Direct answer — How do you decide whether to scale, extend or cancel a mobile game after soft launch? You decide by writing the soft launch kill criteria before the first cohort arrives: for each metric, a go line, an extend band and a kill line, plus a named decision owner, a review date and a budget cap. At the review, the owner compares the data to the pre-registered lines and picks one of three outcomes; nobody rewrites the lines in the meeting. An extension is only legitimate if it comes with a new hypothesis, a capped budget and a fixed date, otherwise it is a cancellation being postponed.
Most studios do not lack metrics in soft launch; they lack a decision process. The mobile game go/no-go decision slides from month to month because nobody agreed in advance what “no” would look like.
I have spent more than 20 years in the games industry, at Gameloft, SFR, Blacknut and Impulse Media Hub, including a business unit with EUR12M+ of P&L. The soft launches that hurt most had ambiguous numbers and no rule for reading them.
This article covers the governance layer: pre-registering thresholds, who decides, extensions, and pivot versus cancel. For the soft launch metrics thresholds themselves (D1, D7, monetisation and CPI scale bands), use the go / extend / kill matrix in our soft launch guide.
Why soft launch kill criteria fail without governance
Soft launch kill criteria are the pre-agreed conditions under which you stop investing in a game during soft launch. A go/no-go decision is the formal moment when a named owner compares the data to those conditions and commits to scale, extend or cancel.
Even the best-run studios find this hard. Supercell’s Brawl Stars was “on the edge” between a global release and cancellation during its 18 months in soft launch, according to a GDC talk reported by PocketGamer.biz. Its GM Frank Keienburg explained that when metrics are “an absolute no go, from our experience, then the team will probably kill their darlings”, but “in Brawl Stars’ case, for example, it was very much on the edge.” The game launched globally in December 2018 and has since “generated more than $2 billion in worldwide revenue”.
The lesson is not “keep going, it might be Brawl Stars”. Borderline results are the normal case, and that is exactly where an unstructured review defaults to inertia.
Pre-register the go/no-go decision before the data
The single most effective practice I know is a one-page decision memo, signed before the soft launch starts and reopened at every review. It is the soft launch equivalent of a pre-registered study: the lines are fixed before anyone sees the result.
| Memo field | What it contains | Common failure it prevents |
|---|---|---|
| Question | The one thing this phase must answer (core loop, retention curve, monetisation, scale) | Reviewing everything at once and deciding nothing |
| Metric and lines | Go line, extend band, kill line, taken from your matrix | Moving the line to wherever the data landed |
| Iterations allowed | Number of builds before the kill line applies | “Let’s see one more build” forever |
| Decision owner | One named person | Split votes that default to extension |
| Review date | Fixed calendar date | Reviews that drift with build delays |
| Budget cap | Maximum spend for the phase, including UA and team | Runway disappearing before the answer arrives |
| Kill consequence | What happens to team, IP and players if the answer is no | Fear of the unknown keeping the game alive |
Take the thresholds from a matrix you trust, such as the one linked above, and adapt them to your genre and economics; do not invent them in the memo. Market choice belongs in the memo too, because a market picked for cheap installs can answer a different question than the one you are asking; our guide to choosing soft launch markets covers that trade-off.
When studios bring us in through our mobile game consulting work, writing this memo is often the first deliverable, before any metric review. It changes every meeting that follows.
Who decides, and how often
A game greenlight is a P&L decision, not a vote. My working setup:
- Decision owner: the GM or head of product at a studio; the portfolio manager at a publisher. One name.
- Recommender: the game lead presents the data against the memo and a recommendation.
- Reviewers: finance checks the cost of the next phase; UA checks whether the CPI and LTV read is credible.
- Cadence: a short weekly metrics check, and a formal decision review only at the end of each iteration cycle, on the memo date.
Supercell runs this the other way round. Keienburg said: “We’re not making top down decisions.” That works where teams genuinely “kill their darlings”. In most studios I have worked with, a team judging its own game finds a reason to continue, so a single owner is the safeguard.
