data from a real group.

Event - RSVPs - Attended - Turnout rate - Returning attendees

1 - Tuesday, new venue - 20 - 13 - 65% - -

2 - Thursday - 24 - 17 - 71% - 9

3 - Thursday - 27 - 20 - 74% - 13

4 - Tuesday, wet evening - 30 - 19 - 63% - 12

5 - Thursday - 32 - 23 - 72% - 16

Five events in, this organiser knows more than any industry average could tell them: a median turnout around 71%, a Thursday effect worth roughly eight percentage points, and a returning-attendee count that is climbing faster than the RSVP count. That last column is the one that actually predicts whether the group survives.

The fifth column is deliberately not "satisfaction". Repeat attendance is the strongest signal a community organiser has, because it is behaviour rather than opinion, and it is free to collect. If people come back, the size is right, whatever the number says.

Four numbers people confuse

Occupant load The maximum number of people a space may legally hold under the applicable building code, calculated from floor area and the occupant load factor for the intended use. A safety limit, not a plan. Planned capacity The number of attendees you deliberately design the event around - the lowest of your four ceilings. Almost always well below the occupant load. RSVP count The number of people who have indicated they intend to come. An intention, collected at a moment in time, which decays as the date approaches. Actual attendance The number of people physically present. The only one of the four that tells you whether the event worked, and the only one most organisers forget to record.

Common sizing mistakes

• Treating the venue's capacity as the target. The occupant load answers a fire-safety question. It has no opinion about whether your discussion will work.
• Reading RSVPs as attendance. They are different measurements with a variable gap between them, and the gap is a property of your community, not of events generally.
• Benchmarking against a mean. Attendance distributions are right-skewed, so the mean sits above the typical event and makes normal turnout look like decline.
• Comparing across categories. A conference average tells a book club organiser nothing. Match the population before you borrow the number.
• Scaling the guest list instead of the format. Doubling attendance without changing the structure usually halves participation.
• Assuming the ceiling that binds is fixed. It moves. Add a second facilitator and the facilitator ceiling stops binding; the next one takes over.
• Not writing down the number afterwards. Three recorded events beat any published benchmark, and the recording costs thirty seconds.

When a meetup should deliberately stay small

Growth is not the only success condition, and treating it as one has quietly killed a lot of good groups.

Keep an event small when its value depends on depth: peer mentoring, mastermind groups, hands-on instruction, roundtables, support-style conversations, anything where the point is that everyone is known in the room. In these formats, adding people does not add value - it converts participants into an audience. A discussion that grows from 10 to 30 has tripled its attendance and cut the share of people who speak by roughly two thirds.

Grow the event when attendees do not need continuous direct contact with everyone else: talks, demonstrations, celebrations, performances, large-scale networking with structure. Here scale genuinely adds something - more variety, more energy, more chance that the right people are in the room.

A first-time organiser is usually better served by aiming for 8 to 20 actual attendees, because the goal of a first event is not to prove demand. It is to find out whether the format works: whether people talk, whether they stay to the end, whether they come back. Those questions are answered more clearly by fifteen people than by sixty. If you are at that stage, our guide to starting a meetup group covers the first three events in detail.

Frequently asked questions

What is the average meetup size?

There is no authoritative cross-industry figure. The best-documented dataset - 4,418 offline meetups of the German-language Wikipedia community recorded between 2001 and 2020 - reports a mean of 8.71 attendees and a median of 7. That is a defensible anchor for grassroots interest-group meetups and should not be applied to conferences, training events or commercial gatherings.

Is 10 people too small for a meetup?

No. In the largest well-documented meetup dataset available, the median gathering had seven attendees, so ten is above typical for that population. Ten is a strong size for discussion, hobby groups and community building, and group-size research consistently finds satisfaction peaking in the small single digits rather than rising with headcount.

Is 50 people a large meetup?

For a casual community meetup, yes. At around 50 people a single shared conversation is no longer possible and the event needs structure - breakout groups, a short programme, facilitated introductions - to avoid becoming a collection of disconnected clusters. Fifty is comfortably small for a conference and clearly large for a book club, which is why the format matters more than the number.

What is the ideal size for a networking event?

Roughly 15 to 40 attendees for an unstructured mixer. Below 15, the pool of possible introductions is thin; above 40, the number of possible pairs grows far faster than any individual's capacity to use them, so the event needs deliberate structure - speed networking, topic tables, rotation - to keep producing connections.

What is the ideal size for a workshop?

For hands-on instruction, 8 to 24 participants depending on how many facilitators are present. The American Red Cross specifies 6 to 10 participants per instructor for its first aid and CPR training, with a maximum of 12, which is a reasonable model for any workshop requiring individual attention. Presentation-led "workshops" have no such limit because they are lectures.

How many people should I invite if I want 20 to attend?

Divide your target by your own observed turnout rate. At 70% turnout, 20 ÷ 0.70 means around 29 confirmed RSVPs. Use your own historical rate rather than a published one - no credible primary source establishes a general RSVP-to-attendance rate, and the figures circulating online cannot be traced to any measurement.

