Author: John Koetsier

John Koetsier is a journalist and analyst. He's a senior contributor at Forbes and hosts our Growth Masterminds podcast as well as the TechFirst podcast. At Singular, he serves as VP, Insights.

ARPU, ARPPU & ROI: driving smarter mobile growth via mobile analytics

Are you a mobile marketer trying to get more ROI from your mobile app business? You already know that first-party data, measurement, and smart analytics are critical. And while ROI remains the gold standard for determining profitability, average revenue per user (ARPU) and average revenue per paying user (ARPPU) are key building blocks for engineering smarter user acquisition and re-engagement campaigns.

Today, in a post-ATT world where ad networks, targeting, and tracking have changed dramatically, ARPU and ARPPU are even more important.

They give you directional signals about user quality and help you set CPI targets that make sense.

ROI still rules

Let’s be clear: ROI trumps everything.

Neither ARPU nor ARPPU tells you anything about whether your campaigns are profitable. Only ROI reveals whether you’re turning marketing spend into enough return to make it worthwhile. But ARPU and ARPPU matter because they help you make smart, forward-looking budget decisions when testing new channels and campaigns.

At a high level:

  • High ARPU is great, but only if your CPI is low enough to make ROI positive

  • High ARPPU is wonderful too, but similarly, only if your acquisition costs align

  • ROI is the final arbiter of success

ARPU in mobile analytics

Definition:

ARPU = total revenue / total installs (over a given time period).

ARPU is still one of the most useful high-level business health metrics. It quickly tells you if you’re monetizing users well enough to justify acquisition costs.

Important things to remember:

  • Granularity matters: You no longer just calculate ARPU globally; you can calculate it by cohort, geo, device type, or more

  • AI-powered optimization: Modern mobile analytics platforms like Singular help you make ARPU predictions in real time to boost or throttle spend across networks, often by using the power of Claude or ChatGPT

  • Beyond gaming: Retail, fintech, and subscription services increasingly use ARPU benchmarks to forecast payback windows and lifetime value

Example: A streaming app may see higher ARPU from iOS users in Tier 1 Western countries, while Android ARPU may outperform in emerging markets.

Knowing this enables smarter campaign allocation and optimization.

ARPPU in mobile analytics

Definition:

ARPPU = total revenue/number of paying users.

ARPPU shines in businesses where only a minority of users pay, like games, freemium apps, or creator platforms. It isolates the spending power of payers and shows how well you’re monetizing your revenue base.

Important things to remember:

  • Hybrid monetization: Many apps combine subscriptions, ads, and IAPs, so ARPPU helps segment how much real spenders are worth compared to ad-only users

  • Whale detection: AI models can help predict future ARPPU of segments or even individuals, guiding personalized re-engagement campaigns

  • Privacy-era analytics: Because user-level attribution is limited, ARPPU is often modeled at a cohort level, but that still provides clearer signals than ARPU when testing monetization changes

Example: A mobile RPG game might have 1,000,000 installs. Out of those, however, only 50,000 users (5%) make at least one purchase. So total revenue is $2,500,000: ARPU is $2.50 but ARPPU is $50.00. If you only look at ARPU, you see a very different reality than if you look at ARPPU.

Knowing this helps you make the right decisions with your app or game and its marketing priorities.

Using ARPU and ARPPU for smarter ad network decisions

Calculating both ARPU and ARPPU helps you make smarter decisions, but always in context of ROI.

  • If Network A delivers an ARPU of $7 with a $5 CPI, and Network C delivers an ARPU of $3 with a $4 CPI, Network A is far more profitable, even though CPI looks higher

  • ARPU/ARPPU analysis ensures you’re not fooled by low-cost traffic that doesn’t convert into meaningful revenue

5 years ago, you might have compared ARPU across networks manually. Today, you can do it automatically … but the logic is the same.

Pro tip: Always consider incremental ROI. Many marketers use incrementality testing alongside their ARPU/ARPPU calculations to validate which channels are truly adding value.

Real-world examples: ARPU vs ARPPU

  • Gaming: A mid-core RPG runs a creative A/B test. ARPU looks flat across cohorts, but ARPPU shows one variant drove a 15% uplift in paying user value. Without ARPPU, they’d have missed it.

  • Retail: A fashion shopping app compares ad network campaigns. Ad network A’s CPI is 20% higher than ad network B’s, but ARPU is 40% higher, making A the better long-term investment

  • Fintech: A neobank sees that early cohorts from influencer campaigns have lower ARPU but higher ARPPU, signaling a smaller but much more valuable customer base

The bottom line

ARPU and ARPPU aren’t perfect, and they don’t replace ROI. But in today’s fast-moving, privacy-restricted, AI-assisted ad ecosystem, they’re essential directional metrics.

Use them to:

  • Set realistic CPI targets

  • Identify higher-value networks and partners

  • Spot opportunities to improve monetization flows

And most importantly: always tie them back to ROI.

With platforms like Singular providing ARPU, ARPPU, and ROI across channels, you can make the smart investment decisions that grow your app business … even in our current complex landscape.

Beyond the MMP: Growing even faster with more data and richer context

I bet you didn’t have a “beyond the MMP” blog post from an MMP on your bingo card today. And yet, here we are … 

The reality is that there’s an emerging superpower in mobile marketing, and it’s based on going beyond the MPP … or at least beyond the traditional definition of what an MMP does. The superpower is data, which sounds banal. Because, of course everyone wants all the data they can get.

But the superpower is new kinds of data: not necessarily the data user acquisition experts have typically gathered. 

You already have tons of data in your Singular dashboard, and it’s all wonderful and super-useful. But there’s times when you could really use more context. Context on the why: why did this work … and why that didn’t, for instance.

Data to help you build that beyond-the-MMP context can help you grow even faster.

I chatted about that recently with SciPlay’s director of ad product, Gal Karniel. And I just did a deep dive on the same topic with Maayan Schoor, a Singular product manager, on the same topic: getting more data that provides more context around why your performance marketing campaigns are getting the results that they’re getting.

Hit play and keep scrolling …

The key insight?

MMP data is the foundation. If you want to accelerate your growth even more, there’s more you can get.

Beyond the MMP: building on a strong foundation

Obviously, Singular’s MMP functionality gives you what you need to measure campaign performance: installs, costs, ROI, retention, incrementality, and much more, with tons of tools for ad monetization, cross-device attribution, and granular creative performance to name just a few.

There’s a ton of growth functionality there, and that’s great. 

All that reporting contributes to providing the “what:” what is happening as a result of my marketing and advertising?

But if you could add why to that what, you can add even more fuel to the growth fire. The why helps you increase the good results you’re getting in the what … and reduce any negative … because it helps you understand why things are working, or not. And that helps you fine-tune future campaigns.

You get the why, in part, via contextual data: everything that’s happening around the what.

That’s where Singular’s Extract tool comes in. It’s an ELT and reverse ETL tool that can pull raw data from almost anywhere: ad networks, app stores, CRMs like Salesforce, organic social platforms, and even your own backend. And it can load that data directly into your data warehouse for analysis.

How does that provide the why? 

New sets of data have just become easily available.

Why marketers need more data (and context)

Maayan put it perfectly:

“The MMP data is just a foundation … ELT fills in the missing context.”

That “beyond the MMP” context includes:

App store data

Beyond downloads and deletions, you can get ratings, reviews, subscriptions, crashes, and acquisition sources. You can also surface the “top-of-funnel” metrics like product-page views, impression counts, and tap-through rates (TTR). 

Plus, Apple’s App Store Connect provides “Product Page Views” and “Impressions,” letting marketers measure the conversion funnel from impression ➜ store view ➜ install. 

All of these signals reveal how effectively your listing converts attention into installs, and where users drop off. Is an increase in app crashes, for instance, correlated with a decrease in monetization? What happens to installs as we get more reviews, and as our star rating changes?

Getting this insight helps you both improve your ads, and improve the experience when people click on them.

Organic social insights

Looking only at paid campaign performance is like reading half a sentence. Organic performance fills in the rest. 

By tracking post engagement, follower growth, and comments across TikTok, Instagram, or Facebook, marketers can spot what resonates with audiences, and then use that insight to steer paid campaigns.  Organic content shows you what people love: paid distribution amplifies it. 

Together, they tell the full story of your brand’s reach and resonance, and provide important insights about what growth tactics to fund next.

Ad network APIs

Singular shows a ton of ad network data, but large platforms often limit how granularly you can pull data. Or which breakdowns you can pull at the same time.

For example, you might get creative-level metrics or geo breakdowns, but not both at once. 

Extract fixes that. It can grab multiple granularities, plus bids, budgets, placements, and timestamps directly from the platform APIs, giving you a far more complete picture. Even if privacy or aggregation rules prevent full user-level joins, you still get rich aggregate-level performance context that helps you optimize campaigns faster and smarter.

Put together, you can see the full picture: not just how many users you acquired, but what they thought, how they engaged, and what really made your metrics move.

