Game Analytics 101: Why waiting "One more week" can kill your game!
This episode, Two and a Half Gamers sit down with analytics legend Russell Ovans—the mind behind Professor ARPDAU and the author of game analytics’ “bible.” Learn the brutal truth about retention, ARPDAU, LTV, and the “golden cohort” trap. If you want to scale your game profitably in 2025, you need these formulas, tools, and lessons - no hype, no BS, just results.
You’ll learn:
Why retention is the only metric that matters (until ARPDAU takes over)
How to model and predict your LTV, player days, and payback window using real retention curves
The difference between analytics and data science, and why cohorts are your best friend
How to avoid burning your UA budget chasing “golden cohorts” or missing your D7 ROAS target
Free tools at arpdow.com to plug in your data and get real answers
Key Takeaway:
Retention plus ARPDAU equals real growth. Track cohorts, model your curve, and never fall for the golden cohort trap or ad network happy talk. If you want more, arpdow.com has the tools.
Get our MERCH NOW: 25gamers.com/shop
--------------------------------------
PVX Partners offers non-dilutive funding for game developers.
Go to: https://pvxpartners.com/
They can help you access the most effective form of growth capital once you have the metrics to back it.
- Scale fast
- Keep your shares
- Drawdown only as needed
- Have PvX take downside risk alongside you
+ Work with a team entirely made up of ex-gaming operators and investors
---------------------------------------
Vibe. Vibe is the leading Streaming TV ad platform for small and medium-sized businesses looking for actionable advertising campaign performance.
https://www.vibe.co/
---------------------------------------
For an ever-growing number of game developers, this means that now is the perfect time to invest in monetizing direct-to-consumer at scale.
Our sponsor FastSpring:
Has delivered D2C at scale for over 20 years
They power top mobile publishers around the world
Launch a new webstore, replace an existing D2C vendor, or add a redundant D2C vendor at fastspring.gg.
---------------------------------------
This is no BS gaming podcast 2.5 gamers session. Sharing actionable insights, dropping knowledge from our day-to-day User Acquisition, Game Design, and Ad monetization jobs. We are definitely not discussing the latest industry news, but having so much fun! Let’s not forget this is a 4 a.m. conference discussion vibe, so let's not take it too seriously.
Panelists: Jakub Remiar, Felix Braberg, Matej Lancaric
Special Guest: Russel Ovans
https://arpdau.com/
https://arpdau.com/ltv
Join our slack channel here: https://join.slack.com/t/two-and-half-gamers/shared_invite/zt-2um8eguhf-c~H9idcxM271mnPzdWbipg
Chapters
00:00 Introduction to Analytics in Gaming
05:38 Russell Owens' Journey in Game Analytics
08:29 Understanding Analytics vs. Data Science
11:37 The Importance of Cohorts in Game Analytics
14:43 Retention Metrics and Their Significance
18:44 Key Performance Indicators for Game Success
21:39 The Relationship Between LTV and CPI
24:37 Predicting Retention and Its Impact on Game Design
30:41 Understanding Retention Metrics
33:42 The Importance of Retention in Monetization
36:31 Expected Player Days and LTV Calculation
43:43 Tools for Predicting LTV and ROAS
50:41 Final Thoughts and Homework for Game Developers
---------------------------------------
Matej Lancaric
User Acquisition & Creatives Consultant
https://lancaric.me
Felix Braberg
Ad monetization consultant
https://www.felixbraberg.com
Jakub Remiar
Game design consultant
https://www.linkedin.com/in/jakubremiar
---------------------------------------
Please share the podcast with your industry friends, dogs & cats. Especially cats! They love it!
Hit the Subscribe button on YouTube, Spotify, and Apple!
Please share feedback and comments - matej@lancaric.me
Listen and follow along
Transcript
So, why is retention the most important metric?
And it's pretty obvious, but you can only monetize and install on the days they play.
If they don't play your game, you can't make money off them.
Yeah, but you can't buy your dinner with retention.
No, no, and you can't.
Yeah, I made that point very clear.
You can't buy dinner with five-star reviews either.
Yeah.
It's 4 a.m.
and we're rolling the dice.
Mate drops, knowledge made of gold and ice.
Felix with ads making those coins rise.
Jack up designs, worlds chasing the sky.
We're the two and a half gamers, the midnight crew.
Talking UA adverts and game design too.
Matej, Felix, Shaku, bringing the insight.
We're rocking those vibes till the early daylight.
The J-U-A, master eyes on the prize.
Tracking data through the cyberspace skies.
Felix stacks colors like a wizard in disguise.
Jackups crafting realms, lift us to the highs.
Two and a half gamers talking smack.
Slow, hockey, sick, got your back.
Ads are beautiful, they like the way.
Click it fast, don't delay.
Uh-huh.
Uh-huh.
Uh-huh.
Uh-huh.
Uh-huh.
Felix, are you gonna take us?
Hello, everyone, and welcome to another fantastic episode of the Two and a Half Gamers podcast, where you can stay two and a half steps ahead of the mobile gaming industry.
Today, we have a very special episode because we are joined by someone who will take us where we've never been before and dive deep into analytics.
But before introducing our guest, I'm Felix Brauberg.
My name is Matiel Antevich.
And I'm Jakobremier.
And
we
are
your host.
So we have the privilege today to be joined by Russell Owens.
