Let’s get two things straight about soft launching a game in 2025.

First: The old playbook is dead. The days of finding your KPIs and then flipping a giant UA switch on launch day are over. Growth is gradual. Strategy is nuanced. Multiple changes to the platforms and app stores have had a significant impact on the volume and type of traffic downloading your game. Teams are also (thankfully) much more sophisticated when it comes to UA strategy and thinking about UA as a source of growth. So that means for the most part the days of assessing the KPIs of your game and going all in with a huge push on launch day are behind us. It makes much more sense to gradually ramp up your UA efforts to achieve scale. In other words, you don’t (or shouldn’t) use the soft launch process to turn on a switch when things are ready.

Second: Your KPIs are a lie. Well, sort of. A game doesn’t have KPIs. It has features. Your KPIs are just the reflection of a specific audience interacting with those features. Change the audience with a new UA campaign, and your KPIs will change too. So you can’t consider the metrics of your game in a vacuum. You always have to consider the KPIs from the perspective of your UA strategy – and think about performance in terms of ROAS. 

Despite those caveats, you still need a soft launch phase to look specifically at your game metrics. The key insight: soft launch isn’t about minimizing costs—it’s about maximizing data quality per dollar spent. UA isn’t marketing spend; it’s your data production budget. You have to invest time and money into your game at a lower scale to assess what performance could look like at a larger scale. And even though you can’t evaluate the performance of your game independently from the UA strategy, when you improve your game’s performance you are making changes in terms of features and balancing. And there are some core features/balancing principles will get metrics up no matter the campaign type. Your day 1 conversion could be 0.5 or 2% depending on the campaign you’re working with. But whatever the number might be, surfacing a conversion offer to your players within 10 minutes of install will get it up.

So what has not changed is that a soft launch process is the process in which you produce data about your game’s performance to decide whether or not to scale it. And get data along the way to see if you’re on the right track or if you need to pivot – or kill. Since the soft launch process is a process used to produce data about your game, you need to be aiming for 2 things: 1) produce data that is relevant and 2) produce data that is accurate. This creates a paradox nobody talks about: You crave perfect, exhaustive data, but you’re forced to make million-dollar decisions with messy, incomplete information. On one hand, you want data that is as perfect and exhaustive as possible. On the other hand, you need to be strategic and make decisions with imperfect and incomplete data. I’ve been exposed to a lot of conversations that focus on the former – much less on the latter.

The solution isn’t to wait for perfection. It’s to be ruthless about focusing on the KPIs that give you the most accurate reading, the fastest.

1) The Only KPIs That Matter: Retention, Conversion, and Redeposit

Making sure you have a potentially profitable game on your hands means validating the retention and monetization potential of your game. Retention is the straightforward one to measure. You can get very accurate and reliable readings with few users

Let’s say you want to measure retention with 95% confidence. Below are the users you need in each cohort to estimate your retention numbers – that number varies with the margin of error you’re comfortable with. In any case you don’t need that many users.

Expected Retention	Margin of Error ±1%	Margin of Error ±2%	Margin of Error ±3%	Margin of Error ±4%	Margin of Error ±5%
45%	9508	2377	1057	594	381
30%	8068	2017	897	504	323
20%	6147	1537	683	384	246
15%	4899	1225	544	306	196
10%	3458	865	384	216	139

Let’s say you have a soft launch and expect your day 30 retention to be at 10%. You shouldn’t be ok with a 5% margin of error (meaning your d30 retention is between 5 and 15%). But if you were, you would need a measly 139 installs to validate that retention number (don’t believe me? I used Gemini to build a cohort calculator to prove it. You can check it out here.)”

The same rationale applies for conversion. Only here because the % of players converting (making an IAP purchase) is lower than retention, you’ll need more users to assess your conversion numbers with the same confidence. But the number of installs needed is still reasonable.

Expected Conversion	Margin of Error ±0.25%	Margin of Error ±0.5%	Margin of Error ±0.75%	Margin of Error ±1.0%	Margin of Error ±1.5%
1.50%	9,083	2,271	1,010	568	253
2.50%	14,706	3,677	1,634	920	409
3.50%	20,027	5,007	2,226	1,252	557
4.00%	22,665	5,667	2,519	1,417	630
5.00%	27,948	6,987	3,106	1,747	777

Now the tough part comes from when you want to assess LTV.

So, what about LTV in all this? Avoid it. In soft launch, trying to measure LTV is a trap. It’s a continuous metric with wild variance. And the challenge here lies in the fact that there is huge variation between individual LTVs. Not only do the vast majority of players not spend. Among those spending users there is also great variation in spending. Even among spending users, a few whales can skew your numbers into oblivion, giving you a dangerously false sense of security. It’s costly and unreliable when you need to be cheap and certain.

