Reported: 29 August 2026
Experience: GET OUT (extraction shooter, live since February 2026)
Summary
Running Sponsored Ads on my experience causes measured playtime, session length, and D1 retention to fall by large margins. Turning ads off causes all three to recover immediately, with no other changes to the experience.
Because these are the same signals discovery uses to evaluate experience quality, the effect is that buying Sponsored Ads makes an experience look worse to the recommendation system than not buying them.
I do not believe this is intended. It appears to be an unintended interaction between two systems: the advertising product is feeding traffic into the ranking product that does not behave like organic traffic, and nothing separates the two.
The evidence
I ran Sponsored Ads consistently for months, then turned them off for roughly one to two weeks in August. No updates shipped during that window. Same build, same experience, adjacent weeks. Every relevant graph changed immediately on ads being disabled.
- Screenshot shows ads turned OFF, and then ON for the remainder of the graph.
Average playtime
| With ads | Ads off | |
|---|---|---|
| Average playtime | 11.1 – 12 min | 28.0 – 29.9 min |
| Percentile band, ads off | 50th – 90th (24.0 – 46.6 min) |
Average session time
| With ads | Ads off | |
|---|---|---|
| Average session | 9.2 – 10 min | 19.0 – 21.3 min |
| Percentile band, ads off | 50th – 90th (14.5 – 25.9 min) |
D1 retention
| With ads | Ads off | |
|---|---|---|
| D1 retention | ~3% | 5 – 8% |
The effect appears to scale with spend. Higher spend produced lower averages.
Across all three metrics the effect is consistent and large: playtime roughly halves, session time roughly halves, and D1 retention falls by around 40 to 60%. Three independent measures moving together in the same direction, on the same schedule, with no build change between them.
The important part: with ads off, this experience sits in the 50th to 90th percentile band on playtime and session time. That is a normal-to-healthy experience by the measures that describe whether players are enjoying themselves. With ads on, the same experience reports numbers suggesting something is badly wrong with it.
Nothing about the experience changed. Only the traffic source did.
Why this is a bug and not just poor ad performance
If ad traffic simply converted poorly, that would be an advertising effectiveness problem and my problem to solve.
The issue is that the same numbers are used to decide how the experience is recommended organically. The mechanism:
- I pay for Sponsored Ads.
- That traffic produces short sessions and does not return.
- My averaged playtime, session length, and retention fall sharply.
- Discovery reads those averages as evidence the experience is worse than it is.
- Organic reach falls.
Paying for the platform’s advertising product measurably harms performance in the platform’s own recommendation system. That is a perverse outcome and I do not believe any part of it was designed to work this way.
Supporting observations
New players are not reaching a single first-screen button
In a recent week where roughly 85% of new plays came from Sponsored Ads, 45% of new players dropped off at the Deploy step.
Deploy is one button on the first screen a player sees. No navigation, no menu to find, no prior step. Nearly half of new players never pressed it.
Recent campaign spend has contributed roughly 5 to 10 concurrent players.
CTR decline on unchanged creative, reported by multiple developers
Several other developers who run ads reached out independently after noticing the same thing. Identical creative that performed well earlier in the year now returns measurably worse CTR at comparable spend, starting around the changes announced on 15 June 2026.
This is the observation I am most confident about, because it is not specific to my account.
Ads Manager reporting delay
Play data no longer appears for days after the plays occur. Expected behaviour is that campaign performance surfaces promptly enough to adjust spend while a campaign runs. Actual behaviour is a multi-day lag, so spend decisions are made blind and corrected after the budget is already committed.
This compounds the above, because it becomes very difficult to tell whether a campaign underperformed or whether the data has simply not arrived yet.
Reproduction: start a campaign at a known daily spend. Record what the manager reports at one hour, six hours, and twenty-four hours. Record the same figures at seventy-two hours. Play counts for the earlier periods continue increasing well after those periods ended.
Wider context
I develop in the extraction shooter genre and track other titles in it. Figures observed 29 August 2026:
| Title | Peak | Now |
|---|---|---|
| HAVOC (paid access, ~250–350 Robux) | 1000+ sustained | ~180 |
| DESERTSTORM | ~1000 | ~80–120 |
| Escape from Roblox | low hundreds after a TikTok surge | ~15 |
| GET OUT | 500–800 sustained (May) | declining since mid-June |
Four titles, different teams, different monetisation models, different update cadences. All shipped major updates through their declines. All are down since roughly the same date and none are recovering.
I raise this as context rather than as the central claim. These are experiences that depend heavily on population because the gameplay is PVP, which makes them unusually sensitive to any reduction in discovery. Whatever is happening shows up in this genre earlier and more sharply than it would elsewhere.
What I considered and ruled out
- Genre fatigue. Probably a factor. Does not explain titles declining while shipping major updates, or a paid-access experience losing most of its population months after launch.
- My own updates. I have made large changes and not all landed well. That explains my experience. It does not explain three others, and it does not explain the ads-off recovery, which happened with no update at all.
- Seasonality. Mid-June to late August is normally a strong period.
- Normal post-peak decline. Expected individually. Four titles simultaneously with no recovery over ten weeks is not.
What would resolve this
-
Separate ad-attributed sessions from organic in the experience-quality signals used for discovery. At minimum they should not be weighted identically. An experience should not be penalised in recommendations for buying traffic through the platform’s own advertising product.
-
Surface the split in analytics. I cannot currently see organic and ad-attributed metrics side by side. I only found this by turning ads off for two weeks and watching every graph move. That should not be how a developer discovers it.
-
Verify whether ad-attributed plays behave differently from organic joins on session length and on progression past a first-screen button press. My data says they do, by a factor of roughly two to three on playtime.
-
Reduce or document the Ads Manager reporting lag.
I have detailed analytics for GET OUT going back to February, including the full ads-on and ads-off comparison, and I am happy to share them privately with anyone on staff who wants to look.
Page URL: https://create.roblox.com/advertise