And more … much more

There’s so much more you can do for user acquisition and growth campaigns with non-traditional datasets:

  • Maybe your app is weather-dependent, like a pollution tracker, so you want to pull in weather forecasts, or forest fire data, or other data that can factor into whether or not you pull the trigger on campaigns, or even automatically start up campaigns
  • Maybe you have a sports clothing app, and you want to pull recent data on scoring in specific leagues for specific players on specific teams, which could help you feature recently hot players in your near real-time ads
  • Maybe your app is fitness-related, so you want to pull in Apple Health or Strava trend data to understand when people are most active — and launch campaigns just before peak workout times in each region.
  • Maybe your app helps people save money or invest, so you pull in stock market volatility, crypto prices, or consumer sentiment data to trigger specific creative or messaging when the markets are heating up (or melting down)
  • Maybe you run a travel or events app, so you integrate flight cancellation data, festival calendars, or hotel occupancy rates to adjust bids or swap creatives dynamically when travel demand spikes.
  • Maybe your business depends on macro-level conditions, so you bring in economic indicators, fuel prices, or employment data to time ad spend with shifts in disposable income or consumer confidence
  • Maybe you run a sports betting or fantasy app, so you pull in injury reports, team rosters, or match odds to push hyper-relevant real-time ads featuring players or matchups that fans are obsessed with right now

The possibilities are virtually endless …

Real-time reaction: pulling data on your own terms

One of the most interesting use cases Maayan mentioned: some marketers are now pulling bid and budget data in near real time to react faster to performance changes: when our bids do this, our results do that …

Others are using Extract to build a change log: a unified view of every campaign adjustment made across networks. That way, they can ensure no optimization decision goes unnoticed or unanalyzed.

That’s serious power. 

It transforms your data warehouse from a dry historical archive into a live situation room for your marketing operations.

Giving the right data to the right teams

Sure, Extract works beautifully with enterprise BI tools and cloud data warehouses like Snowflake or BigQuery. But it can also push directly into a Google Sheet.

So if your CEO just wants to see “daily installs by region” or “App Store revenue by version,” you can automate that report: no SQL, no dashboard logins, no “pulling a report” … just the data that you need, automated, simple, easy.

And it’s not just execs:

  • Agency partners can get limited but relevant data … maybe not the full detail for privacy reasons, but enough aggregated performance metrics to see what’s working across channels
  • CFOs can get daily cost and revenue summaries automatically refreshed and formatted for easy reconciliation with finance systems
  • Creative teams can get their top three performing ad creatives each morning, ranked by installs, engagement, or ROI, so they know what to replicate, remix, or retire

That’s the power of flexibility: the same data can flow into dashboards, databases, or plain old spreadsheets, wherever your people actually work.

It’s the same accessibility whether you’re a massive gaming studio or a lean startup with a part-time analyst.

Integration with Singular: one pipeline to rule them all

Soon, Singular’s user-level ETL will run entirely on Extract. That means MMP users will be able to:

  • Move data seamlessly between Singular and Extract
  • Blend user-level performance data with app store or social data
  • Manage all of it via one login and one data flow

It’s a major step toward a single, flexible growth data stack built around you, not around the limitations of someone else’s schema.

Beyond the MMP: scaling higher

More data doesn’t necessarily mean more noise. It means you’re finally hearing the music beneath the metrics.

And you’re getting the context to understand why campaigns are working 1 day but failing the next. You’re connecting dots that you never had before.

So you’re not just tracking performance anymore. You’re orchestrating it.

That’s engineering growth at a higher level: beyond the MMP.

There’s much more in the full video:

  • 0:00 Intro – More data? More growth!
  • 1:05 What is Extract?
  • 3:00 Why MMP data isn’t enough
  • 5:10 App store data: reviews, ratings, revenue
  • 7:05 Organic social insights (TikTok, Instagram, Facebook)
  • 9:15 Granular ad network data and blind spots
  • 12:00 Near real-time bidding and budget data use cases
  • 14:30 ETL vs. Extract: what’s the difference?
  • 17:20 Flexibility: moving data where you need it
  • 20:00 Who’s using Extract: BI teams, marketers & beyond
  • 23:00 Getting started – free trial, 1M records/month

RIP Privacy Sandbox (we never really knew you)

Shall we raise a glass in memory of the privacy-focused attribution solution that never really launched? RIP Privacy Sandbox: we never really got the chance to know you.

And RIP Privacy Sandbox: you were a solution that the adtech ecosystem has just moved past.

But you were a lot of work:

Now, that’s all gone.

In an announcement on the Privacy Sandbox website, Google VP Anthony Chavez posted this today:

“We’ve decided to retire the following Privacy Sandbox technologies: Attribution Reporting API (Chrome and Android), IP Protection, On-Device Personalization, Private Aggregation (including Shared Storage), Protected Audience (Chrome and Android), Protected App Signals, Related Website Sets (including requestStorageAccessFor and Related Website Partition), SelectURL, SDK Runtime and Topics (Chrome and Android).”

In slightly fewer words, RIP Privacy Sandbox.

What Privacy Sandbox wanted to be

In 2018, Apple launched SKAdNetwork, and no one really noticed.

Until WWDC 2020, of course, when Apple announced App Tracking Transparency and elevated SKAdNetwork as the new standard for privacy-centric attribution on iOS.

But there had been some noise in the ecosystem about it, and it was on point with the privacy-sensitive, GDPR-aware, tracking-is-bad ethos. So in 2019, Google leaned into a framing of marketing measurement it hoped would satisfy everyone: kill the dangerous parts of tracking while building privacy‑preserving alternatives so that publishers, advertisers, and platforms can still thrive.

The idea: avoid a world where privacy kills the business model of free content on the internet. And the business model of advertising-led app growth.

On the web that meant replacing third‑party cookies with APIs like Topics, FLEDGE, and Attribution Reporting. On Android that meant eliminating device IDs … the GAID (Google Advertising ID), enabling on‑device interest signals and audience building, and re-inventing attribution via privacy-safe APIs.

It was ambitious. It was messy. 

It was a larger revolution from a marketing measurement perspective than SKAdNetwork, it broke more, and it required even more technological change.

RIP Privacy Sandbox: doomed from the start?

But from the start, Privacy Sandbox also had massive internal tensions between utility and privacy, between Google’s interests and ecosystem fairness, between timelines and engineering complexity, and between advertisers, ad networks, and privacy advocates.

For some, it didn’t go far enough.

For others, it degraded measurement too much.

But ultimately, the world just kinda moved on from the perceived need for Privacy Sandbox. Somehow, we’ve sort of evolved beyond the assumptions that made it seem urgent in the first place.

In 2020, privacy felt existential. SKAdNetwork dropped hard, and GDPR/CCPA heat was rising. Now, privacy still matters, and transparency is key, along with consumer choice, but the heat has been dialed back just a few notches.

Regulators haven’t dropped a bomb, consumers haven’t revolted en masse, and marketers have mostly adapted using the tools they already had: probabilistic models,  first-party data strategies, data clean rooms, and MMPs.

It might sound a bit self-serving, but MMPs are a big part of the story here. As a trusted and audited third party, major platforms give MMPs like Singular deeper access to their marketing data, which we can then share with advertisers at an aggregated level under specific privacy-related conditions.

Also, it’s just hard to change the world.

Apple owns iOS, owns the iPhone and iPad platforms, owns the App Store, and makes all the ecosystem software. And yet SKAN/AAK is not the default and only measurement methodology on iOS. Google doesn’t own Android in the same way — much of it is open source — and most Android devices are made by third parties who may or may not follow Google’s software and advertising lead. How then can Google enforce a massive conversion to Privacy Sandbox on mobile, never mind the open web?

The result: RIP Privacy Sandbox.

Why Privacy Sandbox is being retired, in the Blade Runner sense of the word

As always, there’s a bunch of reasons:

  1. Too big a leap, too fast
    Replacing fundamental pieces of ad infrastructure (cookies, identifiers, retargeting, measurement) at scale is brutal. Noise (data uncertainty) and limited granularity made ROI, attribution, and optimization harder.
  2. Ecosystem drag and cost
    Small to medium ad tech players struggled under the burden of re‑engineering their tech stacks. Some simply lacked the resources. Meanwhile, bigger players didn’t necessarily see a win for them in re-architecting everything.
  3. Regulatory and antitrust pressure
    Privacy Sandbox always carried a whiff of conflict: Google replacing a somewhat neutral pipe with one where it created and to some extent controls the APIs. Regulators like the CMA in the UK kept a wary eye.
  4. Demand side pushback
    Advertisers and marketers were burned by past “cookie apocalypse” deadlines. Many were slow to adopt sandbox APIs or test aggressively, waiting for more clarity or stability.

All in, it was a massive change. And it just ended up being too much.

Unified measurement from multiple signals is still the way forward

We’ve said it many times: you need multiple data sources to triangulate truth in our complex and messy era of marketing measurement. 

Arguably, the result is better than relying on a single identifier, even 1 as good as the GAID, and last-click measurement. 

Because ultimately, you want to know what’s incremental, not what got the credit, and a richer, deeper, more nuanced understanding of marketing measurement from multiple angles and sources delivers exactly that.

And thankfully, the big platforms have stepped up and are delivering more data — via trusted MMPs like Singular — to ensure we have more touchpoints to build reliable attribution.

RIP Privacy Sandbox

RIP Privacy Sandbox: you never really made it out of the cradle. 