Is that how I say it right?
Or am I butchering that?
No, Ovens.
That's right.
Ovens, Owens.
And we used to work together a brief amount of time at Eastside Games.
And one of the things that you were extremely good at was analytics.
But why don't you introduce yourself?
Sure.
Well, first of all, thanks for having me, guys.
Big fan of the show since I was a little kid.
Used to gather around the wireless and listen to you talk about ad placements.
So today's talk, I asked Felix, like, you know, if I came on, what do you want me to talk about?
Should I do a deep dive into something specific that you guys are interested in?
And he said, no,
let's keep it broad.
So this is is everything you always wanted to know about analytics, but two and a half gamers were afraid to ask.
And yes.
So,
to start with,
I'm going to do a quick.
Oh, hello.
Hello there.
I didn't see you.
Thank you very much for coming to this episode.
It's brought to you by our sponsors, PVX Partners, the simplest and most effective credit line for marketing.
If you're scaling your business and need funding, here's the deal.
With PVX Partners, you can grow your game easily and your business.
There is no equity dilution, you can repay as you get paid, and it's flexible, fast, and fully tailored to your growth.
So you can secure funding in 30 days and get a term sheet in just 24 hours.
Don't you wait, take control of your UA budget without giving away
a piece of your company.
Visit pvxpartners.com and see how they can help you grow your game faster and stronger.
Don't wait.
Go to pvxpartners.com.
And while you have you here, we also want to thank our sponsors, vibe.co, because this episode is brought to you by them as well.
They can help you grow your brand on CTV ads.
With Vibe, you can launch high-impact campaigns across top streaming platforms like Hulu, Roku, Toby, and whatever else.
I mean, all from one simple dashboard.
We don't have all of these in Europe.
Anyway, so it's built for brand all sizes.
It's rated on 4.8 out of 5 on G2 for a reason.
Join over 5,000 marketing teams already scaling with streaming TV.
Visit vibe.com to get started.
That's vibe.co.
Join them.
And thank you very much for joining us for this special episode.
Now let's get back to it.
Thank you.
Kind of.
Yeah,
take on my journey into games and what I've been involved with and why I
think I might know what I'm talking about.
The story begins in 2007 when a company that I had called Backstage Technologies
put an app out on Facebook.
This was the very first days of when the Facebook API was available.
And I wanted to do an app where you could buy virtual scratch lottery tickets and
send them to your friends and play them.
And my team, who were all into games, they gamified it.
So it became a collecting thing where you had prize sets to collect.
And
some of the prizes you could only win with Scratch tickets.
So we launched this thing at the end of 2007.
And by the beginning of 2008, We'd accidentally invented free-to-play social games because we were the first studio to implement microtransactions with Spare Change.
We were the first studio to implement an offer wall with Super Rewards, which is Jason Bailey's former company, which is why I got involved with him later on, Felix.
And then
in 2010,
I sold Backstage to Real Networks, to Game House, and left the industry.
Not
right?
Yeah.
Yeah, I came back in because, you know, it turns out it's as much as I like to pretend it wasn't my passion, it was something that I was good at.
And so, in 2018, I came back to the industry to work for Jason Bailey, who is now the CEO of Eastside Games, and that's where I met Felix.
So, Felix did he managed our ad monetization for the entire studio for over a year, I believe.
So,
and I thought, you know, he was really good, and he was the reason why
you were
before that doing Tetras.
Yeah,
so uh i was at e si games and like analytics for five years and then i retired and i published a book called game analytics retention and monetization in free-to-play mobile games that was my gift back to the industry it was basically everything i'd learned over the previous five years because one of the things about analytics is it's really hard to find out information about how to do it for mobile gaming And so I put down 300 pages of what I thought was stuff that you needed to know if you were going to work.
300 pages,
that's a lot.
That's a good idea.
So, yes, it is a lot.
So, what I did in 2025, this earlier this year, I launched a website, artdow.com.
The whole point of Professor ArtDAO, as it's called, is that you don't need to read my book.
So, instead, you just go to the website and everything you need to, all the tools are there for free.
So,
before I begin, so the way that I've structured this presentation is just a bunch of set of questions that I think people are afraid to ask, but why not?
So the first one is: what is analytics and is it the same or different than data science?
And does anybody really care?
Do you three care?
I'm not sure.
Not really.
Yeah, not really.
Data scientists get paid more is
the short answer.
That's a great answer.
That's the best answer.
I'm not a data scientist, never pretended to be, but I get asked this question periodically.
And the best answer I've come up with, I believe, is this, that if you're a game analyst and you're working in a game studio, you've got tons of data about your players, and what you want to do is be able to make predictions about what your new players are going to do.
And you work with cohorts.
And we're going to talk a lot about cohorts today, because it's really the foundational piece that you work with as a game data analyst.
My focus is on analytics.
My book is called Game Analytics.
It's not called Game Data Science.
If you're a data scientist, what you do is you use typically machine learning, trained on historical player data.
So, what you want to do is try and model and predict individual player behavior, ultimately with the goal of tailoring gameplay, offers, whatever it might be, predict churn
on the individual user level.
Okay, so in either case, you're doing predictions, right?
I mean,
analytics, though, uses statistics statistics and in particular regression as its main predictive tool.
Okay.
Clear?
Does that make sense?
Yeah.
Good explanation.
