# Co-op game analytics: party size, invites and K-factor

> Track party size, invites sent and accepted, and how many new players each player brings, to measure the friend loop that sells co-op and friendslop games.

By Yos Riady. Updated 2026-11-18. Published 2026-11-18. Canonical URL: https://manalens.dev/blog/coop-multiplayer-game-analytics

Co-op games grow through friends, so measure party size, invites sent, invites accepted and new players per player (K-factor). Friendslop is growing: across the top games, peak concurrent players rose from 197,000 to 350,000 (How To Market A Game, July 2026). Track the friend loop in your own game.

## Key points

- K-factor is invites per new player times the share that bring a new player, so 3 invites at 20% gives K of 0.6.
- With a K-factor of 0.6, 1,000 starting players grow to about 2,500 over many waves, and K of 0.8 gives about 5,000.
- Record Remote Play guests as a property of the host's match_start, because guests never run your build.

## Why do co-op games need different metrics?

Because one player brings others, and a group plays together or not at all. A solo game counts players. A co-op game also counts parties: how big they are, how they form and whether the group comes back.

The market data shows the pattern. Chris Zukowski tracked 16 co-op party games. Lethal Company brought the combined peak to 197,000 players, R.E.P.O. to 281,000 and MECCHA CHAMELEON to 350,000. Top games took about 50% of the genre's players, then faded to about 10% within 3 to 9 months. Mage Arena peaked at 11.75% of the genre and still sold over 624,000 copies in a week, per GameDiscoverCo estimates quoted in the post.

Sources: How To Market A Game (30 July 2026).

## Which numbers should a co-op game track?

Start with five: party size, solo share, invites sent per player, invite acceptance rate and K-factor. Then add group retention: did the same group play again.

*Co-op metrics and how to compute them.*

| Metric | Formula | What it tells you |
| --- | --- | --- |
| Party size | Average players per match_start | How groups play |
| Solo share | Matches with 1 player / all matches | Players who cannot find friends |
| Invites per player | invite_sent / players | How much players share |
| Acceptance rate | invite_accepted / invite_sent | How well invites land |
| K-factor | Invites per player x acceptance rate (new players only) | New players each player brings |
| Group return rate | Groups that play again within 7 days / groups | Whether the group sticks |

Sources: David Skok, Lessons learned, viral marketing; Manalens track event documentation.

## How do I calculate K-factor for a game?

Multiply the invites each new player sends by the share of invites that turn into new players. David Skok gives the formula as K = i x conv%. If each new player sends 3 invites and 20% of them bring a new player, K is 0.6 (our arithmetic). Each 10 players bring 6 more.

Count only invites that bring a new player. An invite to a friend who already owns the game helps retention, not growth. So send a flag on invite_accepted that says whether this is the friend's first session.

Cycle time matters too. Skok defines viral cycle time as the time one pass through the loop takes. A friend who joins the same evening grows the game faster than one who joins next month.

Sources: David Skok, Lessons learned, viral marketing.

## Which events should I send?

Send invite_sent, invite_accepted, party_formed, match_start and match_end. Put the party size, the host flag and a party ID on each. Keep the party ID random, not a Steam ID, so you do not send personal data.

On Steam, the invite usually goes through the overlay. ISteamFriends has ActivateGameOverlayInviteDialog for a lobby and InviteUserToGame with a connect string. When the friend accepts, the game gets a callback with the connect string, which is your hook for invite_accepted.

``` Godot
Manalens.track("party_formed", {"party_id": party_id, "party_size": 3, "is_host": true, "invites_sent": 2, "first_session": false})
```

Sources: Steamworks ISteamFriends documentation; Manalens game events documentation.

## How do I count players who join through Remote Play?

As guests of the host, not as owners. With Remote Play Together, only the host needs to own and install the game. Friends join the host's game and stream it. Each connected device has its own session, but the guests never run your build.

So your SDK sees one install. Record the number of Remote Play guests as a property of the host's match_start. A party of 4 with 3 guests is one sale today and up to 3 potential sales later.

Remote Play also hides the friend loop from your invite events. A guest who joins this way may never send invite_accepted from their own install. Count Remote Play parties as their own group, and compare how often a guest later shows up as a new install with a first session soon after the match. That share is the conversion of your free guests.

Sources: Steamworks Steam Remote Play documentation.

## What does a K-factor add over time?

Each wave of new players brings a smaller wave, if K is below 1. Start with 1,000 players and K of 0.6. The first wave brings 600, the next 360, then 216. Over many waves the total gets close to 2,500 players, or 1,000 / (1 minus 0.6). This is our arithmetic from the Skok formula, with no churn.