You were conceived with fanfare, delayed repeatedly, reshaped under pressure, and ultimately unceremoniously retired.

Maybe your ideas will live on … mutated, rebranded, resurfaced under new acronyms and frameworks. The tension between privacy and performance isn’t going anywhere, after all.

But in the end, complexity alone doesn’t win. Utility does.

And for now, the industry has spoken.

RIP Privacy Sandbox!

LTV > CPI = success: here’s how you do that consistently

Achieving LTV greater than CPI is the foundation of success in mobile growth … obviously. But how do you consistently achieve LTV > CPI? According to AppAgent’s Roberto Sbrolla, with a full app lifecycle strategy.

I recently sat down with Sbrolla for a Growth Masterminds deep dive into successful long-term user acquisition and monetization, and the levers you can pull to build a sustainable and profitable mobile game business.

Hit play and keep scrolling for the highlights:

LTV > CPI: how smart strategy beats viral hopes in mobile growth

Everyone wants virality. 

Everyone wants that mythical ASO-driven organic wave that crests at the perfect time, never breaks, and delivers millions of users. (Cheaply, of course. And top-quality users who monetize all the time.

But let’s be honest: hope is not a strategy. Virality is awesome when it happens, but hard to engineer.

That’s why on a recent Growth Masterminds episode, I sat down with Roberto Sbrolla, Head of Growth at AppAgent and a 15-year digital marketing veteran, to talk about what actually does work.

Spoiler: it’s not luck. It’s kinda hard work.

But it is pretty simple. LTV > CPI wins.

Simple, but not easy.

The only math that matters: LTV > CPI

It’s no big secret, right? 

Everyone knows you have to get more out of your marketing campaigns than you put in. At least if you want to stay in business.

That’s why LTV > CPI is the golden equation in mobile marketing:

“Achieving LTV greater than CPI is the foundation of success in mobile.” — Roberto Sbrolla

That’s the game. But how do you do it successfully, and more than just once in a while? That’s where the strategy comes in: being able to achieve this consistently week after week, and month after month.

You have 3 key options when launching a mobile game:

  1. Go head-to-head with the giants
  2. Be the big fish in a small pond
  3. Create a new pond entirely

Each has trade-offs, and each requires different resources and planning.

David vs Goliath: the head-to-head strategy

Think you can take on Royal Match? Maybe you can. 

But you’d better bring a lot of firepower.

“You need a lot of money because you need to fuel the user acquisition at a very high level,” Roberto says.

match villains

Roberto points to Match Villains by Good Job Games as a rare success story with this strategy. Good Job Games went toe-to-toe with Goliath here. They saw what Royal Match publisher Dream Games was doing right, innovated inside the core game mechanics, and carved out their own niche in the matching game space. 

But take note: deep pockets and sharp planning were key.

Big fish, small pond

Most studios aren’t flush with $50 million in ad budget. And even if you are, do you really want to potentially flush it all down the drain?

So here’s the alternative:

Find a niche. Specifically, find and serve an underserved audience … one that the big successful publishers in the space have ignored or neglected. For example, if most match-3 games target women, what about one that appeals to men?

chrome valley customs

“Chrome Valley Customs from Offroad Games did exactly that,” Roberto says.

They built a match-3 game designed for male players — who doesn’t like restoring old cars —  and found success in a space where competition was lighter and CPIs were lower.

3. Make your own genre

Or, you can just completely invent your own brand-new genre.

Hard? Yes. Rewarding? Massively.

“Triple Match 3D reinvented the genre,” says Roberto. “They created something that wasn’t 2D, but 3D — and carved out a whole new market.”

match triple 3d

It’s the riskiest play, of course, because you need creativity, patience, a lot of user research, and a ton of luck. 

But … when it works, you own the space.

It’s not just where you compete … it’s how

Time is another critical variable. Are you pre-launch? Mid-scale? A 10-year-old game? 

Your tactics should change based on your maturity.

“Creating a successful game is a marathon,” Roberto reminds us. “And if you’re running a marathon, you’d better train.”

Understanding your game and your target audience — and what success looks like — are critical.

That means planning from the beginning: understanding CPI benchmarks, knowing your target audience, knowing what monetization levels you need to hit, and crafting the experience accordingly.

Tuning the CPI levers

So: back to LTV > CPI.

Lower CPI and higher LTV is the equation we’re all trying to solve.

One way to attack the equation, of course, is making CPI lower. Sbrolla outlined a few key strategies on this side of the equation:

  • Market research early
    Validate your concept before building the game
  • Marketability tests
    Run creative tests before dev is complete to test themes, visuals, characters
  • Mini-games
    Integrate marketable mechanics into the core gameplay to align ads with actual experience
  • Creative strategy
    Find great unicorn creatives that over-deliver: they’re the result of process, iteration, and clarity

“The most impact happens when users see the ad, install the game, and the gameplay matches exactly what they were promised,” Roberto says.

Oh, and don’t forget IP. It can help or hurt.

“IP can lower CPI by bringing in fans,” Roberto explains. “But it has to match the core gameplay. If there’s a mismatch, you’ll pay for it.”

Unlocking LTV: the other side of the coin

The other side of the LTV > CPI equation is monetization.

And while CPI is mostly market-defined, LTV is where your game design team earns its keep.

There are also some big levers to pull on this side:

  • Retention
    If people stay, they pay … if they bounce early, you’re burning budget
  • First-time user experience (FTUE)
    Remove friction and onboard well to get players to fun fast
  • Live ops & social features
    Leaderboards, challenges, and teams drive retention and revenue

“The higher the retention at the beginning of the curve, the better everything else will be,” Roberto says.

Royal Match, for instance, introduced teams early, and that had a big impact on revenue and retention.

ROAS campaigns are your friend

Still running CPI or cost-per-event campaigns? Maybe it’s time to switch.

“ROAS-focused campaigns consistently outperform,” Roberto notes. “They let you turn your money faster, which lets you scale faster.”

They also help you hit high margin goals, not just reasonable margin goals. That’s something too many studios ignore.

“Margin is the silent killer,” he says. “It’s not enough to get LTV > CPI … you need that plus margin to sustain your business.”

We all know creative is crucial for acquisition. But Roberto made a smart point about retention, too: “You target with creatives. Who you attract matters. That’s part of LTV.”

Attracting the right user … one who actually wants what your game delivers, is half the battle.

Final thought: strategy wins. Luck doesn’t scale.

You can get lucky. People do. Studios do. Developers do.

But you can’t count on it.

“If you want to be more than a one-hit wonder, you need strategy,” Roberto says. “And every strategy starts with a strategist.”

In other words, you need someone thinking from day one about market, positioning, competition, audience, and monetization.

Because if you want to win the marathon …

 … you gotta train for it.

So much more in the full podcast

Check out the full podcast on YouTube or any major audio platform for much more …

  • 00:00 Introduction to Growth Masterminds
  • 00:51 The Importance of LTV and CPI in Mobile Games
  • 02:24 Strategies for Achieving Higher LTV than CPI
  • 02:41 Competitive Positioning in the Mobile Game Market
  • 04:22 The Role of Funding in Strategy Execution
  • 04:47 Niche Markets and Innovation
  • 13:55 The Importance of Market Research
  • 15:44 Leveraging CPI for Competitive Advantage
  • 17:18 Understanding Market Response and CPI
  • 17:54 The Role of Mini Games in CPI and Retention
  • 19:17 Monetization Strategies and Hybrid Models
  • 21:03 Campaign Types and Their Impact on ROAS
  • 24:55 Importance of Retention and Early User Experience
  • 31:57 Social Features and Their Impact on Retention
  • 33:31 Strategy vs. Luck in Game Development

12 best console games of 2025: biggest sellers, highest rated, most loved

What are the best console games so far in 2025? And, what does that list reveal about how the industry is changing?

Of course, asking what the best console games are is a bit of a different question than which are the top selling console games. Anyone can grab a list of the top-selling games for 2025, and it’ll probably look a lot like this:

  1. Monster Hunter Wilds
  2. Assassin’s Creed Shadows
  3. The Elder Scrolls IV: Oblivion Remastered
  4. Call of Duty: Black Ops 6
  5. MLB The Show 25

And you’ll probably see some Elden Ring, some Doom, maybe Forza Horizon 5, and a mix of Minecraft, NBA 2K25, Grand Theft Auto, and EA Sports FC 25 in the mix, as they’ve led sales at various months over the past year.

But what are the best console games so far in 2025?

Answering that question means we need to look beyond sales numbers and look at what makes these games awesome … according to critics, sure, but especially players. And it reveals massive trends in the console gaming industry:

  • Indie games have never been stronger
  • Quality over quantity is winning
  • Cross-platform play is table stakes
  • Innovation within genre matters more than reinventing the wheel
  • Pricing strategy matters … but quality matters more
  • Social and creative elements trump graphics
  • The death of the 60-day review cycle
  • Players crave completion, not endless engagement

Here are 12 that made it for me. 

(Oh, and game developers: while you’re checking that out, take a look at Singular’s PC and console attribution and analytics product, as well as our cross-device attribution product. We can help you grow faster … more at the bottom of this post.)

Best console games: games that conquered 2025

Monster Hunter Wilds

This game sold well, but also had great gamer appeal.