So if we're working with cohorts, what is a cohort?
But simply, it's just a set of players.
So we're dealing with player behavior in the aggregate, right?
We want to take a set of players that say
they've come from the same source, right?
So this is really important when you're doing user acquisition.
So in particular, attributing users to a particular ad campaign.
And you can go as deep as you want to this particular creative.
So we tended to work at ESI games just on the ad campaign level.
So this ad network running, targeting these particular users.
So a cohort suggests a set of players.
They have something in common.
The most common thing is that you look at the players who who all installed the game on the same date or had their first open on the same date.
Other ways we tend to break them down is by platform
and by country, example.
So, um,
here's here's a definition of a particular cohort: we want to look at all the organic Slovakian Android users who installed our game in June, or maybe we look at the release, and then there is also one attribute missing installed on an emulator because he's a cheater.
That is, yeah, Jakub only plays on emulator.
Yeah, so we know we filter them out because they're the one making all of the $99 purchases, one right after the other.
So we can
crack the APK.
So why are cohorts important?
I've alluded to it.
It's because it's how we can monitor how well a paid user acquisition campaign is going.
And you really want to keep track of your new cohorts, right?
Cohort behavior evolves time.
And you want to understand where your good users are, what cohorts they belong to.
A typical mistake some big games have made in the past is that they'll launch and they'll have this massive success in their first month and they keep buying users based on that particular behavior of the golden cohort, but their new cohorts are not behaving the same way.
So you can light a pile of cash and find
a lot.
So we're clear on what a cohort is.
Yeah, we see the golden cohort behavior too often, I guess, still.
Yeah.
Yeah.
It's intoxicating, right?
Because it's like, yeah, this game's amazing.
It's super good.
Yeah.
Yeah.
Yeah.
Okay.
So in the spirit of you guys' podcast, I made some 4 a.m.
doodles at a hotel bar.
I was
trying to explain what cohorts, what their behavior looks like over time, right?
So the thing about cohorts, when you're looking at how they behave, you're looking at typically charts where the x-axis is an integer and it's the days since install.
So the origin on the x-axis is day zero.
It's when the cohort starts playing.
The cohort people don't have to have the same install date like when I was talking before about all the Slovakian players in June.
Because what we do is we just shift and looked at their behavior from their days since they started playing the game.
You know, what do they do on the first day?
If they come back, what do they do on day one, et cetera, day two, three, seven?
So some of these graphs that I've looked at is a way of looking at, say, Dow
top left as a cohort metric, as opposed to, say, just a time series when you're looking at how's your game's daily active users doing.
Dow as a cohort metric leads directly into how we calculate retention, which is the primary topic of this presentation.
So, the things we want to know or assume
These are assumptions we're going to make in a lot of the predictions that we make on cohorts, is that cohorts kind of behave in this way, particularly once they get to be a certain size.
Then, you know, small cohorts, you're basically looking at individuals, they dominate what happens.
Larger cohorts, 100,000, 10,000 users, they tend to get this smooth behavior of how they play the game or leave the game churn,
how their LTV grows asymptotically.
And how, and this is the surprising one that I get a lot of people are surprised, is that ARP DAO remains pretty constant for cohorts regardless of how old they are.
So on day one, you might have an ARPDAO of 50 cents.
Day 360, you're still going to have an ARPDAO of 50 cents for a cohort.
And that's because the daily conversion rate tends to stay the same for the cohort.
So even as users drop off, the ones that remain, they continue to spend money and that purchase isn't relatively same.
What's the ideal minimum size of a cohort?
I like round numbers like everyone does.
So I tend to say 100.
Like if you've got less than 100, so when we're looking at, say, cohorts based on a geo by country, we'll typically filter out ones where we don't have at least 100 installs.
Yeah.
And then what's the minimum cohort size on a daily basis, let's say, when you're having a soft launch game and you want to measure retention?
I would typically not look at it on a daily basis, but look at, like, say, group them.
Because again, it's cohort metric.
You don't, they don't all have to come on the same day, right?
So I would wait until we've gathered enough.
And you can do the power calculation on your statistics and all that, but
intuitively, we know fewer users, the less reliable the estimates.
Just want to mention for people who don't do statistics, daily power calculation means the statistical significance, which means how many users we actually need for the result to be statistically significant.
So yeah,
what would that number be in terms of the,
let's say, day one retention?
If I want to see
retention
proportion is
slightly differently there.
In this case, because the numbers are big, retention
that you don't need that many.
100 would be fine.
But if you're looking at conversion, when you only have like three or four out of your hundred that are making a purchase, that's when it gets tricky.
And that's where the power calculation is going to reveal that you need like a thousand users because the events of interest happen so infrequently yeah one one note regarding the great charts over there
the red curve which is growing the ltv1
99 of games in the market don't see this they see curve that grows for five days and then gets flat forever and that's that that's how the market works just if you if you see if you see a curve if you see this you see a gold mine yeah you mortgage your house you mortgage your house and you spend it on ux that's like day 18 that's day 18 there.
Is day 18?
That's not, yeah, yeah, no, but
Felix knows this from our time at Eastside Games, is that
our narrative idol games, which had really strong live ops,
had
curves like this all the way out to 9180 days.
Because the retention was good.
That's what's really driving this.
And that's, again, the kind of the theme of today's talk.
Still the highest rewarded IMPTAW I've ever seen in any game.