So a small change in K matters a lot. At K of 0.3 the same 1,000 players end near 1,429. At K of 0.8 they end near 5,000. Raising acceptance from 10% to 20% can double the friends each player brings, so fix the join flow before you buy more traffic.

*Total players from 1,000 starting players by K-factor. Our arithmetic, no churn.*

| K-factor | First wave | Total after many waves |
| --- | --- | --- |
| 0.3 | 300 | About 1,429 |
| 0.6 | 600 | About 2,500 |
| 0.8 | 800 | About 5,000 |

Sources: David Skok, Lessons learned, viral marketing.

## How do I find where the friend loop breaks?

Build a funnel: game_open, invite_sent, invite_accepted, match_end. Compare the share at each step. A low invite rate means the invite is hard to find or the game gives no reason to share. A low acceptance rate often means friends do not own the game or cannot join fast.

Then split by party size and solo share. If many players start solo matches, they either have no friends playing or cannot find a lobby. Track a "searching for party" event with its wait time to tell these apart.

Compare group return rate by party size too. In many co-op games a group of 3 or 4 is the core of the game. If full parties come back and duos do not, the content may be tuned for one group size only.

Sources: Manalens funnels documentation.

## Where can I see invites, parties and retention together?

Manalens is the analytics and growth engine built for games. Send invite_sent, invite_accepted and party_formed from the Unity or Godot SDK, then build a funnel from game_open to match_end and break it down by the party_size property. Retention and acquisition source sit in the same project.

- [D1, D7, D30 retention calculator](https://manalens.dev/tools/retention-calculator)

Sources: Manalens product documentation.

## Frequently asked questions

### What is a good K-factor for a co-op game?

A K-factor above 1 means each player brings more than one new player, so the game grows on its own. Few games keep that for long. Even 0.3 to 0.5 matters, because it adds that share to every player you buy or earn. Track it by week and by acquisition source.

### Should I track Steam IDs to link friends in a party?

No. Use a random party ID and the anonymous install ID from the SDK. That is enough to count parties, sizes and returns. Steam IDs are personal data, and you do not need them to see whether invites work or whether groups come back.

### How do I measure a friend pass or free guest access?

Treat each guest as a potential buyer. Send a guest flag on the guest's events, then count how many guests later start a session as owners. With Remote Play Together, guests do not run your build, so record their number on the host's match event instead.

### Which acquisition source brings the most groups?

Break down the party funnel by first touch source. A creator video often brings friend groups that buy together, while an ad may bring solo players. Compare party size, invite rate and K-factor by source, then put more budget into the sources whose players bring friends.

### Is the friendslop genre too crowded to enter?

How To Market A Game argues it is not saturated: the combined peak rose with each hit, to 350,000 players. But top games take about half the genre and fade within months. A new game must win the first weeks with streamers and friend groups, so measure invites from day one.

## Sources

- [How To Market A Game: Is Friendslop saturated? (Chris Zukowski, 30 July 2026)](https://howtomarketagame.com/2026/07/30/is-friendslop-saturated/)
- [David Skok: Lessons learned, viral marketing (For Entrepreneurs, read 18 November 2026)](https://www.forentrepreneurs.com/lessons-learnt-viral-marketing/)
- [Steamworks documentation: ISteamFriends (Valve, read 18 November 2026)](https://partner.steamgames.com/doc/api/ISteamFriends)
- [Steamworks documentation: Steam Remote Play (Valve, read 18 November 2026)](https://partner.steamgames.com/doc/features/remoteplay)

## Related guides

- [Game marketing measurement guide](https://manalens.dev/blog/game-marketing-measurement-guide)
- [Player segmentation for small studios](https://manalens.dev/blog/player-segmentation-small-studios)
- [D1, D7 and D30 retention benchmarks for PC and indie games](https://manalens.dev/blog/pc-indie-game-retention-benchmarks)
- [Difficulty tuning with data](https://manalens.dev/blog/game-difficulty-tuning-analytics)
- [D1, D7, D30 retention calculator](https://manalens.dev/tools/retention-calculator)
- [CCU to DAU and MAU estimator](https://manalens.dev/tools/ccu-to-dau-mau-estimator)
- [Game marketing guides](https://manalens.dev/blog/topic/game-marketing)
- [Game metrics every indie developer should track](https://manalens.dev/blog/game-metrics-indie-developers)
- [Game analytics dashboards: which charts matter](https://manalens.dev/blog/game-analytics-dashboard-charts)
- [UGC and mods: effect on game sales and retention](https://manalens.dev/blog/ugc-mods-sales-retention)
- [Estimate Steam sales from CCU: does demo CCU predict success?](https://manalens.dev/blog/estimate-steam-sales-from-ccu)
