Every mod on the Steam Workshop has exactly one thing that is unique to it and no other item on the platform: nothing. Not the name.
There is no uniqueness check when a creator publishes. Two mods in the same game can carry the same title, spelled identically, and the Workshop will accept both. So they do. In the live subscribed catalogue, 245,880 names are claimed by two or more mods inside the same game — and those names carry 976,915 mods between them, 16.53% of everything on the Workshop.
That is a structural fact nobody has measured, and once you measure it you get a number that looks like a finding and is partly an artifact. Both halves of that sentence matter, so let us take them in order.
The pooled gap
Split the 5,911,464-mod subscribed catalogue by whether the mod's name is unique within its game:
| Name status | Mods | Share of catalog | Reached 1,000 subs | Share of all subscriptions |
|---|---|---|---|---|
| Unique within its game | 4,934,549 | 83.47% | 15.96% | 90.46% |
| Collision, different creators | 846,219 | 14.31% | 6.50% | 8.72% |
| Collision, same creator | 130,696 | 2.21% | 7.29% | 0.83% |
A mod with a unique name clears a thousand subscribers 2.46x as often as one sharing its name with a rival. The gradient is monotone in how crowded the name is, and it does not stop:
| Mods sharing the name | Mods | Reached 1,000 subs | Share of all subs |
|---|---|---|---|
| 1 (unique) | 4,934,549 | 15.96% | 90.46% |
| 2 | 311,286 | 9.37% | 4.57% |
| 3–4 | 179,718 | 7.45% | 2.17% |
| 5–9 | 147,367 | 6.16% | 1.39% |
| 10–49 | 201,794 | 4.75% | 1.10% |
| 50+ | 136,750 | 2.42% | 0.32% |
A mod is published into a name used by fifty other mods in its own game and it is 6.6x less likely to reach a thousand subscribers than a mod whose name belongs to it alone. Every band in the ladder is one way down. There is no sweet spot.
That is the finding as the pooled data gives it, and the pooled data is the wrong way to read it.
Why the pooled number is too big
The collision-heavy games are not a random sample of the Workshop. They are a specific kind of game.
Human Fall Flat sends 51.7% of its catalogue into a shared name. A Dance of Fire and Ice 53.1%. SpeedRunners 41.1%, Trailmakers 40.3%, WorldBox 36.5%, Besiege 33.9%, Castle Crashers 32.3%. And every one of those is a game where the Workshop page is not chiefly a mod store — it is a file-sharing slot. Trailmakers blueprints, Besiege levels, Plague Inc scenarios, Terraria builders' workshops. Names like blueprint #1, example levels, untitled scenario and transformation slot are not two creators colliding by accident. They are the default label an in-game editor stamps on user content, and thousands of people leave it.
Those games also sit at the bottom of the reach distribution, because their content is not competing for subscribers the way a mod does. So they contribute a huge, low-performing collision block to the pooled average, and they drag it down.
The clean test is to count games instead of mods. Across the 125 games with at least 500 mods in each group:
- 86 games favour the unique name. 39 favour the collided name.
- Mean of the per-game reach rates: 12.86% unique vs 9.70% collided — 1.33x.
- Pooled over the same 125 games: 14.69% vs 6.42% — 2.29x, the number the table above would have given you.
Same data, same games, and the penalty shrinks from 2.3x to 1.33x just by giving every game one vote. That is textbook aggregation bias, and it is the first thing to say in this article. Drop the two most collision-saturated ecosystems — Trailmakers and Besiege — entirely and the pooled gap barely moves (16.85% vs 7.96%), which tells you the composition problem is broader than two games: it is distributed across a whole class of in-game-content ecosystems, not concentrated in a couple of outliers.
So the headline is not 2.3x. The honest headline is a range, and the low end of it is thin.
The control that holds
Which is why the paired test matters. This series has now killed three separate axes — publishing hour, tag count, and the audience-portability headline — because a pooled effect evaporated once the same creator's own mods were compared to each other. If name collisions are just a marker for which ecosystem you are in, the paired test should erase them too.
It does not. Across 236,968 creator-and-game pairs where the same creator published both a uniquely-named mod and a collided one in the same game:
| Mods | Reached 1,000 subs | |
|---|---|---|
| Uniquely named | 1,920,507 | 15.58% |
| Collided | 595,868 | 8.36% |
That is 1.86x for the same people, in the same games, shipping into the same audience. Of the pairs where the two groups differed, 23,457 creators did better on their unique names and 8,232 did better on their collided ones (205,279 tied). The name-collision gap survives the control that killed the timing axes.