Okay, good to know.
So, what are we going to learn today?
So, I'm going to do a little introduction to my website and Professor Arpdow's free analytics tools that are going to help you model your game's retention.
And once you have a model of your retention, you're going to be able to predict things like your player lifetime value, predictive LTV,
and your expected days to break even for ad campaigns.
You have that click calculation thing there
on the website?
Yeah.
The one where we input, oh, yeah, I really, really need to do this because
every time I talk to people, and I just tell them, like, guys, have you modeled your CPI?
Like, how much money you need to throw at this?
Like, oh, think it will work out, or it won't work out.
Well, I mean, no, I haven't modeled how.
If you're looking for a model of how your CPI changes with spend.
Hello, everybody.
Thanks for joining this episode.
Our sponsor, Fast Spring, has delivered D to C at scale for over twenty years.
They power top mobile publishers around the world and are reliably fast growing, profitable and very financially secure.
So FastSpring isn't just figuring out D to C on the heels of the recent VC investment.
Their developer first platform and deep gaming and payment expertise have helped democratize global commerce for over two decades.
As no drama D2C partner.
With real-world experience at scale, Fast Spring is a partner that you can trust with your players.
Quickly launch a new web store, replace an existing D2C vendor, or just add redundant D2C vendor at fastspring.gg.
Thank you very much, guys, for sponsoring this episode.
Now let's get back to the content.
Bye-bye.
That one's trucking.
Not even that one.
Just the basic one, the linear one.
Okay.
Even that one.
No,
you can check it out.
And we're not going to get there today, but because it's kind of the more advanced one of the free tools.
But yeah.
What I really want to get across today is why your game's retention curve is fundamentally important.
What it means to your game's success and
how it it really should be your focus.
And when you building a game, managing a game, you're the product owner, whatever, really focus on retention because monetization can follow.
Yep.
So that's definitely true.
Which goes back to the
question then.
I mean, if you were expecting me to come in and talk about all the game studio metrics and what the key performance indicators are, that's not what I'm going to talk about today because I think, you know, your audience, if they're listening to you guys, they're well beyond that.
So really what I want to talk about is
what are the most important metrics and how you should be modeling them.
Okay, so I'm going to start with a quote from Jason Bailey who sat beside me for five years and he used to always say the games teams, because they would look at like top line revenue and they would look at their daily active users and they would come to the growth team and complain that their DAO was falling.
And Jason would tell them to go away and say you need to worry about two things and two things only.
What's your retention and what's your average revenue per daily active user?
And everything else will follow.
Why is that?
Well,
obviously, it's under their control.
So if you're managing a game as a product, what can you change?
You can change the things that are going to bring your users back.
You can change how your live ops is working.
That will improve your retention.
And then how you monetize the users you have, that's under your control.
Once you have those two things trending in the right direction or at a baseline level, then the growth team can go ahead and profitably acquire users for you.
Right.
And that's profitable acquisition of users for any free-to-play game is only possible if the lifetime value is greater than the cost per install.
That's the basic hypothesis.
LTV formula.
So it turns out, though, that lifetime value is just the sum of your retention times ARPDAO.
So, well, there you go.
Right there in one formula explains why the games team should only worry about retention and ARPDAO because ultimately that's what's driving LTV.
And if LTV is high enough, we can grow this game relative to your cost per install.
Yeah, that's the fact I wanted to add here that in current market, the CPI thing is kind of a little bit more important than the rest of economic, unfortunately.
No,
everybody's fixating on CPI, right?
And you gotta look, it's LTV is just as important.
It's like, you know,
but people focus on CPI because it's something they think they can game,
I don't know, we just need the right creative.
We just need to lie to our users about what the game's like, and then they'll come into the game, you know, all the techniques there.
Yeah, definitely.
And it becomes, I don't want to get into why our industry
is troubled by the skyrocketing costs per install.
What I just want to talk about is you can build a game that you can grow.
if your LTV is greater than whatever your cost per install is and however you want to go about fixing CPI.
You guys can talk about that with someone else who knows anything about that because I don't.
But what I do know about is that your LTV is your sum of your retention times.
I want to point that out, and you guys are going to get kind of tired of hearing me say sum of retention in the same way that my wife is tired of hearing me say ARPDAO.
It's okay.
It's okay.
I got back at her by incorporating ARPDAO industries.
Okay, so quickly, what's ARCDAO?
It's just revenue divided by DAO, right?
So here's a chart, a simple one that
an analyst at a game studio might put together.
It's a time series where a calendar date is along the x-axis, and then we can plot how are we doing?
What's our revenue?
What's our DAO?
ArcDAO follows nationally.
So these are good at revealing what your long-term trends are, but
again, the CEO at ESI Games used used to say that
your DAO and revenue are vanity metrics.
You go and focus on what you can control, and the growth team will do the rest.
So, and they're vanity, but they're important.
We have a business way of them.
So,
we talked about DAO.
Now, how do we define retention?
It's one of those metrics that people new to the industry might have a little bit of difficulty get wrapping their head around.
It's just this: we look at
we have a cohort of users, how many of them come back to play the next day?
What proportion of them?
And that's your D1 retention.
How many of them come back to play
on the seventh day?
We don't care about the sixth or the eighth.
Do they play exactly on the seventh day?
That's your D7 retention.
And so.