  • Sales: 10.5+ million units
  • Critical reception: 89 on OpenCritic (196 reviews)
  • Verdict: This is a commercial juggernaut with broad appeal.

Monster Hunter Wilds didn’t so much launch as detonate. Capcom’s latest entry became the fastest-selling game in the publisher’s history, selling 8 million copies in just 3 days. 

By the end of its first month, it had sold over 10 million copies, a company record. Monster Hunter Wilds dominated U.S. sales charts for months.

What makes Wilds remarkable isn’t just its commercial success but how it achieved it. 

The game introduced seamless crossplay for the first time in the franchise, breaking down barriers between PlayStation, Xbox, and PC. 

That’s huge.

The elephant in the room? 

Performance issues on PC at launch led to mixed Steam reviews, showing that even massive commercial success can’t escape technical scrutiny. However, Capcom’s commitment to post-launch updates, including the addition of classic monsters and quality-of-life improvements, has gradually won back player goodwill.

Clair Obscur: Expedition 33

If Monster Hunter Wilds was kind of the expected blockbuster, Clair Obscur: Expedition 33 was 2025’s lightning bolt from a clear sky.

  • Sales: 5+ million units
  • Critical reception: 92 on OpenCritic (177 reviews, 97% recommended)
  • Verdict: The year’s biggest surprise and critical darling

This became Metacritic’s highest-rated new game of the first half of 2025, scoring 93 on the platform while earning near-universal praise from critics.

And it sold pretty well too:

  • 500,000 units sold in 24 hours
  • 1 million within three days
  • 3.3 million after 33 days
  • Over 5 million copies as of October

For a $50 indie title from a new studio, these figures are … impressive. 

What’s the secret? 

Clair Obscur blends turn-based RPG mechanics with real-time action elements all in a cool Belle Époque-inspired dark fantasy world. The game has voice acting from Charlie Cox and Andy Serkis, production values that you’d think came from AAA studios, and a super-cool story.

The really good news: innovation and craft can still overcome massive marketing budgets.

Hades II

When Hades II finally left Early Access just a month ago, it immediately claimed the crown as the highest-rated new game of the year. 

  • Sales: Early access numbers strong … sales are unclear so far
  • Critical reception: 94 on both Metacritic and OpenCritic
  • Verdict: Roguelike perfection refined to its peak

Supergiant Games’ sequel doesn’t just meet the impossibly high bar set by its predecessor … it may even surpass it.

The game keeps everything that made the original Hades addictive while adding more. Critics praised its “impeccable combat,” “flawless characterization,” and the way it evolves the formula without abandoning what made it special.

As GameSpot noted in their review, it’s “Hades 2 improves upon its predecessor in every way, making it a masterfully crafted sequel that is essential to play.”

Hollow Knight: Silksong

Few games have been as anticipated as Hollow Knight: Silksong. 

  • Sales: 6+ million units
  • Critical reception: 92 on OpenCritic
  • Verdict: The indie phenomenon that broke the internet

After 6 years of waiting, Team Cherry shadow-dropped the game on September 4 with just 2 weeks’ notice. Online storefronts crashed, the game hit 587,000 concurrent players on Steam, and it became the biggest game on Twitch within minutes.

Whoa.

3 million copies sold on Steam alone in the first few days, with total sales across all platforms reaching 6.4 million within a month. On Xbox, Game Pass downloads tripled PlayStation’s numbers, while Steam accounted for roughly 4 million of those downloads. 

Also, Silksong earned widespread acclaim for improving upon the original Hollow Knight in nearly every way: faster movement, more complex combat, brilliant level design, and the trademark Team Cherry attention to atmospheric detail. 

The fact that this entire phenomenon was created by essentially two people makes it one of gaming’s most remarkable success stories of 2025.

Split Fiction

Developer Hazelight has become synonymous with innovative co-op experiences, and Split Fiction continues that legacy.

  • Sales: Undisclosed
  • Critical reception: 91 on Metacritic
  • Verdict: Excellence from proven masters

Following the critically acclaimed It Takes Two, this game tells the story of 2 writers — one specializing in sci-fi, the other in fantasy — who become trapped inside their own stories by a machine seeking to steal their ideas.

(Sounds like modern AI to me.)

What makes Split Fiction special is its creativity and the variety of its gameplay. Playing in split-screen, you and a partner can traverse wildly different environments, each requiring unique approaches and cooperation. 

GameSpot awarded it a rare 10/10, calling it “hilarious, compassionate, and delightful” and praising its multiplayer experiences.

Best console games: major releases that delivered

Death Stranding 2: On the Beach

More action, more combat, still in a great setting.

  • Sales: 1.4 million (estimate)
  • Critical reception: Strong (PS5 exclusive)
  • Verdict: Kojima’s vision fully realized

Hideo Kojima’s sequel to his divisive 2019 masterpiece is a sequel worthy of the masterpiece mantle. The game features more action and combat while addressing nearly every criticism leveled at the original. 

Set primarily in Australia, it maintains the bizarre post-apocalyptic world while expanding it with improved gameplay systems.

Kingdom Come: Deliverance II

A medieval masterpiece with depth and complexity.

  • Sales: 2+ million on Steam, 600k+ on consoles
  • Critical reception: Strong across platforms
  • Verdict: Medieval immersion at its finest

The sequel to the 2018 medieval action-RPG doubles the size of the original’s open world and continues the story in 15th century Bohemia. Critics praised it as “a medieval masterpiece” with “staggering scope, depth and complexity.” 

The game allows players to be diplomats, craftsmen, knights, or cutthroats … or any combination thereof.

Assassin’s Creed Shadows

Yeah … it’s Assassin’s Creed …

  • Sales: Top 3 in U.S. year-to-date charts
  • Critical reception: 81 on OpenCritic (206 reviews)
  • Verdict: Solid entry 

While maybe not quite reaching the critical heights of some competitors, Shadows sold well, spending much of the year in the U.S. top 5. The most-reviewed game of 2025 with 206 professional critiques, it shows Ubisoft’s continued dominance in the open-world action space, even if some critics said it follows familiar formulas.

Donkey Kong Bananza

Donkey Kong will always be awesome, right?

  • Sales: Uncertain
  • Critical reception: Strong Nintendo showing
  • Verdict: Switch 2’s platforming showcase

The Switch 2 launch lineup benefited enormously from Donkey Kong Bananza, which captured the same Nintendo magic and creativity and pure fun that we’ve seen for decades. Digital sales weren’t fully reported, but the game ranked highly on Nintendo platforms and reminded everyone why the company remains the king of platformers.

And yes … some wildcard successes

EA Sports College Football 26

Currently sitting at #2 in the U.S. year-to-date sales charts, this sports title shows the enduring appeal of the college football license and of course EA’s sports dominance.

Ghost of Yotei

Launching in October, Sucker Punch’s sequel has received strong critical reception (87 on OpenCritic, 99 reviews), with critics praising its “gorgeous landscapes and satisfying, fluid action” while noting it doesn’t revolutionize the formula.

NBA 2K26

The basketball simulation continues to dominate its genre, ranking #5 in U.S. year-to-date sales. As GameSpot noted, NBA 2K earns another year as the best annual sports game available.

What does this tell us about console games right now?

1. Indie games are stronger

The success of Clair Obscur: Expedition 33, Hollow Knight: Silksong, and Hades II is impressive. These aren’t just “good for indie games” … they’re competing with AAA blockbusters for critical acclaim, player attention, and sales.

Clair Obscur’s 5 million sales would be considered a success for plenty of mid-tier publishers. And Silksong’s 6 million copies in a month rival major AAA releases. 

As gaming analyst Rhys Elliott noted in his breakdown of Silksong’s success: “There’s a lot of shared DNA between souls-likes and the two Hollow Knight games … [the] ‘lower ceiling’ argument was already hilariously outdated, but Silksong has buried it completely.”

2. Quality over quantity

The concentration of success at the top of the market tells a compelling story. 

In some subgenres, the top two games account for 80% of the category’s revenue. Players are becoming more discerning. They’re spending time with fewer games but engaging more deeply with the ones they choose.

Monster Hunter Wilds’ 10+ million sales and Clair Obscur’s 5 million show players are voting with their wallets for polished, complete experiences.

And they don’t want rushed live-service games or half-baked releases.

3. Cross-platform play is table stakes

Monster Hunter Wilds’ introduction of crossplay wasn’t just a feature, it was essential to its success. 

In 2025 and 2026, players expect to be able to play with friends regardless what hardware each of them has bought. Games that deliver this see expanded communities, longer lifespans, and better word-of-mouth growth.

4. Innovation within genres matters

Almost every successful game on this list improves upon established formulas rather than creating entirely new ones. 

Hades II is a better Hades.
Silksong is a refined Hollow Knight.
Clair Obscur blends turn-based combat with real-time elements.

In other words: familiar genres combined in fresh ways.

This suggests that players appreciate evolution over revolution. They want studios to understand what makes a genre work, then enhance it thoughtfully rather than discarding everything in pursuit of novelty.

5. Pricing strategy matters, but quality matters more

Hollow Knight: Silksong’s $20 price point and Clair Obscur’s $50 tag certainly helped their success in an era of $60-80 games for consoles. 

