13 Rules not to f**k your studio

May 10, 2024
Jon Bellamy
min read

Let's start with this: we all know how hard forecasting can be. Whether you've got Excel sheets cluttering your desktop or scrambling through last-minute projections the night before a board meeting, we've all been there. But if we keep clinging to outdated norms and predictable assumptions like "365 days is the perfect payback period," we'll look back in a few years and wonder what on earth we were thinking.

We're still stuck in the taxi era of forecasting, but the landscape is changing quickly (think of the introduction of Uber), and we've got to embrace that disruption or get left behind. So, I've put together 13 rules you need to follow to stop f**king your studio.

You're going to want to do these things - don't.

1. Don't rely on benchmarks

It doesn't matter what the person next to you is doing - it's irrelevant. What matters is your numbers, against your costs of install, and nothing else.

Sometimes it's handy to look at other competitors in the space,  to get a feel if you're within the ballpark, but it's utterly meaningless. The calculus is your own, and it only applies to you.

2. Don't assume CPIs stay flat or decline over time

I know a lot of people are guilty of this, but don't do it. We all assume that in 2026 CPIs are going to be roughly where they are today. That's rubbish because the more you spend, the higher your CPIs go. Nothing novel here, but the point is, context is important. You should always be looking at payback periods with a scale of spend included.

3. Don't assume that your product CLV will "grow into" CPIs

I've heard "We're going to spend this year on year" or "We can't really afford to spend at this level" and "We might lose some money on some of these campaigns" followed by "But don't worry, the product will catch up and we'll make the environment better. We'll run some LiveOps campaigns. We'll push ARPDAU up. We'll grow into these CPIs." That's a big risk and it's so dangerous to do.

4. Don't grow installs/DAU to "show growth"

We're not looking at installs and DAU. It really doesn't matter how many installs you have. One user at $10,000 a year is the same as 10 users at $1,000 a year. The math is the math. It may feel different. You might like a high DAU or MAU or install number, but unless you're hypercasual and in very few other cases, your installs and DAU numbers really are somewhat irrelevant.

5. Don't assume your first (golden cohorts) from iOS, US are indicators of expected performance

Don't assume that your golden cohorts, and your iOS tier one, will continue to get you amazing results. You may think "We're getting paybacks in six days. It's going to be like Supercell."

Never is. Remember, early spending looks great. Middle spending looks okay. At some point you run out of runway. Don't project based on your very best cohorts right at the beginning.

6. Don't extend your payback periods tolerance without statistical rationale

This is critical. If I'm going to get a tattoo, it will be this: "Don't extend your paybacks". Seriously don't do this.

You're not going to want to do these things - do them.

7. Assume CPIs increase every year, no matter what you do

This happens. People are getting better at targeting. People are spending more and more every year. It's a more competitive environment that breeds inflation, like in any other market for any other product. Installs are still just a product.  They go up over time.

On average, I think over the last five years or so, the increase been something like 20% on average per year of natural CPI inflation across the boar. This means, just to stay still, your product must be improving by 20% every year.

8. Be honest about whether you can focus on blended ROAS or paid-only ROAS

What normally happens is people spend a bunch of money. They look at paid only ROAS and then the numbers start to slip and they say, "Well, we can look at blended. That's okay, we have 20% more headroom." That's, once again, super risky.

"Paid only is always safe. Blended isn't always safe. I'm a fan of paid-only ROAS analysis."

9. Use net revenue in your ROAS/payback calculations

Use net of platform fees and net of IP licensing royalties. It's your dollars in and your dollars out. Be really harsh with yourself. If there's money coming out that is unavoidable, take it out of your revenue and use that as your ROAS calculation.

10. Align data points from your MMP and your internal KPI datasets

If you're working with the likes of an AppsFlyer or an Adjust, it can be really hard and messy to align that data set with all of the LTV stuff that you have in Firebase or some other database -- especially when you're doing your attribution and you're looking at how your UA spend is performing.

It's hard. Find a way to connect the two and the world opens up for you. There is no one-size-fits-all solution for this, but it's getting slightly easier over time.

11. Reduce UA spend when payback periods extend

When your payback periods extend, remember the blue line from point 6. You set this risk threshold and said no matter what happens, day 180 is the limit for me.

Don't get lax with that. If you hit this wall and you go beyond this wall, reduce your spend. It may hurt your growth numbers, doesn't matter. It's discipline, it's profitability, it's the right thing to do. If you run a studio, this is your responsibility.

12. Set a minimum rate of return of UA spend

Ultimately, whether we like it or not, we're all building creative products that are amazing. These games serve to bring joy to thousands of millions of people. But we are, as studio leaders and as people responsible for departments, asset allocators. Our job is to deploy resources, people, capital, time, focus, and expect a rate of return on those things that justifies why these businesses were started in the first place, why we're running these departments, and why we built this game.

"Our job is to deploy resources, people, capital, time, focus, and expect a rate of return on those things that justifies why these businesses were started in the first place."

13. Adjust your ROAS to reflect the time value of money

Lastly, you need to adjust your ROAS for the time value of money. I'm not going to get too detailed on this or go all corporate finance. Angus Lovitt goes into detail on this is his presentation: You're probably wasting money on UA.

However, simply put, a dollar today is worth much more than a dollar tomorrow. And dollar today is worth much, much more than a dollar two years down the road. So, if you have a payback point that is two years into the future, you have to ask what it's actually worth in today's money once you adjust for inflation, you adjust for all of the risk? The answer: significantly less than the nominal value.

If you employ these 13 best practices, life gets easier. You will be less likely to:
  • Overspend on UA
  • Underspend on UA
  • Back the wrong product(s)
  • Set false expectations
  • Make poor roadmap decisions
  • Make poor distribution decisions
  • Face cashflow issues
  • Be negatively surprised

All of this is the manifestation of being empowered by data and following the numbers. You build forecasts this way and it's unavoidable.

Remember, forecasting is like eating healthy. Feels good on January 1st - sucks every other day for the rest of the year. But if you want to survive and live long and be healthy, it's necessary.


This blog post was adapted and condensed from Jon's GDC 2024 presentation: How not to f**k your studio: Forecasting done right. You can download his full presentation here.

Related Articles