And what do you say are good retention metrics for D1, D7, and D30
in mobile?
40 and 20.
If you can hit that, you're doing fine.
So East Games just put out a
ball
game, and D1 retention is like 53%.
That's awesome.
D1 retention will anchor.
D7 is going to define the shape of what happens after that.
That's what we're going to see.
Do you have also a benchmark for D90?
Well,
you'll see what's in the next in the next slide.
I'll talk about what ultimately these numbers, how they should relate to each other.
So D90 should be half of your D30, which should be half of your D70.
So
spoiler alert.
So,
yeah, let's go and just jump ahead and look at the next slide here.
So here's, you know, graphs you've probably seen.
And this is not a time series.
Well, it kind of is, right?
The x-axis, though, is the install date of the players.
So we're looking at each day we got a bunch of players and what's their D1 retention.
And you can't, you have to wait till tomorrow to figure that out.
And then, what's their D7 retention?
Well, you have to wait seven days, eight days actually, before you can calculate it.
And then here's D30, the poor purple one, way back there a month ago.
We have no idea.
We call these cohorts are not fully bait.
But it doesn't mean we have to wait.
We don't have to wait 30 days.
We predict what for each of these days or install cohorts, we can easily predict what the D30 retention is going to be.
And let me show you how.
But first, we have to think about, you know, why do we just measure retention on D1, D7, D14, whatever, you know, studios pick their favorite days.
But the numbers,
this
standardization around D1, D7, and D30 retention and so on, I think really comes from Eric Suffert's book, Freemium Economics, because he talks about this recurring pattern that he's seen with free-to-play games, which is your D7 retention is typically your D1 retention.
Which is a wishful thinking.
I mean, yeah, it's an idealized game.
You're right.
It is wishful thinking.
But again, if you've got a compelling end game, lots of content, a live ops is pushing out
events every weekend, right?
These things are important.
If you're just throwing out a hyper-casual game with some ads that you bombard the user with in the first session.
Love it.
Love Love it.
Yeah.
You're not going to see this retention curve, right?
And you're in a very different business then.
I'm not sure what it is, but you know, this is the business that ESA Games has been in and my own studio before that.
Long tail retention is crucial and live ops is crucial.
So this is what Eric said.
He pointed this out and he said, isn't that interesting?
But what he failed to do is to actually generalize the concept.
And what he was describing is just a power function.
We just needed to do a little bit of math.
And that realized that retention follows this formula where you have two coefficients, A and B.
And B is the exponent, and A is the coefficient.
A is typically the value of D1 retention,
and B is negative.
And so, which means you're just dividing by the number of days
typically.
So I'll show you in a minute
what these formulas
derive in terms of
its shape.
Because really, we can, we don't need to just measure retention for a set of industry standard dates.
We can instead derive a retention curve and we can measure them for any day we like.
And once we have a curve, which we can usually build from like the first seven days of data, we can predict what our retention is going to be out to day 360
or whatever, right?
With varying degrees of accuracy.
So just quickly,
I said, you know, it's a way of defining what a retention profile is for a game as a whole.
You know, if you're looking at your D1, D7 retention as every day, and you're looking at a bounce up and down,
you can get distracted by really what you want to know is what's my game's D7 retention.
And you can build a weighted average of these retention values through cohorts, like just taking them and smashing them all together.
And then you can derive a retention curve, which is just a formula, a function.
And it's a compact way that you can represent a game's retention.
And if you have a formula, so the example here I give, say your retention by day n is defined by the formula 0.4n times n to the negative 0.5.
That's actually a very typical kind of retention curve that you'll see in
games.
You can use that to predict what your retention is for any day.
So what's day 53 retention?
Well, if we plug 53 in for n,
we will see that it's 5.5%.
I picked that as because 53 is a funny number, and no one would ever think.
Oh, what's your day 53 retention?
So, what's the retention curve look like?
So, all it is is we're just going to use statistical regression to fit a curve to our observed retention values.
This graph,
I'm connecting a value for D0 retention, which is always one, as you play on the day you install, because the install is the first one.
Day one retention of 0.4, day three retention of 0.23, 0.3%, and the day seven retention,
I think I put in 18 or something, 16.
You end up then with a formula that fits this data.
And in this case, it's that retention by day n is equal to 0.396 times n to the power of negative 0.472.
And so as you move those dots around, the shape of this curve will change.
It never goes to zero.
It asymptotically goes on forever, which is, of course, not realistic, but you can easily fix that by just forcing it to be zero beyond someday when you know your game never has players.
But again, going back to you know
tires of Eastside games, and Trigger Play Voice came out in 2018, 17, the year before I joined.
It still has players from the original golden cohort, right?
So
they're on to whatever, you know,
day thousands, and they're still retention, we're still retaining users.
Love idle games.
Actually, day 3,000 plus.
Yeah, exactly.
So, again, so to take it back to those
scribbles I did on the back of the app at the 4 a.m.
in the hotel bar, the retention curve is just cohort DAO normalized.
Normalized in the sense that we know how many installs were part of the cohort.
If we just take every day's DAO and divide by that number, you'll get a proportion.
And you're basically calculating retention
in your retention curve.
Basically, how many users left?
How many are left?
You know, like what
the thing about retention is it's got a funny statistical meaning, which is like if you look at, say, say your day nine retention
is
5%.
