Every dataset has a floor: the value that the most items pile up on. On the Steam Workshop it is not zero. Zero is where untracked items sit, and that is a different story. The floor of the live catalogue — the 6,003,574 mods that carry at least one subscription today — is one.
654,035 live Steam Workshop mods have exactly one subscriber. Not "few". Exactly one, to the digit, as reported by Steam. That is 10.9% of the live catalogue in a single bucket, and it is the largest single-value population in the entire dataset. Another 401,005 sit at exactly two. Between them, more than a sixth of everything moddable on Steam is parked on the first two rungs.
What that population holds, collectively, is almost nothing: 654,035 subscriptions out of 12,458,624,755 — 0.005%. The 16,184 mods above 100,000 subscribers hold 42.7%.
That much is arithmetic. The interesting question is what the floor means. Is it a verdict on 654,035 bad mods? Or is it something the ecosystem does to them? We spent 41 days of snapshots and a paired creator control trying to tell those two apart, and the answer is not the one the pooled table suggests.
The shape of the live catalogue
| Subscribers | Mods | % of live catalogue | Subscriptions held | % of all subscriptions |
|---|---|---|---|---|
| 1 | 654,035 | 10.89% | 654,035 | 0.005% |
| 2–5 | 1,033,392 | 17.21% | 3,223,197 | 0.026% |
| 6–20 | 1,050,229 | 17.49% | 11,929,103 | 0.096% |
| 21–100 | 1,159,249 | 19.31% | 57,301,286 | 0.460% |
| 101–999 | 1,251,757 | 20.85% | 448,519,530 | 3.600% |
| 1,000–10,000 | 672,728 | 11.21% | 2,202,750,240 | 17.681% |
| 10,000–100,000 | 166,000 | 2.77% | 4,408,496,117 | 35.385% |
| 100,000+ | 16,184 | 0.27% | 5,325,751,263 | 42.748% |
Two things to read off this table. The first is that the distribution is not a long tail that tapers smoothly — it is a wall at the bottom. More mods sit at exactly one subscriber than in any band above it. Nearly half the live catalogue (46.9%) has never got past twenty subscriptions.
The second is that the audiences and the inventory are almost disjoint populations. Delete the 654,035 one-subscriber mods and you remove a tenth of the platform's inventory and five thousandths of one percent of its usage. Delete the 16,184 mods above 100k and you remove a quarter of a percent of the inventory and 42.7% of everything.
Does anything ever leave the floor?
A snapshot tells you where a mod is. It does not tell you whether it is stuck. For that you need history, so we used ModSignal's snapshot ledger: 27,511,046 subscription readings across 41 days, 15 August to 27 September 2026.
Take every mod whose first reading in that window was exactly one subscriber, and ask what its highest reading was afterwards:
| Started at 1 subscriber | Mods | Share |
|---|---|---|
| Highest reading was still 1 | 613,656 | 97.88% |
| Ever gained a second subscriber | 13,275 | 2.12% |
| Ever reached 10 subscribers | 1,585 | 0.25% |
| Ever reached 1,000 subscribers | 12 | 0.002% |
613,656 mods spent six weeks at one subscriber and never moved. The median member of the cohort finished the window at exactly the value it started on. Twelve of 626,931 climbed to four figures; the single best performer in the entire group peaked at 7,704.
This is worth stating plainly, because it is easy to misread as a moral about quality. Over a six-week window a mod sitting at one subscription is not currently unloved — it is structurally stationary. The floor is not a slow lane. It is a parking space.
The control that changes the story
Here is where the pooled numbers would mislead you if you stopped.
The obvious next step is to go looking for what makes floor mods different — and our own database will cheerfully supply an answer. Of the 654,031 one-subscriber mods we hold, 100% have no description and no preview image. Compare mods at ten subscriptions and above, where only 13.1% lack a description and 1.9% lack an image. That looks like the cleanest finding in the dataset: a blank listing is a death sentence.
It is not. It is an artifact of ours, and we should own it here rather than let it stand.
ModSignal runs a housekeeping command that strips the heavy payload — description text and preview image URL — from items below ten subscribers, to keep a six-million-row table affordable on a four-gigabyte box. The threshold is a storage decision. It sits at exactly ten subscriptions, and the seam is visible in the data: all 3,466,771 items under ten subs have both fields empty, with no exceptions. That is not how creators behave. It is how a where subscriptions_current < 10 behaves.