The “one more month” trap: when to extend a soft launch
The most expensive word in soft launch is “almost”. The sunk cost effect, the tendency to keep investing because of what has already been spent, is strongest when results are close to the line and the team has worked on the game for years.
Supercell’s Clash Mini shows how long “almost” can last. It closed “nearly two and a half years on from beta release”, and the team wrote that it “was a good game but not the game that would ultimately fulfill our dream”, adding that “our quality bar was not met even after all those improvements”.
An extension is legitimate only if it meets four conditions, all written in the memo:
- A new hypothesis. A specific reason why the next build should move the failing metric, not “more data”.
- A capped budget. As a GGA rule of thumb, an extension should cost a clearly bounded fraction of the original phase, not reset it. Treat this as an illustrative template, not a benchmark.
- A fixed date. One iteration cycle, then a new decision.
- A trend. The metric under review must be improving build over build. A flat line after several builds is a kill signal, not an extension signal.
If two of the four are missing, the honest answer is no. Approve extensions one cycle at a time, never open-ended.
Stuck between extending and cancelling? Book a Strategy Call and we will review your soft launch data against a decision memo before your next review.
Pivot vs kill: what to keep
To cancel a mobile game in soft launch is not always to throw everything away. The useful distinction:
| Situation | Decision |
|---|---|
| A mode, feature or theme clearly works; the rest does not | Pivot: new project, new memo, new budget |
| Core loop does not retain after several builds | Kill the game; keep tech and learnings |
| The fix is a rebuild of the core | Kill; a rebuild is a new greenlight, not an extension |
| Strong concept, wrong host | Fold it into an existing live game |
Clash Mini followed the last path: the team said it would commit “to bringing the Clash Mini experience to Clash Royale to reach its fullest potential”. Hay Day Pop took the clean exit. Supercell said “we feel that Hay Day Pop does not meet the standards we have set for ourselves and our players”, and PocketGamer.biz reported that, citing Sensor Tower, “the game has generated 360,000 downloads and accumulated $990,000 in revenue”. Everdale closed on the same logic after “Failing to meet expectations”, with purchases transferable to other Supercell games. For the player-facing side of a closure, see our playbook for sunsetting a mobile game.
How publishers’ portfolio managers read the decision
A portfolio manager does not ask “is this game good?” but “is this the best use of the next euro across the slate?”. Three questions matter:
- Was the soft launch long enough to answer the question? Squad Busters is the counter-example. PocketGamer.biz reports on its release that “Internal pressure and the feeling that Supercell’s risk appetite was getting too constrained played a role”, and that “the one month soft launch didn’t allow for testing of the meta and long-term retention and monetisation beyond D7.” Game lead Eino Joas wrote: “We had 40 million pre-registrations, 75 million installs, but not a lot of those players stuck around.” It later became the first live game the studio shut down.
- Is the extension cheaper than the next prototype? If not, the portfolio should fund the prototype.
- Was the memo respected? A team that moved its own lines is a governance risk on the next project too.
What we could not verify
- Industry kill thresholds. No Tier A dataset publishes the thresholds at which studios cancel games. The memo, the extension conditions and the budget-cap rule above are GGA’s own P&L-based method, labelled illustrative.
- Supercell’s internal KPIs. None of the cited articles give the metrics behind Brawl Stars, Clash Mini, Hay Day Pop or Everdale decisions; the Brawl Stars talk only says the team “compared against Clash of Clans and Clash Royale data”.
Conclusion
Soft launch kill criteria are a governance tool before they are a metric. Write the lines, the owner, the date and the budget cap before the data, review against the memo, and approve extensions one cycle at a time with a new hypothesis. Borderline games will still be hard calls, but they will be calls rather than drift.
Need a senior read on a go, extend or cancel decision? Talk to Game Growth Advisor to turn your soft launch data into a defensible decision.