So what that means is that if you have a cohort and you pick a user out of the cohort at random, there's a 5% chance that they're going to play on day 90 or did play if you're looking at historical data.
So it's the probability that a user will play
on this day after they install.
So if your day seven retention is 20%,
there's a one in five chance that a new install is going to play on day seven.
Given that, we can do some pretty cool things with it.
So, well, first of all, let's take a break and one of you or all of you go over to arpdow.com slash retention.
And you can play with
a tool that I built so that you can come up with a model for your game's retention curve.
And all you need to do is plug in some observations for what your early retention signals are.
you know
my website will fit a curve to that data and show it to you so for people looking at the slides you go to the retention curve creator you plug in some observer observations that you have from any cohort you like
so say we have 40.2 d1 retention and 16.7 day seven retention I'm going to leave the D3 and D30 at zero, meaning we're going to exclude those.
You need to just provide two.
And obviously, of course, the more data points, the more accurate the model is going to be.
You hit submit.
I tell you what your retention curve is.
I give you the formula for it.
And I'll give you a little picture of what it looks like.
Nice freebie.
Really, this is not hard.
Like how I'm doing this is explained in detail in my book.
Rather than, like I said, you don't have to go read my book.
Just go play with this tool.
It's simple statistical regression.
So, yeah, I mean, this looks pretty sweet, eh?
Day 365 retention of 2.81%.
We'll take that.
How realistic is that?
Sure.
Sure.
The thing that's
alluded to this as Professor Arcto talking on the corner of one of the slides back there is that
retention and Arctow are often in tension with each other.
What's a great way to make a user turn right away is to show them an infrastructure ad every three seconds once they install a game, right?
Great for monetization, so your ARCDAL would be high, but your retention is
going to take a hit.
The gold mine is building a game with a day 365 retention of 2.81%,
but still has
monetization.
Nothing to blame, like a dollar or something.
Yeah, the dollar art balance, D365 retention of 3%, you'd be okay.
Why is retention the most important metric?
Felix, do you agree retention is the most important metric, or do you believe ad monetization, ad
revenue is the most important?
I have a call at 9 a.m.
about exactly this with a client, but I do agree that retention is exactly the most important.
With this particular client, we've actually managed to double the ads, and the day one retention is exactly the same.
Amazing.
Very slow base, right?
Yeah, but still.
Well, that's good.
All right.
So
why is retention the most important metric?
And it's pretty obvious, but you can only monetize and install on the days they play.
If they don't play your game, you can't make money off them.
You can't buy your dinner with retention, though.
No, no, and you can't, yeah, I make that point very clear.
You can't buy dinner with five-star reviews either.
Yeah, right, exactly.
Yeah, yeah.
You need to,
but what you need to be able to do is,
and here's the analogy I use.
So I have this recurring cafe in my book to kind of ground these ideas in bricks and mortar businesses.
So imagine you own a cafe and you have a customer cross the threshold.
Well you're there they're going to be you're going to monetize them.
They're in your store.
They're coming in to buy a coffee, croissant, whatever.
What you really are trying to sell them though is a return visit.
Because you're going to monetize them today, but you want to monetize them tomorrow as well.
So it's that whole tension between, you know, provide them good service, provide them a compelling reason to want to come back
and to make money off them today as well.
So, but within mobile gaming or free-to-play gaming, what we want to know is how many days is an install going to play?
Like, how many chances am I going to get to monetize them before they churn?
And it turns out that the expected number of unique calendar days that one of your new users is going to play your game is simply the sum of your game's retention curve.
It blew my mind when I realized this.
After playing with analytics data and leading analytics at East Games for years,
this
really obvious, it's obvious in hindsight.
But when you are working in the industry without any mentors or guidance or whatever,
this might be something that
takes you a long time to discover or you may never discover it.
So I put it in my book.
I'm like, here, this is something that's pretty cool.
You fit a curve to your observed retention, that's your retention curve.
Then, if you take the sum of that,
that's the number of unique calendar days, and install is going to play your game.
Why is that important?
Well, as we just said, you can monetize them when they play.
So, expected player days is what I call that.
The acronym is PB for player day.
It's the sum of the retention curves.
So, as part of the output on the Professor Arcado site that I include with the retention curve, is not only just predictions for what your day and retention is going to be, but how many days will a player be expected to play, unique calendar days that they'll play the game by that time frame.
So, you can see here for this slide that by day 180, given this particular retention curve, we expect that users will have played the game 13 different days.
You have 13 cracks at monetizing them.
It's very much an average
in the sense of the average is the fulcrum of which you balance a histogram.
So if you were to look at actually how many, if you were to look at a histogram of how people play your game, you're going to see that most installs play only one day in turn.
They install, they kick the tires, they go, this game's not for me, and they're gone.
And then out at the other end, you've got a bunch of players who love your game and they play every day
and you've got this massive valley in between right so players either play once and quit or they play all the time
why is it important to know the expected player days
what do we derive from it yeah how can you use it
you're gonna derive ltv from it
Because if you if if so going back to this example here by day 180 we know that that player
will have played 13 times.
So
we know what our RPAO is.
You multiply the two, and that's your LTV 180.
Okay.
There's nothing magic about that.
It's just the algebra.
I devote an entire chapter of my book to showing the equivalence between monetization,
monetization curves, and the way that we typically look at.
data regarding revolution.