And so does the listing-quality control. Restrict to mods that have both a preview image and a description — the standard presentation bar this series uses to separate a formatting signal from a listing-quality proxy:
| Mods | Reached 1,000 subs | |
|---|---|---|
| Complete listing, unique name | 2,970,196 | 24.21% |
| Complete listing, collided name | 382,474 | 13.70% |
1.77x, essentially unchanged. A well-presented mod with a contested name still reaches a thousand subscribers 1.77x less often than an equally well-presented mod whose name is its own.
The gap also holds in every publish year we can bucket, widening as the Workshop gets more crowded:
| Published | Unique | Collided |
|---|---|---|
| 2015 | 21.88% | 6.67% |
| 2020 | 18.89% | 9.61% |
| 2024 | 13.75% | 5.09% |
| 2026 | 9.09% | 3.23% |
So: a real effect, of contested but real size. Somewhere between 1.33x (game-weighted) and 1.86x (within-creator paired). Anyone quoting you the pooled 2.46x is quoting composition.
What a shared name actually does
Here the mechanism is legible, and it is not mysterious. Compare the medians of the two groups:
| Median views | Median favourites | Median subscribers | |
|---|---|---|---|
| Unique name | 63 | 1 | 39 |
| Collided name | 9 | 0 | 10 |
A collided mod is seen seven times less. The gap is at the top of the funnel, not at the bottom — it looks like the discovery problem this series documented in the discovery lottery and the one-subscriber floor, not a conversion problem. Nobody is arriving at a contested mod and deciding against it. Nobody is arriving.
Which fits the platform's search behaviour: when seven mods are called the same thing, a query for that name is a list of undifferentiated rows, and the click goes to whichever one the ranking already favours. The rest are, in practice, unreachable by the one search term they chose.
Inside a collision, position is everything
The striking part is what happens within a shared name — and this is the clearest number in the dataset. Rank the mods inside each collided name by publication date:
| Position in the name | Mods | Reached 1,000 subs | Subscriptions held |
|---|---|---|---|
| First published | 245,821 | 12.28% | 761,310,866 |
| Second | 245,821 | 6.79% | 218,134,056 |
| Third–fifth | 178,746 | 5.06% | 119,950,624 |
| Sixth or later | 306,224 | 2.81% | 88,154,767 |
The first mod to claim a name reaches a thousand subscribers 4.4x as often as the seventh-plus. And the concentration is stark: the top mod in a collided name holds a median 66.7% of that name's entire subscription count (p25 52.6%, p75 85.5%).
Being early is not a small advantage here. It is most of the outcome. A name is a finite resource on the Steam Workshop, and it is allocated by timestamp.
84% of shared names are graveyards
Put the clusters rather than the mods in view and the picture gets bleaker:
| What the whole name achieved | Shared names | Share | Mods in them |
|---|---|---|---|
| No mod in the name ever reached 1,000 | 207,497 | 84.39% | 759,283 |
| Exactly one mod reached 1,000 | 24,600 | 10.00% | 111,700 |
| Two or more mods reached 1,000 | 13,783 | 5.61% | 105,932 |
In 84.4% of collided names, nobody in the group ever gets to a thousand subscribers. A shared name is overwhelmingly a dead end for every party — the pool is not being divided, it is failing to fill.
And when a name does produce a hit, it usually produces exactly one. Multiple winners in one name is a 5.6% event.
That is also where the "same creator" cluster pattern shows up. When one creator owns every mod under a name — a numbered series like blueprint #1, #2, #3 — the reach rate is 7.29%. When two to five creators contest it, 9.23%. When six or more creators fight over one name, 3.66%. The most contested names are the worst by a wide margin.
The exception worth naming
Not every game punishes a collision, and the ones that don't are informative. Wallpaper Engine — the largest ecosystem on the Workshop — has 93.40% of its uniquely-named mods reach a thousand subscribers and 96.51% of its collided ones. There is no penalty at all, because the reach rate was already near the ceiling: when almost everything succeeds, sharing a name cannot cost you much.
The clearest inversions are where the Workshop page is a discovery surface rather than a distribution channel. Counter-Strike 2 runs 23.86% unique vs 42.99% collided; Left 4 Dead 2 31.39% vs 37.60%; Tabletop Simulator 14.25% vs 22.10%; RimWorld 29.50% vs 31.81%. In these ecosystems a common name appears to help — most plausibly because the common names are the established, canonical ones the community already searches for, and a newcomer adopting them is riding a known term rather than inventing one.