But if these games weren’t excellent, they wouldn’t have sustained their sales trajectories. Players are willing to pay premium prices for quality, like Monster Hunter Wilds at $70, but nobody wants to pay AAA prices for CCC experiences.

6. Social and creative elements matter more than graphics

One of the most telling trends from 2025 gaming research is that players value social features, creativity, and gameplay over cutting-edge graphics.

(I mean, we learned this from Nintendo Wii way back in the day, no?)

This explains why platform-style games like Fortnite, Minecraft, and Roblox continue to dominate while gorgeous but ultimately hollow AAA titles struggle.

The success of stylized games like Clair Obscur and Hollow Knight: Silksong over graphics-first titles supports this. Beauty matters, power is great, and the latest consoles are impressive … but artistic vision trumps polygon counts.

7. The death of the 60-day review cycle

Team Cherry’s decision to not send out review copies for Silksong and still achieve massive success represents a major shift. Streamers and online word-of-mouth can generate more buzz than traditional review outlets.

The old gatekeepers matter less. 

This doesn’t diminish critical analysis … it just means power is more distributed.

8. Players crave completion, not endless engagement

The success of focused, complete experiences like Split Fiction, Blue Prince, and Clair Obscur’s 30-hour campaign suggests fatigue with games-as-service models. 

Players want games that respect their time, tell complete stories, and offer satisfying conclusions … maybe more than endless grinds designed to extract maximum playtime (and perhaps endless revenue).

Looking ahead … more is coming

As we move into the final months of 2025, a few major releases still loom: Metroid Prime 4: Beyond, Ghost of Yotei’s continued performance, and maybe some holiday surprises. 

But 2025 is already going to be remembered as a year when indie and mid-sized studios proved they could compete. And win.

  • Massive scale and technical ambition are still cool (Monster Hunter Wilds)
  • A clear vision can achieve critical and commercial success (Clair Obscur)
  • Patience, craft, and respecting your audience can create legendary results (Hollow Knight: Silksong)
  • And iteration on a theme can be its own form of innovation (Hades II).

So, in a way, these games are blueprints for console gaming’s future. A future where it’s not always about the biggest budget and the biggest team making the biggest game.

At least, not always.

Growing your top console game? Singular can help

Building a great game is great. 

But unless it’s marketed well, who will know?

Singular offers all the tools you need to measure and attribute your console game growth, helping you optimize your marketing strategy.

Console game makers need clear cross-platform insights. With Singular’s PC & Console Attribution product, studios can unify measurement across PC, console, CTV, web, and mobile, tracking user acquisition, engagement, and LTV all in one place. 

You can ingest in-game events from platforms like Epic, Xbox, Switch, then integrate them with your marketing spend data down to keywords, creatives, and campaigns. You can also send web event postbacks to ad networks to create a feedback loop and optimize UA campaigns using real funnel signals.

Adding extra power: Singular’s Cross-Device Attribution tooling. It connects touchpoints across devices and platforms (desktop, mobile, console, web), de-duplicating users and attributing conversions to the right campaigns. 

That means if a player discovers your game via mobile ad, later launches it on console, and then buys DLC on PC, Singular can help connect those steps and credit the right acquisition source. That helps you optimize spend to your highest ROI channels, understand how cross-platform behavior drives revenue, and reduce blind spots in your marketing funnel.

Oh, and if you need to know how your creative is performing, Creative IQ is AI for delivering only the best and most performant ads.

Talk to us today.

OpenAI is the everything-platform ushering in the era of AI-first marketing

Are you ready for AI-first marketing? I’m not talking about using AI to do marketing. I’m talking about AI platforms becoming the interface, the context, and the locus of your core marketing activities.

A tectonic shift is happening in the world of digital platforms. At its 2025 DevDay just yesterday, OpenAI announced two major launches: apps built inside ChatGPT and AgentKit for creating AI agents.

What makes this especially important: ChatGPT is the next billion-user platform. It is graduating from an app to a platform as we watch, and now we know that this platform will have both agents doing all kinds of work and apps conducting all kinds of business.

This matters, because if there’s anything we’ve learned about platforms from Apple, Google, Amazon, Microsoft, and IBM: successful platforms inevitably accrue most of existing ecosystem value to themselves.

And, they set the rules of engagement for everyone else.

Right now, ChatGPT has 800 million weekly active users, up from zero just a few short years ago. As ChatGPT grows into its emerging platform status, marketers need to recognize that this is likely to be one of the most important shifts in digital architecture in years, if not decades.

  • It’s a challenge to Google Search
  • It’s a challenge to Google Play
  • It’s a challenge to Apple’s App Store power
  • And, as we’ll see, it’s a challenge to much, much more

Because ChatGPT is a multi-platform. 

Maybe an uber-platform … maybe the everything-platform. 

And perhaps, finally, the 1 true western super-app.

ChatGPT … the everything-platform?

This is much more than just ChatGPT getting some new features.

Potentially, this is the beginning of a re-wiring of who controls digital distribution, who owns user attention, and who can monetize both embedded experiences and real-world retail purchases. In other words, the emergence of AI-first marketing. And, incidentally, AI-first products, AI-first games, AI-first … everything.

So it’s not just about mobile apps as we currently understand them. It’s not just about the App Store and Google Play.

Because we do everything in ChatGPT and its competitors. We get work done with LLMs. We buy things based on recommendations from LLMs. We answer health questions via LLMs. We get travel suggestions on LLMs. We do almost everything via LLMs … we’ll probably even play games inside LLMs before too long. 

Apps on ChatGPT will be able to appear inline, expand to full-screen, and use picture-in-picture, meaning games are a definite possibility.

This impacts mobile apps, for sure, which do much of this today, but it also impacts e-commerce platforms like Amazon. It impacts the open web and websites like Booking.com. It impacts productivity software like Google Workspace and Microsoft Office.

And probably much more I’m not thinking about right now.

Because ChatGPT is a multi-platform. It’s not a gaming platform or a productivity platform or a search platform or a commerce platform. Potentially, it’s all of those put together … the everything platform.

Let’s take a few minutes to dig into what this means, what opportunities it presents, where the risks are, and how marketers should position themselves.

Let’s start here: what did OpenAI announce?

First, here’s a quick breakdown of the key announcements:

Apps inside ChatGPT via Apps SDK

  • Developers can now build interactive apps that live within the ChatGPT interface. These apps respond to natural-language prompts, show interactive UI elements (maps, forms, sliders, etc.), and can be invoked within conversations.
  • For users: ChatGPT can suggest apps contextually (if you’re talking about finding a hotel, it might surface Booking.com). Or you can explicitly summon an app (e.g. “Spotify, make a playlist for my party”).
  • For developers: The SDK is open source, built on the Model Context Protocol (MCP). Developers can start building in preview now; later in 2025 OpenAI will open app submissions and monetization.  
  • Early partner apps include Booking.com, Canva, Coursera, Figma, Expedia, Spotify, Zillow.  
AI-first marketing

 

AgentKit for AI agents

  • AgentKit is OpenAI’s new tooling suite to help developers design, deploy, and optimize AI agents (automated workflows or “assistant agents”) more easily.  
  • It offers a visual interface for defining agent logic, connectors, evaluation pipelines, and UI embedding.  
  • Because apps themselves are now integrated, combining agents and apps means agents can trigger app usage or collaborate with them inside ChatGPT.
  • In other words: ChatGPT becomes the orchestration layer, not just a point of interface.

Taken together and building on OpenAI’s existing and growing scale, these changes usher in the first stages of AI-first marketing.

Why this is a big deal: a multi-platform power shift

Below are some of the biggest implications.

1. The platform becomes the interface

Today, to reach users, customers, and players, services need to …

  • Publish on the web and rank well in Google Search
  • Publish in the App Store and Google Play (and rank well)
  • Buy ads to promote their apps or sites, or …
  • Choose organic/owned tactics for getting attention and traffic that they can monetize

If the new ChatGPT platform takes root and wins our time and attention, the interface becomes ChatGPT itself. Users don’t necessarily go out to search or open an app … they stay in the conversation. They invoke the service they want, or ChatGPT suggests it.

(Think how much money Amazon makes by suggesting products. Will ChatGPT monetize app suggestions?)

In this scenario, ChatGPT becomes the “operating system” … the super-app … the uber-platform. And apps merge into the fabric of that operating system incredibly seamlessly, in context. 

Context is critical here. Unlike any other existing operating system or platform, ChatGPT knows all your conversations with and learnings from third-party apps. That’s super-powerful, but also brings up privacy concerns. If Apple, Amazon, and Google have huge advantages in their ecosystems because they have more data, OpenAI could have much, much more.

That switches up the conventional go-to-market model.

2. Distribution via OpenAI, not Google or Apple

OpenAI is positioning itself as the gatekeeper of access. Rather than your app being discovered via Apple’s or Google’s app stores, your app (or agent) is discovered in ChatGPT, suggested by context, surfaced via ChatGPT’s directory, or invoked conversationally. 

(That’s the environment for AI-first marketing.)

Because ChatGPT has 800 million weekly users and continues to grow fast, the potential reach is enormous.

It’s literally getting comparable to or even exceeding many existing platforms and ecosystems.  