This is why redeposit rate is a more practical metric to measure the depth of spending. Like conversion and retention it’s a binary metric. It refers to something that happens or not. So you can get much more reliable readings with fewer players. It is also a very good proxy for the quality/depth of spending. Conversion is great (and important) to monetize your game. But you can expect 40-50% of your paying users to spend just once. And those paying users making just 1 IAP purchase will probably account for 5-10% of your title’s revenue. Redeposit rate helps get a sense of how well your game will monetize.

Now that you know WHAT to measure—retention, conversion, and redeposit—the question becomes: how much data do you need to make a confident decision? This is where sample sizes become critical, and where most teams make expensive mistakes.

2) Day 30 is Your Decision Point (Here’s Why)

I’ve discussed previously the different cases where testing monetization from the start makes sense. Most of the time, your KPI phases should not be retention first, then monetization second. Having a strong retention by no means guarantees you have a strong business on your hands. You need to test different slices of your entire game (retention + monetization). Day 1, then day 7, and day 30.

The whole point of the soft launch process is to get a partial snapshot of your game’s performance in order to assess what it might end up looking like if you operate that game for years. Your business model might be based on the user’s d360 performance. But you don’t want to be validating the game’s business potential on d360 metrics in soft launch. You need a good proxy for the timeframe you are looking at.

Here, it’s not one size fits all. But there is a method to making the right choice for you. Say you  want to make a game and have some strong data on how similar games perform over a long period of time. What I recommend is looking at the relevant timeframe for your business model, and see what % of that is realized at different points in time. In the example below I’m taking a d360 target timeframe, and looking at what to expect in terms of % of d360 metrics achieved by days 1/7/30/90.

Day	% of d360LTV	% of d360Conversion	% of d360Redeposit
1	10%	30%	20%
7	20%	50%	40%
30	35%	75%	65%
90	60%	85%	80%

As you can see not all metrics are realized at the same speed. Only 35% of d360 LTV is achieved by day 30. But you will see that 75% of players converting and 65% of players having spent more than once will have done so by day 30. In other words a majority of revenue still needs to be realized post d30. But both in terms of conversion and redeposit, you’ve captured the vast majority of players who are going to contribute to your bottom line by then. The reality is that successful games perform well from the start (i.e. day 1/3/7). I have never seen or heard of a game that started poorly and somehow managed to make up for lost ground day 14 or 30 onwards (here also, hope and wishful thinking will do more harm than good). If you are looking at comparable performance, you will more likely than not have the info you need by day 30 to make a call. And if your game doesn’t appear to be performing well by day 30, then it most certainly means it will never perform well.

These metrics are useless without sufficient sample size to make reliable decisions. You need enough users to hit statistical confidence on retention, conversion, and redeposit by Day 30. That’s where UA becomes your data engine—not a marketing expense, but an investment in decision-making speed.

3) UA is Data Production, Not Marketing Spend

The reason you are soft launching is to assess whether or not your game could be a viable business. The data you need to make that assessment is something valuable you need to ‘produce’—and UA is your production engine.

If you adopt a red light by default mindset, you’re operating under the assumption your game will more likely than not fail. This changes how you think about UA spend entirely. I’ve often seen situations where various stakeholders worry about spending ‘too much’ on UA in soft launch—but worry much less about the running OPEX costs while you’re producing the data needed to assess your game’s potential.

This thinking is backward. Let’s do the math on two approaches to getting the statistical confidence you need on your Day 30 KPIs:

Scenario A: The “Cautious” Approach
You spend a conservative $30k/month on UA. It takes you 3 months to get sufficient sample size for confident decisions.

• UA Spend: $30k × 3 = $90k
• OPEX: $200k × 3 = $600k
• Total Cost to Kill: $690k

Scenario B: The “Data-Driven” Approach
You spend an aggressive $100k/month on UA. You get statistical confidence in just 1.5 months.

• UA Spend: $100k × 1.5 = $150k
• OPEX: $200k × 1.5 = $300k
• Total Cost to Kill: $450k

So on one hand yes you are spending more on UA ($150k vs $90k). But spending more allows you to kill the game faster – and in doing so saving 35% of your total costs. The purpose of soft launch UA isn’t just to get data. It’s to buy a decision. The faster you can buy that decision, the cheaper it is. The purpose of soft launch UA isn’t acquisition. It’s buying statistical confidence. The faster you can buy that confidence on the metrics that matter, the cheaper your ‘red light by default’ process becomes. The cost isn’t the UA spend. The cost is the time you waste with insufficient data.

The soft launch paradox has a clear solution: Focus ruthlessly on binary KPIs (retention, conversion, redeposit) and use aggressive UA spend to validate them by Day 30.

Every dollar spent on UA is buying you certainty—and certainty is the most valuable currency in game development. The teams that understand UA as their data microscope, not their marketing budget, are the ones making faster, cheaper decisions.

Your game’s fate is sealed by Day 30. The question isn’t whether you can afford to spend on UA—it’s whether you can afford not to.