The losses are concentrated in the sandbox and builder games, and they are large: Cities: Skylines 41.60% vs 11.17% (3.72x), Paralives 13.25% vs 2.81% (4.72x), Avorion 10.66% vs 3.23% (3.30x), Besiege 4.01% vs 1.31%, ARK: Survival Evolved 28.37% vs 11.09%, Ravenfield 42.66% vs 19.25%, Stormworks: Build and Rescue 10.47% vs 4.95%, Human Fall Flat 9.24% vs 3.73%, Space Engineers 7.58% vs 4.26%.
The big clusters are all in exactly that class of game, and they are all placeholder labels:
| Game | Shared name | Mods | Subscriptions | Reached 1,000 |
|---|---|---|---|---|
| Kerbal Space Program | untitled space craft |
3,969 | 55,377 | 4 |
| Besiege | example levels |
2,668 | 4,232 | 0 |
| Plague Inc: Evolved | untitled scenario |
2,082 | 2,527 | 0 |
| Trailmakers | transformation slot |
1,668 | 26,630 | 0 |
| Terraria | builder's workshop |
1,485 | 69,395 | 7 |
| Terraria | all items |
506 | 321,400 | 10 |
| Besiege | plane |
648 | 76,399 | 7 |
untitled space craft, 3,969 times, and four of them ever reached a thousand subscribers. That is not 3,969 mods competing. That is a default filename.
What it means
If you are a creator, the name is the cheapest lever you have and you are probably wasting it. A shared name cuts your odds of reaching a thousand subscribers to somewhere between two-thirds and half of what a unique name gives you, and if you are not the first to publish under it, the person who was keeps a median two-thirds of everything the name ever earns. My Mod is not a name. Neither is anything the game's own editor prefilled.
If you are publishing into a builder game, know that the collision rate is not about you. Human Fall Flat, Trailmakers, Besiege, A Dance of Fire and Ice and SpeedRunners all sit above 34% collision — because their editors write the filename for you. In those ecosystems the median blueprint #1 is one of thousands, and it is competing against a default.
If you are reading a Workshop statistic, the pooled average is lying to you about this axis — including the first table in this article. Giving every game one vote instead of every mod cuts a 2.29x headline to 1.33x. This series has now found the same shape four times: a convincing pooled gap that is partly composition, and a real residue underneath that survives a paired control. Both parts are the finding. Report the residue; do not quote the pooled number.
Related reading: the discovery lottery on the 94% of mods that get zero views in a week, the one-subscriber floor on the 654,035 mods parked at a single install, the short-title problem on the title-length axis this one sits beside, the tag cliff on the 1.8 million mods carrying no labels at all, and the blank-page problem on listings that never told anyone what they were.
Method: 5,911,464 Steam Workshop items tracked by ModSignal as of 29 September 2026 across 392 games — every item that is is_active and carries at least one subscription. Names are compared as lower(trim(title)) within the same game, so the same title in two different games is not a collision; cross-game reuse is not measured here and would raise every figure. Rows whose normalised name matches a default-editor label (untitled, unnamed, new, test, 1, dream, blank, default, example, template, sample, aaa, asd, asdf, xxx, qwerty, no name, noname, nothing, placeholder, map, level, scenario, playlist, blueprint, new map, new level, new project, new scenario, test map, test 1, each optionally followed by digits) are excluded from the pooled, ladder and paired figures above — 63,271 items — because they are an editor's default filename rather than a shared name; including them moves the collided share from 16.53% to 17.26% and the pooled reach rate from 6.50% to 6.28%, with no change to any conclusion. The two largest clusters discussed by name (untitled space craft, example levels, untitled scenario, transformation slot, builder's workshop, all items, plane) are quoted with their full cluster sizes so their placeholder character is visible rather than hidden by the exclusion. "Reached 1,000 subscribers" means subscriptions_current >= 1000 at a single consistent read; it is a point-in-time threshold, not a tracked trajectory, and a mod that cleared and later lost subscribers is not counted. The paired test compares creators who published both a unique and a collided name in the same game and reports the aggregate rate on each side, with ties (a creator with zero hits on both sides, or equal hit counts) excluded from the win/loss tally — they are the large majority, which is why the decided pairs number 31,689 of 236,968. The game-weighted figures are unweighted means of the individual per-game reach rates across the 125 games with at least 500 mods in each group, so a 700-mod game counts the same as a 180,000-mod one; that is deliberate, and it is the whole point of showing it beside the pooled number. The within-name position ranking orders mods by time_created inside each game-and-name group and is computed over the full colliding set including default labels, since position within a cluster is meaningful there too. Subscriptions are the last reported values and are never decremented by Steam, so a delisted mod, if any, would still be counted. Counters move continuously; a re-run will differ by a handful of items.