In that scenario, OpenAI could disintermediate (or at least compete with) the traditional app-store model controlled by Apple and Google, and the typical SEO model controlled by Google. (Plus other distribution mechanisms, like the physical product model currently largely controlled by Amazon.)

3. Monetization inside the interface

OpenAI is not just offering distribution: it’s baking commerce into ChatGPT itself. There is already talk of “Instant Checkout” and integrations with Etsy/Shopify, enabling users to purchase goods inside ChatGPT.  

That means marketers (especially e-commerce brands) could convert interest to sales directly within ChatGPT, rather than sending users off to a shop or website. 

That could be incredibly frictionless, a revolution in retail similar to how Amazon’s one-click purchase button changed the e-commerce world in 1997. 

The more friction is removed, the more powerful the conversion, and OpenAI is building not just commerce into ChatGPT, but agentic commerce … meaning perhaps your agents will eventually make purchases that you’ve pre-authorized.

In fact, that’s exactly what marketers expect from agents.

4. The rise of conversational UX and agents

Instead of thinking of your app or website as a static interface, product managers and marketers will now need to think in terms of agents, flows, and conversational experiences. 

  • How does your service work in a dialogue?
  • What triggers, states, decision logic, handoffs to UI, or fallback to web are needed?
  • What does engagement look like?
  • How do you impact retention?
  • How do you measure conversions?
  • How do you understand CAC?
  • What data will you have to calculate ROAS?

With Apps SDK and AgentKit, OpenAI is lowering the barrier to create these experiences. 

Marketers and product teams will need to think in these terms or risk being left behind. Some testing is going to need to happen, and happen fast.

5. Google search and SEO under threat

As users start relying more on ChatGPT to fulfill information, product, or service tasks inside the AI interface, the centrality of web search will continue to erode. We’ve seen that already since late 2024, when OpenAI rolled out ChatGPT Search.

(I joked with a CEO the other day that web search is like cable TV … only for old people. That’s not 100% true in either case, but it’s directionally accurate.)

That means your brand and content must adapt to appearing not just on web pages but via conversation-based apps or agents. “Ranking” in an AI-first marketing reality becomes less about keywords and more about alignment with conversational intents, app matchmaking, and contextual suggestions.

And web-based marketing, whether mobile or desktop, changes drastically.

What marketers should be thinking about (and doing)

There’s clearly a lot of change coming in the era of AI-first marketing. Where do you start?

Given this shift, here’s how marketers can proactively adapt:

  1. Start building or prototyping ChatGPT-native experiences
    • Identify core brand services or features that could be reimagined as conversational apps (e.g. product configurators, quizzes, concierge, content recommendation, booking).
    • Use the Apps SDK to prototype mini-apps or embed parts of your service into ChatGPT contexts.
    • Think in modular flows: rather than an entire app, start with micro experiences.
  2. Design for conversational UX, not just UI
    • Map user journeys as conversations, with branching logic, fallbacks, state management, context carryover.
    • Be mindful of how prompts, clarifications, system vs user messages, and UI renderings (sliders, forms, maps) interplay.
    • Plan for error handling, “Are you sure?” confirmation, fallbacks to web links when needed.
  3. Leverage agent workflows
    • Use AgentKit to create intelligent agents for recurring tasks: onboarding help, content recommendations, or even full customer service flows.
    • Agents can trigger app usage internally (e.g. “Your agent can invoke the booking app now”) or orchestrate multi-step tasks.
  4. Embed commerce and reduce friction
    • If your brand sells products, explore how Instant Checkout (or its equivalent) might let users transact inside ChatGPT.
    • Rethink funnel drop-off: fewer context switches, fewer page loads, more in-chat conversion.
  5. Content strategy for prompts & intents
    • Recognize that your content and messaging need to align to user intents as expressed conversationally (not just web SEO).
    • Consider prompting frameworks: “When a user says X, your app could respond Y; next prompt Z.”
    • Monitor which intents are most invoked and optimize responses or UI embed accordingly.
  6. Think about app discovery & positioning
    • Because apps will likely have a directory or catalog inside ChatGPT, your app or agent will need to meet high standards for usability, design, speed, and clarity to be surfaced or featured.  
    • Invest in onboarding, branding, and contextual prompts that encourage users to “install/enable” your app.
    • Think about how you’ll market your app on ChatGPT in ads and organic experiences. “Word of mouth” will have a whole new meaning in contextual LLM-driven app invocations.
  7. Watch data, privacy, and interoperability
    • Users will be prompted to connect apps and grant permissions; your app must be transparent and trustworthy.  
    • Consider how to link your app’s backend with your existing systems while respecting privacy, compliance, and security.
    • Be prepared to interoperate with other ChatGPT apps or modules (e.g. combining your app with analytics, content, or more).
  8. Monitor performance and feedback loops
    • With usage inside ChatGPT, metrics will shift: usage sessions, app invocation counts, conversation abandonment, agent success rates.
    • Build feedback loops to refine prompt flows, UX, error handling, and context retention.
  9. Experiment early, but manage expectations
    • As this is still new and evolving, early results may be uneven.
    • Focus on high-value verticals (e.g. travel, commerce, media) where integrated experiences matter most.

Of course, as with any major platform shift, there are dangers to keep in mind:

  • Centralization & dependency
    If your entire business becomes deeply tied to ChatGPT, you are dependent on OpenAI’s rules, ranking logic, monetization terms, and changes.
  • App discoverability competition
    Only some apps may be highlighted or surfaced prominently. You might get buried unless your UX or value is strong.
  • Privacy & data control
    Handling permissions, user trust, and data security will be essential.
  • Performance constraints
    Latency, context limits, memory retention, or model hallucinations may degrade experiences.
  • Regulation, antitrust, and scrutiny
    As OpenAI becomes more of a “platform” provider, it may attract regulatory or competitive pressure.

Entering the age of AI-first marketing

For marketers, this is kinda like when web browsers first became the dominant interface, or when mobile apps became the new new thing.

In a similar way, the emerging shift to AI-first marketing, or conversation-first product and brand experiences is not incremental … it’s structural.

Brands that begin thinking inside the conversation … building apps and agents for ChatGPT …  will be better positioned than those still fixated on web pages or mobile apps alone. 

The new gatekeepers of discovery and conversion in AI-first marketing are very likely to be AI platforms like OpenAI, not Google or the App Store.

But don’t panic.

None of this will happen instantly. And though TV killed radio, radio still exists … as does TV, even though streaming is bigger now, and the internet is bigger than both.

So: the sky is not falling. At least not instantly.

But there very well may be a new sheriff in town.

And it’s always a good idea to get to know the new powers that be.

Mobile retargeting: 12 actionable takeaways from global experts

Is user acquisition the art of getting attention? If so, mobile app retargeting is the science of not losing it. And, ultimately, making it profitable.

Mobile marketers spend 10s if not 100s of billions running ad campaigns, driving installs, encouraging sign-ups, and pursuing purchases. But as all UA pros know, the hard horrible truth is that most users vanish within hours or days. If you have 15% of your acquired users in a cohort 30 days later, you’re an all-star.

But of course … that means 85% have jumped ship. 

Mobile retargeting is the quiet hero of growth marketing. It can turn that fleeting attention into lasting value. But only if you do it right.

That’s why we just released 2 resources for you … 1 you can read, and 1 you can watch:

They’re free and available on-demand right now: check them out.

While you do that, here are just a few of the key learnings from the webinar, which featured experts who run mobile retargeting campaigns dozens if not hundreds of times each month.

  • Gabriel Oyarzabal, VP LATAM, Jampp
  • Karl Berta, VP Americas, Appier
  • Jonathan Yantz, Managing Partner, M&C Saatchi
  • Martje Abeldt, CEO, RevX
  • Olivia Sears, Head of Account Management, Smadex
  • Angela Humphrey, VP of BD, YouAppi
  • Mike Gadd, Customer Success Director EMEA & India, Singular
  • John Koetsier, VP Insights, Singular (moderator)

Here are their biggest takeaways:

1. Mobile retargeting and re-engagement: better together

We kicked off with an important distinction: re-engagement is owned, retargeting is paid. Think email, push, and SMS for re-engagement; think DSPs, social ads, and programmatic for retargeting. 

But think about them together.

Both matter, and both should be part of a holistic strategy. 

As Olivia Sears noted, segmentation is the key: different cohorts need different nudges. Cart abandoners might require urgency; lapsed users may just need a reminder.

“ Retargeting is more about intent,” she says. “Reengagement rather, is more about how to add value, or rebuild the relationship.”

Critically important: unifying the data layer for owned and paid channels, says Mike Gadd. Data has to flow into one system, enabling real-time audience updates and consistent messaging across push, email, CTV, and mobile ads. 

Frequency caps are essential: nobody likes feeling stalked by an app.

And consistency matters, says Karl Berta. A user who gets an SMS with a discount and then an ad with a completely different message is confused, not convinced. Or, worse, annoyed.

And that won’t boost sales.

2. Budget allocation: start with 20–30% for mobile retargeting

How much of your growth budget should go to retargeting? 

Jonathan Yantz suggested 20–30% as a starting point, but emphasized: it depends. 

App lifecycle, vertical, and audience maturity matter. Brand-new apps, of course, spend zero on mobile retargeting because they have no or few lapsed users to retarget.