Should I benchmark against average or median attendance?

Median. Attendance distributions are strongly right-skewed - in the Wikipedia meetup dataset the standard deviation of 9.28 exceeds the mean of 8.71, and the largest event drew 206 people. One unusually large event pulls a mean upward for months and makes ordinary turnout look like a decline.

How do I know what my venue can actually hold?

Divide the room's usable square footage by the occupant load factor for your layout. Under the 2021 International Building Code, that is 15 net square feet per person for tables and chairs, 7 for chairs alone and 5 for standing. Then treat the result as a legal maximum and plan for meaningfully fewer - a room rated for 80 seated is usually comfortable for 45 to 60 people who need to move around.

Where Ontaym fits

Most of this article is arithmetic you can do on paper. The part that is genuinely tedious is the fourth ceiling: keeping a current count of who has said yes, watching it change in the week before the event, and recording what actually happened afterwards so the next event has a real turnout rate to plan against.

Ontaym exists for that part - one event page, one link, responses that stay current in one place rather than scattered across a group chat. It will not tell you how big your event should be. It will tell you, accurately, how big your last one was, which is the number this whole article has been arguing you should be using.

Stop guessing your turnout rate. Start measuring it.

The number worth having

The question "what is the average meetup size?" has no honest answer, and the answers on offer are mostly citations of citations. The question "how many people should come to my next event?" has a precise one, and you can calculate it this afternoon.

Work out your four ceilings. Take the lowest. Divide by your own turnout rate to get the RSVPs you need. Run the event, write down what actually happened, and after three or four rounds you will have a benchmark that describes your community rather than someone else's dataset.

An event is the right size when the format it is built on still works at that size. Eight people in a room where everyone contributes is a successful meetup. Eighty in a room designed for a shared conversation is not, however impressive the headcount looks in the recap.

Sources Schwitter, N. (2023). Bridging the offline and online: 20 years of offline meeting data of the German-language Wikipedia. Journal of Computational Social Science, 6(2). link.springer.com - 4,418 meetups; mean 8.71, median 7, SD 9.28, max 206. Pramanik, S., Gundapuneni, M., Pathak, S. & Mitra, B. (2016). Predicting Group Success in Meetup. Proceedings of the International AAAI Conference on Web and Social Media (ICWSM). ojs.aaai.org - group, event and RSVP volumes for three US cities, 2015. International Code Council (2021). International Building Code, Section 1004.5 and Table 1004.5, Maximum Floor Area Allowance Per Occupant. codes.iccsafe.org. Factor values as adopted in state law at Washington Administrative Code 51-50-1004. National Science Teaching Association. Science/STEM laboratory occupancy load. nsta.org - 50 net square feet per person for shops and laboratories. Events Industry Council (2023). Global Economic Significance of Business Events, 2019 baseline year. eventscouncil.org - 1.6 billion participants; $1.15 trillion direct spending; no attendance breakdown by event category. US Bureau of Labor Statistics. Occupational Outlook Handbook: Meeting, Convention, and Event Planners. bls.gov - occupational data only; no event attendance statistics. American Red Cross (2017). First Aid/CPR/AED Program Fact Sheet. redcross.org (PDF) - 6–10 participants per instructor, maximum 12; one manikin and one AED trainer per two participants. Latané, B., Williams, K. & Harkins, S. (1979). Many hands make light the work: The causes and consequences of social loafing. Journal of Personality and Social Psychology, 37(6), 822–832. PDF - 71% at two people, 51% at four, 40% at six. Hackman, J. R. & Vidmar, N. (1970). Effects of size and task type on group performance and member reactions. Sociometry, 33(1), 37–54. scholar.harvard.edu - satisfaction optimum around four to five members. Centers for Disease Control and Prevention. Evaluation Briefs No. 13: Data Collection Methods for Program Evaluation - Focus Groups. cdc.gov (PDF) - focus groups defined as approximately six to twelve people. Zhou, W.-X., Sornette, D., Hill, R. A. & Dunbar, R. I. M. (2005). Discrete hierarchical organization of social group sizes. Proceedings of the Royal Society B, 272(1561), 439–444. royalsocietypublishing.org - fitted layers of 4.6, 14.3, 42.6 and 132.5, scaling ratio close to three. Finn, J. D. & Achilles, C. M. The Tennessee Study of Class Size in the Early School Grades. PDF - small classes of 13–17 versus regular classes of 22–25; effect sizes of 0.23 in reading and 0.27 in mathematics. Hattie, J. Visible Learning ranking of influences and effect sizes (2017 edition). visible-learning.org - reducing class size d = 0.21; small group learning d = 0.47. The synthesis method has been criticised in the peer-reviewed literature and is presented here as contested.

Ontaym Editorial Team

Ontaym builds tools for organising real-world gatherings, so the team spends its days on the coordination problems this article describes. Articles are researched against primary sources, reviewed before publication, and revised when the underlying facts change rather than on a schedule.