Sorry, I'm trying to
wrap my head around it a little bit because usually when I I don't balance economies and stuff like that I balance it mostly on like session time session length which gives me daily play time on calendar day and from that point on I know like how much actual you know battles the guy does or whatever therefore I I can estimate like by this point in I don't know two hundred days he had the roll gotcha this many times which means this has this percentage of collection yada yada yada.
But what's the translation of this number regarding that?
So this has nothing to do with actual playtime, does it?
No, it just has to do with they fared up the game.
And yeah,
it's a bit of a more, it's a blunt object, but it's one that user acquisition managers
would use as opposed to economy game design.
Like you're
looking at the data through a kind of a different lens in terms of how engaged is this player, how
and engagement will impact MarkDAO for everyone?
But you can't engage them unless they open the app, right?
They have to come back and
play it.
But really, what you're starting to look at there, Yakub, is I think the,
to my point at the beginning, looking at individual user behavior is very different than looking at cohort behavior.
Yeah, so you're using a microscope and I'm using like a magnifying glass.
Okay.
Yeah, kind of the difference.
So
let's go to the next.
To average it out by whole cohort, that's the point.
Yes.
Okay.
Right.
And LTV is an is a cohort metric.
It's the average, right?
It's yeah.
So I can do because I'm modeling it for every single user.
Right.
So
this player days thing is so here's the the running sum of retention plotted as a curve.
And it looks like an LTV curve
has the same shape as the LTV curve, because the players that play are the ones that you monetize.
Ad archow is relatively constant.
I feel like for an ad-driven game, your ad archow is pretty constant, right?
It doesn't really change.
Like if a player's playing, if it's the 180th day they're playing your game, they're still going to generate roughly the same amount of ad archive as they did on the first day they're going.
Yeah, the only difference is how close it is to Christmas, right?
Or Black Friday.
That's right.
That's right.
So the professor on the right there is chiming in to say,
just sort of reiterating what I said on the last slide.
Since you can only monetize a user on the day they play, player days by N is how we predict LTV.
It's also how we predict DAO and revenue.
So all of the predictive tools that are freely available on ARPDAO.com use the monetization curve as
the Swiss Army knife of doing your predictions as to where you're going to get.
I actually give you the formula for the player days curve, which again, these are all stated in an easy way for you to just plug them into Excel and you can play with them yourself if that's what you want.
So, yeah, to you know, wait, we can predict LTV with the retention curve.
Like, it's kind of surprising, but if again, if you look at it, it's just a way of defining what LTV M is.
A lifetime value is the average total revenue derived by a user during their first 10 days after they install your game.
Well, algebraically, it's just the number of days they play times ARPDAO.
So
it falls out of this notion that PDN is the expected number of distinct calendar days they're going to play the game.
And ARPDAO is just how much money an active player generates each day that they play the game.
So if you want to improve your LTV, which is necessary to grow, because you want an LTV to be greater than cost per per install, you need to either increase retention or you need to increase ARPDAO.
And you need to do one without affecting the other.
Ideally, you improve both.
Easier said than done.
Yeah.
Or you don't.
So
the second free tool on the Professor ARPDAO website is this LTV predictor.
And so you can go and play with that while I'm
talking about it here.
And it works very similar to the retention curve creator.
All you do now, though, is you also add in your ARP DAO,
right?
So if you have an estimate of your retention curve and you have an estimate of what your ARP DAO is, and LTB is just the two multiplied together.
And so I give you like the LTV curve so you can get estimates of where you're going to be.
What's your LTB 90 going to be?
In this case, it's $4.68 and so on.
And so you can see that
if you increase your ARP value, you're going to increase revenue.
If you increase retention,
depending on how you increase retention, retention curves are weird though, right?
Like if your D1 retention was 80%, but your D7 was only 16%,
what would that mean?
That means people are dropping off like crazy, right?
And then
you would expect that that trend would continue and you'd have no users by day 90.
Yeah, but then you need to show millions of interstitials day one and then you're fine.
You do.
Much of your program game economy.
Love it.
Yeah, exactly.
But that's that's the right answer.
If you've got 80% D1, yes.
And
you can maximize that day.
Exactly.
Yeah.
And you know, well, we've only got them for a week.
Carpetium.
Carpet DM.
Is there a
CPI input somewhere?
Because on the side I see the seven ROS target to break even by day.
Well,
here's the cool thing
the seven raws targets don't have anything to do with CPI or LTV
like they're based on your ratio of what your
your LTV7 is to your cost per install and that can be like a five dollar LTV seven and a you know fifty dollar cost per install or it can be you know five cents on a 50 cent CPI it's it's about the ratio of those two and the monetization to that point is going to follow the shape of the retention curve
So
the two are divorced.
So here, yeah, let's go to the next slide to see what you're talking about here.
On that LTV prediction tool, I include what your LTV by day N is predicted to be based on your current retention on ARPDAL.
And then another column at the end is what your D7 Rho S target should be if you want to break even by day N.
So, for example,
N for 90.
So, day 90,
we know by then each of our installs has played an average of nine calendar days.
Their LTVN by day 90, LTV 90 is $4.68.
28.03%
is the D7 ROAS target.
So if we have an ad campaign that by day seven, we're seeing a ROAS of 28% or greater, we know it's likely to break even by day 90.
And that has, and it doesn't matter what the CPI is.