On the flip side, Martje Abeldt shared that mature apps often spend 50%+ on retargeting because they’ve already saturated their UA channels and maybe their markets, and it’s much more about getting users back.

3. When does the clock start? Right away

When should you retarget?

The answer might surprise you.

Gabriel Oyarzabal dropped one of the most striking insights: in some cases, mobile retargeting should start within 15 minutes of install.

As in, you’ve just won them, and BOOM … you’re instantly retargeting them.

Why?

Because attention drops off fast. I can’t count how many times I’ve installed an app or game, then kept playing my game or scrolling my feed. When the game’s over, I don’t even remember to go back to the app that I just got. A few months later I noticed this app or game icon and think … what is this?

That’s why Oyarzabal essentially telling us that the retargeting clock doesn’t start at D7 … it starts at minute 15.

The moral of the story … if you wait, you risk losing users and not even getting that first open. Push notifications and emails can help, but they’re limited. 

Paid retargeting can catch those who slip through the cracks and ensure that the users you’ve acquired actually open your app or game.

4. Measurement: incrementality or bust?

If you’re not measuring incrementality, you just might be flying blind. Multiple panelists hammered this point home. 

You can do it lots of different ways — talk to Singular about how — but whether you’re using geo-holdouts, ghost ads, A/B testing, or just wide divergences in spend and activity, incrementality is probably the best way to know if your retargeting spend is actually driving conversions you wouldn’t have gotten anyway. 

As Gabriel Oyarzabal emphasized, it’s not a one-time thing: incrementality testing should always be on.

5. iOS is a challenge for mobile retargeting, sure, but increasingly, so is Android

We know mobile retargeting on iOS is tough with ATT and IDFA opt-outs, but it’s not impossible.

But it’s also getting tougher on Android in some geos.

 ”In some European countries, Scandinavia, for example, opt-out rates on Android are very, very high … about 50%,” Abeldt said. “ So when you look at the global landscape, there’s a lot of audiences that are opting out these days also on Android, and it’s a total blackout, basically on the device.”

That means you have to adopt different tactics.

Which ones?

Angela Humphrey highlighted broader privacy-compliant segmentation plus a shift to aggregated signals like location, device type, OS version, app version, and more. Then for measurement, incrementality still works great because it’s not dependent on user-level signals.

6. Creative: personalization, relevance, and DCO

Retargeting is only as good as the creative you use. (Of course, timing, targeting, and offer matter too.)

Segmentation is critical to get the right creative and offer in front of the right users.

Ideally, you want to align creative with segments, says Olivia Sears. Big spenders should get different messaging than casual users. One way to personalize creative that doesn’t rely on user-level data: context-aware creative. These are ads that change based on weather, time of day, or market conditions.

And dynamic creative optimization (DCO) is an increasingly big deal. DCO has broken out of retail and is now effective in gaming, finance, travel, and subscription apps.

One of the reasons I personally like it is that DCO is like a mini-shopping experience in the middle of your scroll … just pause, see if you’re tempted, and if so, tap it for more info.

7. Seasonality: don’t just shout louder in Q4

Holidays are expensive and noisy. 

You want to have specific campaigns that align with the time of year and the events that are going on, but it’s risky: you don’t want to test brand-new creative when ads are the most expensive they’ll be all year.

So both Olivia Sears and Mike Gadd recommended evergreen campaigns that run year-round, so you’re not scrambling in Q4 with untested strategies. 

And, if you’re going to some holiday-specific messaging, build off one of your high-performing templates.

Start early, extend into post-holiday windows (Q5!), and complement paid retargeting with owned channels to control costs.

8. Future of retargeting: AI and cross-device?

AI is changing everything, so why not mobile retargeting too?

As most marketers, our panelists expect AI to revolutionize creative, from generating video ads for every SKU in an e-commerce feed to crafting hyper-relevant messages at scale. 

Beyond AI, however, cross-device marketing looks to be big.

Cross-platform and cross-device marketing, done right, builds that surround sound marketing effect that I’ve talked about before. It’s the ability to seem like your brand is everywhere … even when you’re tiny.

And that can be huge in driving action.

9. Orchestrate frequency across owned + paid

This is really hard to do when you don’t have IDFAs and GAIDs might be scarce in the geo you’re targeting, but make an effort to coordinate messaging caps across push, email, and SMS so you don’t over‑serve the same person. 

That’s just annoying, and we all know it.

Best-case scenario here is that you have users/players/customers who have willingly given you first-party data like email addresses or phone numbers, and you can use those cautiously but effectively.

Ideally, you centralize audiences and regulate overall pacing of all messaging so that one channel doesn’t nuke performance in another.

Or, worse, cause churn rather than prevent it.

Post‑click matters. 

Nothing kills a conversion faster than confusing, different, or broken destinations.

Always route people to the exact screen, information, or offer promised in your mobile retargeting ads … even after an update or reinstall. It’s the fastest way to turn intent into action.

That means deep links, and it means deferred deep links.

11. Practice audience governance

Separate UA and retargeting with clean suppression lists, “do‑not‑target” uploads, and postbacks of every install to partners. 

It’s easier for an ad network partner to find you “new” users when it’s actually mobile retargeting at work. Ensure that doesn’t happen by practicing good campaign hygiene.

This avoids double-counting, eliminates cannibalization, keeps budgets honest, and ensure you know accurate ROI for both your UA campaigns and your mobile retargeting efforts.

12. Map bottlenecks, then retarget to unblock

The best mobile user acquisition marketers are also (almost) product managers who know their own apps better than almost anyone else.

Run cohort analysis on post‑install funnels (D0, D3, D7, D14, D30) to find drop‑offs. Identify the blocks or hurdles. Then build creative and offers that nudge users to the next milestone.

Now you’ve got a genuine and powerful mobile retargeting message … not just some generic “come back to us.”

Unlocking value with mobile retargeting

Mobile retargeting isn’t just about chasing users. Done right, it’s about delivering value, maintaining relevance, and extending lifetime customer relationships.

It also reduces your per-use cost of acquisition by wringing more value from your initial UA investment.

Or, as one panelist quipped: “Reduce CAC, get them back.”

That’s retargeting that works.

There’s so much more in the full webinar. Check it out here. And, of course, don’t forget to get the mobile retargeting guide, which is available right here.

Here’s what to expect from the full webinar:

  • 01:16 Welcome and Overview of Retargeting Networks
  • 02:39 Agenda and Expert Panel Introduction
  • 04:21 Introduction to Singular and Webinar Purpose
  • 05:19 Key Insights from the Retargeting Report
  • 06:24 Audience Polls and Initial Panel Discussion
  • 07:43 Deep Dive into Retargeting Strategies
  • 18:51 Measurement and Incrementality Testing
  • 23:41 Audience Engagement and Poll Results
  • 25:14 Challenges and KPIs in Retargeting
  • 26:51 Retargeting on iOS and Android
  • 27:33 Privacy Compliance and Measurement Strategies
  • 28:16 Targeting and Measurement Challenges
  • 32:00 Creative Strategies for Reengagement
  • 39:36 Running Effective Re-engagement Campaigns
  • 47:10 Future Trends in Retargeting
  • 52:10 Top Takeaways and Q&A

iOS 26 and the frozen Safari user-agent: why you don’t need to care

TL;DR

  • Safari’s user-agent is frozen in iOS 26
  • It always says “iOS 18.6,” regardless of your real version (and, as far as we know, it always will)
  • This looks scary but has zero impact on attribution in Singular
  • Apple is tightening fingerprinting protections, not targeting UA marketers specifically
  • You’re safe … keep calm and carry on

Time is standing still. While Apple’s iOS mobile operating system is at version 26, the company’s Safari web browser reports iOS 18.6 and probably will in perpetuity. The question is: will the frozen Safari user-agent have a negative impact on your attribution capabilities?

Short answer: no, as long as you’re using Singular.

Longer answer … keep reading!

Frozen Safari user-agent in iOS 26

Apple’s latest privacy move in iOS 26 might look scary at first glance.

Safari now reports the exact same user-agent string no matter which iOS version you’re running. In other words: if you’re on iOS 26, Safari still insists you’re on iOS 18.6. When iOS 26 transitions to iOS 27, Safari will still report iOS 18.6, getting increasingly out of step with the actual underlying operating system.

That’s it. Forever, as far as we know.

This looks like a problem. Marketers will immediately wonder what happens to attribution … measurement … segmentation … will a frozen Safari user-agent break all of these things?

The short answer: no.

The long answer: still no … but let’s unpack it.

Why the change?

The frozen mobile Safari user-agent is just part of Apple’s broader privacy push. 

It’s in the same family as link tracking protection, stripping referrer parameters, and other AAK-style changes that showed up in iOS 26.

On September 15 Apple quietly announced in the Safari 26.0 release notes that the user-agent would be frozen:

“Safari now reports a frozen OS version in its user agent string on iOS 26 and iPadOS 26, showing the last version released before iOS 26.”

The goal is to impair device fingerprinting and make it harder for data brokers to build unique device profiles from small technical signals that devices and browsers have always shared in digital handshakes when communicating with servers and requesting data or resources.