So when a previous article in this series reported that blank descriptions make a mod about ten times less likely to reach 1,000 subscribers, that headline was partly measuring our own trim cursor rather than the decisions of creators. We recomputed it on the retained data — mods at ten subscriptions and above — and the honest version is smaller but still real:
| Listing completeness (mods at 10+ subs) | Mods | Reached 1,000+ |
|---|---|---|
| Description but no preview | 64,159 | 4.1% |
| Preview but no description | 499,767 | 16.0% |
| Description and preview | 3,302,598 | 23.4% |
Among the 3,876,309 mods we still hold full data for, 13.1% have no description. A mod with no preview image reaches 1,000 subscribers 4.1% of the time against 23.4% for a complete listing — roughly 5.7x, not ten. The dramatic ten-fold gap does not survive at the bottom of the ladder where it was originally measured; the direction of the finding does.
The floor is a property of the game
The description trap was a warning: at the bottom of this catalogue, the most striking correlations are the ones most likely to be caused by the collection pipeline. So for the real question — creator, or ecosystem? — we needed a control the pipeline cannot fake.
Start with geography. The floor is wildly unevenly distributed. Across the 234 games with at least 1,000 live mods, 145 keep the floor below 5% of their catalogue and 4 push it above half. Among games with at least 20,000 live mods:
| Game | Live mods | At exactly 1 sub | Floor share | Reached 1,000+ | Subs per mod |
|---|---|---|---|---|---|
| The Escapists 2 | 31,561 | 17,349 | 55.0% | 0.27% | 24.1 |
| KovaaK's | 21,169 | 9,793 | 46.3% | 0.30% | 21.7 |
| Universe Sandbox | 71,940 | 31,737 | 44.1% | 0.26% | 18.3 |
| BattleBlock Theater | 23,284 | 9,855 | 42.3% | 0.34% | 33.8 |
| Human Fall Flat | 96,383 | 37,565 | 39.0% | 5.77% | 621.7 |
| Terraria | 101,305 | 33,298 | 32.9% | 5.57% | 617.9 |
| Space Engineers | 167,330 | 46,671 | 27.9% | 7.20% | 518.2 |
| Portal 2 | 164,311 | 30,253 | 18.4% | 2.78% | 243.1 |
| RimWorld | 60,635 | 491 | 0.8% | 29.49% | 5,892.4 |
| Garry's Mod | 177,144 | 133 | 0.1% | 51.18% | 8,901.6 |
| Euro Truck Simulator 2 | 31,535 | 4 | 0.0% | 39.83% | 5,175.5 |
| Wallpaper Engine | 105,852 | 0 | 0.0% | 93.72% | 18,216.7 |
Wallpaper Engine has no mod at one subscriber at all — 105,852 live mods, 99,204 of them above a thousand. The Escapists 2 has 17,349. Both are Steam Workshop ecosystems of comparable size, and the floor separates them completely.
The floor is also a near-perfect negative proxy for how well an ecosystem works at all. Across the 163 games with 3,000+ live mods, the correlation between a game's floor share and its share of mods reaching 1,000 subscribers is −0.45; against subscriptions per mod it is −0.30. Games that park mods at the floor do not promote them later. Games that do not, do.
But geography alone cannot separate "these games attract worse mods" from "these games fail their mods". For that, use the creators themselves.
Same creator, different game
Take only creators who publish at least five mods in a high-floor game and at least five in a low-floor game. That is 1,162 creators and 44,064 of their mods. Then compare what happens to the same person's work depending on where it lands.
| High-floor games | Low-floor games | |
|---|---|---|
| Their mods | 16,913 | 27,151 |
| Stuck at exactly 1 subscriber | 19.1% | 1.6% |
| Reached 1,000+ subscribers | 8.8% | 12.2% |
The same 1,162 people. Their mods are twelve times more likely to be stranded at the floor when they publish into a high-floor game — 19.1% against 1.6% — and less likely to break four figures there too.
Run it as a paired test instead of a pooled one and the effect gets sharper: among 3,043 creators with at least three mods on each side, 83.6% of creators did better in their low-floor game than in their high-floor game. Geometric mean subscribers: 60.8 in the low-floor ecosystem against 12.7 in the high-floor one — a 4.80x difference for the same creator.
That is the finding. Whatever the floor measures, it is overwhelmingly not a property of the mod or the person who made it. It is a property of the ecosystem the mod was published into.