But the curve needs to go up.
That's the assumption.
The curve's going to follow the same shape.
The retention curve and the monetized, if your RPAO is relatively constant, your LTV curve is going to follow this trajectory of the retention curve, the sum of the retention curve.
And ROAS here, not to be confused for our avid listeners, is not the ROAS hotel in Bodrim, Turkey, but return on ad spend.
Exactly.
I assume
your listeners are very intelligent and they know what a lot of these things are.
My apologies if I'm assuming things are understood.
Thank you for pointing out.
No, it's just a
prominent game CEO in our industry who, after
selling his gaming company, he named the hotel ROAS.
He bought a hotel in the game.
Okay, perfect.
I like it.
That's an actual hotel.
I was hoping ROAS meant something in Turkish.
Maybe it does.
One of your listeners can
let us know.
So let's see what's on the next slide here.
Oh, yeah, yeah.
So this is one of the things that I was interested in.
So I have included it as a sidebar on that LTV
tool.
It's basically what happens if you miss your D7 ROAS targets?
What's its impact on your days to break even?
Because a lot of times
you're starting a ROAS campaign and your
rep from the
network is like saying, oh, yeah, you're really, you're close because you're saying, well, we need a D7 ROAS of 28% to break even by, or 28% to break even by day 90.
Well, our ROAS is only 20%
on the campaigns.
And they're like, oh, that's close enough.
Well, is it?
Our days to break even just went from 90 to 178.
If your day seven ROAS is only 15% in this particular example,
suddenly you got to wait 316 days.
And if it's 10%, you're never going to see it.
Sounds more realistic.
And you get a lot of pressure from account reps to just let the campaign run.
It's learning.
It's still learning.
And you need to spend this amount of money
to get enough signals for it to learn.
And, you know, meanwhile, you're...
And then even then, after like two weeks of spending and spending, you're still, you're close, but you're not, you're not at your ROAS target.
So you just need to be aware that shit gets real here.
Like you're, you're not going to see a return on that ad spend for, you know, it doesn't change linearly.
It goes up exponentially.
So if you miss,
you're going to have to wait a long time.
And even then, if you're lucky, right?
Like again, if your retention curve does have a terminal day, you know, so people just never play your game after, say, 300 days, and you're looking at a 316-day
payback window, you're probably never going to see the money back, right?
At least not all of the money.
So, don't let your account reps bully you.
Give it just one more week.
Yeah.
But we should end on the positive note for sure.
So, actually, we got through it.
We're at the point, it's the end of the lecture.
I have homework for all of you, which is to go to ARPDAO.com and play with
the tools there and do some experimenting with looking at what's better for your game.
Is it a 10% lift in your D7 retention or is it a 10% lift in your ARPDA?
You know, because
I see a lot of product managers really kind of fixate on squeezing more and more money out of their users, battle passes,
more ad placements, because they want to increase that ARP down.
And yeah, that's important, but
would you be better off finding a way to increase your retention by a similar lift?
Battle passes actually can decrease your ARPA.
Yeah.
Yeah, yeah.
I agree.
I agree.
And, you know, or pay to remove ads.
If you do that wrong,
you gotta hurt yourself.
Oh, though, you may increase your retention.
So it may
not be a negative.
So that's your homework.
I expect
you guys to get back to me to that with an answer.
And here, last slide is, so if that's wet your interest at all in
analytics for the games industry, where can I go to learn more?
Obviously, you can go to arcdow.com, come for the free analytics tool, stay for my geographical pricing tool, which maybe I'll come back one day and we can talk about that.
Probably not so much in you guys' wheelhouse.
You can buy my book.
My book's great.
The Bible, there you go.
I have a copy downstairs.
Yeah,
did I buy it for you, Felix?
No, no, I bought it for myself.
Thank you very much.
I owe you a beer.
Yes.
So, which is about how much money I make on each copy.
Jeff Bezos gets
most of the revenue on this baby.
You can buy my book, How to Find It.
Just Google Oven's Game Analytics, or you can visit the book's website, which is gameanalytics.ca.
I'm Canadian, not.com, which is
the good analytics company.
Or you can follow me on LinkedIn, where I post about stuff like this as things occur to me and I learn more and more, I'd like to share my learnings with the community at large.
Yeah.
Thanks.
We'll put all the links in the show notes.
And if you have any questions or comments, just put them under the video, join our Slack channel, or just be a ping or Russell on LinkedIn or on the DM me directly.
Yeah.
Yeah, I love it.
Yeah.
I got nothing better to do at this point.
Yeah, pretty much
thank you very much for for sharing all the info and the graphs and everything and that though both of those tools i will i will definitely play around yeah
some new metrics today yeah yeah exactly
yeah
well you know if you guys can't find any value in in what i've done then
i need to just shut it down and go
quietly into the
into the subject
yeah exactly yeah yeah yeah.
But I'm hoping that you know, and hey, if you guys have any feedback about you know, it would be nice is if you did this or that, but the um, I think
you know, you want to you want to probably look at that days to break even calculator because I think that's um, yeah, that's very interesting.
It's very interesting.
I've seen a lot of LTV calculators, but I don't think so.
I've seen the one with this edition.
Yeah,
very cool, very cool.
There you go.
Thank you for listening, and yeah, see you next time.
See you next time.
Cheers.
Ciao.