This means Safari won’t report (or leak, depending on your perspective) the actual OS version anymore. Instead, it will always report iOS 18.6. Interestingly, this aligns mobile Safari with Safari on Mac, which first started freezing the Mac OS string way back in 2017.

Yeah, a frozen Safari user-agent is a bit of a big deal

If you’re not in mobile user acquisition or performance marketing, “user-agent” might sound a bit like boring plumbing. But it’s one of those background signals that historically helped attribution providers confirm which OS version a user was running. 

That can help with probabilistic attribution, and it can also provide important signals about device age, hardware capability, and software compatibility.

So yeah … when Apple freezes the Safari user-agent, people worry. If Safari won’t report iOS 26, won’t that mess with campaign tracking, SKAN reporting, or attribution accuracy?

No user-agent, no worries

The good news is that a frozen Safari user-agent won’t impact your Singular measurement at all.

We saw this change months ago in the iOS 26 beta and made sure our attribution models were ready. 

Singular attribution doesn’t rely on the Safari user-agent string, and freezing it doesn’t affect how conversions, installs, or revenue get tracked in Singular.

In other words: our systems already account for this, there’s no change required, no loss of accuracy, and no broken dashboards.

So yes, it sounds like this is yet another signal going down the tubes. But, in reality, this particular change is more of a cosmetic tweak than a measurement crisis.

The bigger picture

What Apple is really doing here is reinforcing its privacy-first positioning. User-agent strings can be abused for building device graphs and more, so Apple is trying to cut them off at the source.

It’s essentially the same logic as iOS 14.5’s ATT rollout: limit unnecessary signals, get advertisers and platforms to rely on provided APIs like SKAN, and shrink the surface area for tracking.

The good news is that Singular’s unified measurement isn’t impacted, and we still get all the data we need for accurate and fast attribution.

Something big happened to SKAN 4 while we weren’t paying attention

Something big just happened to SKAN 4. Specifically, to SKAN 4 adoption on the Meta platform. And almost no-one noticed.

I certainly didn’t.

The big question, though, is whether it matters.

Meta’s driving SKAN 4 adoption

Meta is now driving SKAN 4 adoption. Meta isn’t completely on SKAN 4. Not 100%. But currently, 44% of SKAN postbacks that we’re seeing from Meta are SKAN 4, not SKAN 3.

There’s the big blue social network, down at the far right of this chart:

SKAN 4 adoption

 

It’s literally almost the first significant event in the SKAN ecosystem for over a year. Plus the most significant event in SKAN 4 adoption perhaps ever.

And it leaves Google as the single remaining big holdout.

It also marks the first time SKAN 4 postbacks have sustainably become the majority of SKAdNetwork postbacks. We’ve seen a few spikes here and there, but they’ve always been a flash in the pan.

This trend looks sustainable:

skan 4 2025 trend

Thank you Gabriel Rosa

Singular’s had the SKAN Adoption Dashboard for a long time. And for a long time I checked it religiously, scouring for the slightest evidence that something was changing in the SKAN world.

Because, for a while, it really, really mattered.

But I’ve gotten out of the habit.

Not for everyone, apparently. Big thanks to Gabriel Rosa, a user acquisition coordinator for By Aliens, who checked the dashboard recently, and then posted about it on LinkedIn.

Gabriel Rosa SKAN 4

 

(Note: he posted in Portuguese; I’m showing LinkedIn’s automatic translation to English here, and it may contain errors.)

“Meta Ads finally heading to SKAN 4.0?” Rosa asks. 

As he notes, it’s been about 3 years since Apple released SKAN 4, and we really haven’t seen full adoption by the biggest players.

The biggest question about SKAN 4 in 2025

But there’s a big question to ask here.

First off, let’s be clear: it’s nice to see the ecosystem moving forward to SKAN 4. As I wrote over a year ago, SKAN 4 brings a bunch of good things that SKAN 3 can’t deliver:

  1. Fixes SKAN 3’s volume limitations by returning more data at lower volumes
  2. Adds a longer data-collection and attribution timeframe with multiple postbacks up to about 35 days
  3. Enables better creative optimization insight with a more granular source identifier
  4. Provides data to support more effective ad network optimization with richer signals for delivery and targeting
  5. Delivers better measurement of reality with less loss and more accurate conversion capture
  6. Introduces additional complexity with more signals and harder implementation
  7. Improves reporting for subscription apps and long conversion-window models
  8. Improves reporting for low-user-count apps by allowing signal return even at low volume through coarse conversion values

(Get all the details in that post linked above if you want them.)

But the biggest question about increased SKAN 4 adoption in 2025 is, honestly, does it even matter?

And the answer is both yes and no.

Yes, more data is better. And yes, we’ll feed that into our attribution and optimization engines. So yeah, it’s a good thing.

But also no.

The industry has engineered around ATT and SKAN to a fairly significant degree. Singular’s Unified Measurement uses multiple signals to triangulate marketing attribution, and it works so well we’re talking about golden ages of marketing measurement.

So SKAN 4 isn’t as important today as we thought it would be in early 2024.

More is more, and 4 is more

That said, SKAN 4 does offer more data and more insight into what’s working in mobile marketing. Getting that from Meta as well as the rest of the adtech ecosystem is generally a good thing.

And SKAN 4 feeds more data into Unified Measurement.

Also a good thing.

It’s just not quite the salvation that we once thought it might be.

Hello Gemini! Singular now supports Gemini as well as ChatGPT, Claude, Cursor, Visual Studio Copilot

Hello Gemini! Singular’s MCP integration now supports Google’s Gemini so you can talk to your data via yet another LLM.

If AI is eating marketing, Singular’s making sure you get a seat at the dinner table. 

And … that you get some good munchies.

By which we mean, of course, accurate, clean, and trustworthy data straight from the most accurate possible source: your own data in our databases and dashboards.

This week Gemini CLI joined ChatGPT, Claude, Cursor, and Visual Studio Copilot as part of our growing set of LLM integrations. So you can now talk to your Singular data directly through Gemini, asking natural-language questions and getting instant, AI-powered answers: no SQL, no dashboards, no building queries, no waiting on your BI team.

Just you, your voice, and your data.

So … what can you ask Gemini?

Just connect Gemini-CLI and you can ask pretty much anything you want:

  • What are my top performing campaigns?
  • What creatives are working on TikTok?
  • Show me a chart of installs by geo over the last 90 days.
  • What CPI trends are you seeing month-over-month by ad network?
  • Where should I be spending more?
  • Where should I be spending less?
  • And much, much more …

Gemini will instantly give you insights, tables, and visualizations, all generated from your very own Singular data.

  • Gemini brings the brain
  • Singular brings the data
  • You get the insights

Even better, we don’t just support Gemini. There are now so many options for which AI engine you’d like to plug into your growth data …

Gemini’s got company

Gemini joins a growing list of LLMs that Singular supports, thanks to our standard, private, and safe MCP integrations.

So far, they include:

  • Claude by Anthropic
    Singular was the first MMP to enable direct access to your data via an LLM when we released support for Claude in June.
  • ChatGPT by OpenAI
    ChatGPT followed Claude, which makes sense because Anthropic invented the MCP integration protocol that all the LLMs are now using.
  • Cursor by Anysphere
    We didn’t really announce it, because it’s pretty geeky, but if your devs want to play with code, AI, and data, Cursor is a really good way.
  • Visual Studio Copilot (GitHub Copilot)
    We didn’t really announce this one, so I’m slapping my own wrist here again, but the reason was the same: it’s super geeky for developers who want to build cool stuff with their data (and AI).
  • ChatGPT – Developer Mode
    This is still in beta, but yeah, developers like options. So this is available too …
  • NEW: Gemini CLI
    Right now, Gemini supports MCP integrations to Gemini-CLI, which is the Gemini-CLI desktop app. The browser-based Gemini doesn’t support MCP integrations just yet, but we’ll connect it as soon as it does.

And yes, more are on the way … including super-cool agentic AI engines that will support even more advanced use cases and complex jobs.

OK … how do you connect Gemini?

Short version: get all the details from our Singular MCP doc in our Help Center.

Slightly longer version:

  • Install the Gemini CLI app
  • Configure Gemini to use Singular MCP
  • Run Gemini in your terminal app
  • Authenticate with your Singular account when prompted
  • Give Gemini the right permissions

When all that’s done (and yeah, get the exact details from the Help Center) you can easily ask Gemini for any of the data you want. Take a look here to see what people are commonly using our MCP LLM integrations for.

MCP LLM integrations

Worried about LLM hallucinations?

So are we.

Good, clear, accurate data is critical to what we do, because it’s critical to what you do.

That’s why we’ve invested heavily in ensuring that hallucinations are as rare as possible when you use the Singular MCP. We literally have a 7-step approach to forcing any LLM accessing your Singular data to do its best work, simply, strictly, observably … without making stuff up.

This makes bad data extremely rare. 

That said: you still need to be aware, and ensure that Gemini’s answers — or answers from any LLM you connect to your Singular data — pass the sniff test.

Not a Singular client yet? Can’t access this Gemini goodness?

Well, let’s fix that.

You can start for free, and you can also book a product demo to get a sneak peek under the engine that powers growth from some of the best and biggest brands on the planet.

All it takes is a click.

See you soon!