Which ecosystems are generous, and which are not
The same split appears when you rank games by how many of their mods actually find an audience:
| Game | Live mods | Floor share | Reached 1,000+ |
|---|---|---|---|
| Wallpaper Engine | 105,852 | 0.0% | 93.72% |
| House Flipper | 33,222 | 0.0% | 72.65% |
| Garry's Mod | 177,144 | 0.1% | 51.18% |
| Euro Truck Simulator 2 | 31,535 | 0.0% | 39.83% |
| Cities: Skylines | 161,988 | 7.6% | 36.77% |
| Left 4 Dead 2 | 176,948 | 0.3% | 31.52% |
| Don't Starve Together | 25,818 | 2.7% | 30.22% |
| RimWorld | 60,635 | 0.8% | 29.49% |
| Counter-Strike 2 | 29,914 | 2.8% | 27.73% |
| Kenshi | 22,264 | 1.4% | 26.04% |
| Paralives | 23,436 | 16.6% | 11.17% |
| Aimlabs | 83,213 | 19.5% | 7.16% |
| Universe Sandbox | 71,940 | 44.1% | 0.26% |
| The Escapists 2 | 31,561 | 55.0% | 0.27% |
Note what is not correlated here: size. Brick Rigs has 177,330 live mods and a 0.8% floor share. The Escapists 2 has 31,561 and a 55% floor share. Being a big ecosystem does not protect you, and being a small one does not doom you. The pooled floor share does rise with game size — from 2.4% in games under 500 mods to 12.3% in games above 50,000 — but that is a composition effect, and the within-game comparisons above are what survive it.
What this means if you are publishing mods
One. Read the floor before you read your own mod. If you are publishing into a game where a third to a half of the catalogue is parked at one subscriber, you are not being graded — you are being filed. The Escapists 2 has 84 mods above 1,000 subscribers out of 31,561. Garry's Mod has 90,655 out of 177,144. Those are different industries wearing the same UI.
Two. The floor is not a slow start, so do not wait it out. Across six weeks, 97.9% of mods sitting on it never gained a second subscriber, and 78% of the floor population is more than a year old. Only 1.9% was published in the last 30 days. This is a terminal state that happens to be the modal one, not a queue.
Three. If you have published into a high-floor ecosystem, your own track record is the evidence. The 1,162 creators in this dataset who worked on both sides of the line were twelve times less likely to be stranded on the low-floor side, and the paired test says 83.6% of them did better there. The difference was the game, not the mod. Moving ecosystems is a legitimate, measurable strategy — and it is the only lever in this dataset that the creator fully controls.
Four. Beware the bottom of your own dashboard. We nearly published a finding that blank listings kill mods, because our database said 100% of floor mods had no description. Our database was wrong — a storage trim, not creator behaviour. The strongest correlations in the cheapest part of a dataset are usually the ones your own pipeline put there, and we have corrected that claim above and noted it in the method below.
Where the floor sits, game by game
The two ends of this spectrum, if you want to check your own ecosystem:
- Wallpaper Engine — 105,852 live mods, zero of them on the floor, 93.72% above 1,000 subscribers.
- Garry's Mod — 177,144 live mods, 133 on the floor, 51.18% above 1,000.
- House Flipper — 33,222 live mods, 1 on the floor, the best hit rate on the platform at 72.65%.
- RimWorld — 60,635 live mods, 491 on the floor, 29.49% above 1,000.
- The Escapists 2 — 31,561 live mods, 17,349 on the floor — 55% of the catalogue, and only 84 mods above 1,000.
Related reading from this series: the tag cliff on the 1.8M mods that carry no tags at all, the crowding penalty on why rival density beats the hour you publish, and the audience portability study on what happens when a creator moves a hit from one game to the next.
Method: 7,343,108 active Steam Workshop items tracked by ModSignal as of 27 September 2026, of which 6,003,574 carry at least one subscription, across 396 games and 1,692,039 creators. All figures are a single consistent read of the live database on that date; the counters move continuously, so re-runs differ by a handful of items. "At one subscriber" means subscriptions_current = 1 exactly, as reported by the Steam Web API — the value is an integer counter, so the pile-up at 1 is a real population and not a rounding artefact. Escape rates are computed from 27,511,046 snapshots covering 41 days (15 August – 27 September 2026): for each item we take the first reading in the window as its starting value and the maximum later reading as its best, restricted to items active at the end of the window. Because the window is six weeks, these are short-horizon escape rates and should not be read as lifetime survival. "High-floor games" are the ecosystems at or above a 20% floor share among games with 20,000+ live mods; "low-floor games" are those at or below 5%. The creator control keeps creators with at least five mods on each side (1,162 creators) or three on each side for the paired test (3,043), compares mean log(subscribers + 1) per side, and counts ties as non-wins. Two caveats we hold: the Steam lifetime view counter is unreliable in the low bands — 32.7% of live mods show fewer views than subscribers, which is impossible — so no view-based cut is used as headline evidence here; and votes_up_current is zero for all 6,003,582 tracked live items, so no upvote-based claim can be made from this dataset at all. Finally, the description and preview-image fields are stripped by ModSignal's own cold-item trim for all items below ten subscriptions (3,466,771 items); the retained-data comparison in this article is computed on items at ten subscriptions and above only.