Spring break 2026 didn’t happen in a vacuum. It unfolded amid TSA disruptions, rising gas prices, and broader global uncertainty. Despite everything, many people still traveled.
Start.io analyzed mobile user data between March 15 and April 15 to understand how and where movement played out. The result is a clear, data-driven view of spring break travel across the U.S.: from distance to destination to regional behavior.
Nearly Half of Americans Traveled, and They Mostly Stayed Domestic
To start, travel volume remained strong. In impressive 44.8% of Americans traveled domestically more than 100 miles during the spring break period. By comparison, just 2.4% traveled internationally. That’s a nearly 19:1 ratio of domestic to international travel.
And when people traveled domestically, they didn’t stay close:
- 66% went 500–1,000 miles
- 24% went 200–500 miles
- 10% went 100–200 miles
One City Stood Above the Rest
When it comes to domestic destinations, one city clearly stands apart: Chicago. Chicago ranks as the top destination for 25 states, accounting for half of the country.
The states sending the most travelers there span the Eastern half of the U.S., including the Southeast, Midwest, and Northeast. From Georgia and the Carolinas through Ohio and Illinois, up to New York and Massachusetts, Chicago consistently emerged as the central draw.
Where Chicago isn’t the top destination, regional patterns take over quickly:
- Western states (Arizona, California, Colorado, Nevada) favored Los Angeles.
- Florida travelers largely stayed in-state, with 19.9% heading to Miami.
- Texas travelers leaned toward San Antonio (12.9%).
- Mountain states like Montana and Wyoming gravitated toward Denver.
- Gulf states such as Louisiana and Mississippi preferred Houston.
International Travel Followed a Different Pattern
While domestic spring break travel showed a mix of national and regional hubs, international travel is far more concentrated. Canada was the top destination for 46 out of 51 states, accounting for 90% of states and about 15.4% of total international traveler share.
After Canada, there’s a noticeable drop:
- Mexico ranked second at 8.8%.
- Brazil followed at 5.8%.
A handful of states broke the pattern and favored Mexico instead: border states Arizona, New Mexico, and California, as well as Washington and Alaska, where travel skews toward warmer destinations.
Beyond the Americas, travel extended globally, though at a smaller scale, with destinations like South Korea, the United Kingdom, Germany, Japan, Australia, Singapore, and France each capturing a small share of travelers.
A Few States Drive a Large Share of Travel
Not all states contribute equally to travel volume. Texas leads across the board:
- 10.2% of total domestic travelers
- 10.3% of total international travelers
Florida ranks second in both categories. Together with Georgia, New York, and North Carolina, these five states account for 34% of all domestic travel.
Similarly, international travel is not evenly distributed. A small group of states accounts for a disproportionate share:
- Hawaii (23.6% of travelers go international)
- Alaska (22.6%)
- California (12.8%)
- Oregon (8.8%)
- Washington (8.5%)
These figures are well above the national average of 5.1%. At the other end, states like West Virginia, Iowa, Tennessee, Arkansas, and Louisiana show much lower international travel rates, all below 3.5%.
Taken together, the data paints a clear picture of how Americans traveled for spring break 2026: overwhelmingly domestic, often long-distance, and shaped by a mix of national hubs and regional proximity. From Chicago’s dominance across half the country to Canada’s outsized role in international travel, the patterns highlight just how structured large-scale travel behavior remains, even in a more unpredictable environment.
The IAB, during the 2026 NewFronts, showcased new content, new partnerships, and new ad formats. The presentations also revealed something more fundamental: The idea of “channel” is dissolving. What once looked like a clean divide between digital and traditional TV has now blurred into a layered, fluid ecosystem. And at the center of that ecosystem, quietly but unmistakably, is mobile.
Across the week, the contrast was striking. On one side were the connected TV (CTV) heavyweights (Amazon, Roku, Tubi, to name a few) positioning themselves as the natural evolution of linear television: premium, high-reach, big-screen storytelling environments. On the other were the digital-native platforms (TikTok, Meta, Snap, and others) emphasizing constant engagement, creator-driven content, and massive, always-on audiences.
But this framing misses a deeper shift. It’s not that the future of media is about CTV versus mobile. It’s about how all roads increasingly lead to or through mobile.
The Screen Is No Longer the Signal
At a glance, it might seem like big-screen viewing is having a moment. CTV investment continues to rise, with projections showing significant growth in ad spend as brands shift budgets from linear TV.
That momentum is real. But it’s also incomplete.
Even the platforms that built their reputations on the living room screen are now deeply embedded in mobile ecosystems. According to Sensor Tower data, streaming platforms like Netflix, Amazon Prime Video, and HBO Max each surpassed 20 million mobile app downloads in 2025.
In other words, streaming TV is no longer confined to the couch. It’s happening on commutes, in waiting rooms, between meetings, and in the quiet in-between moments of the day. The “TV experience” has escaped the television.
Meanwhile, mobile-first platforms like TikTok, Snap, and Instagram are redefining what viewing even means. Short-form video, creator content, and interactive formats are compressing and expanding engagement at the same time. A user might watch dozens of pieces of content in minutes or spend hours immersed in a continuous feed. The result is a world where the distinction between “watching TV” and “using your phone” is becoming increasingly irrelevant.
Why Mobile Leads the Omnichannel Reality
From our perspective, the increasingly diverse mix of companies showing up at the NewFronts underscores an important reality: Mobile is not just a channel. It is the connective tissue of the entire media ecosystem.
Mobile is the only environment that travels with the user across contexts. It captures real-time signals about behavior, location, and intent. It bridges moments, from discovery to consideration to action, in ways that no other environment can replicate.
A streaming session on a smart TV might begin on a mobile app. A purchase influenced by a CTV ad might ultimately happen on a phone. An augmented reality experience, like those showcased at the NewFronts, lives almost entirely within mobile environments, turning advertising from something you watch into something you experience.
The takeaway from this year’s NewFronts isn’t that one format will win. In fact, the idea of winners and losers is becoming less useful altogether. What matters is how seamlessly brands can move across environments and how well they understand audiences not as viewers of a specific screen, but as people navigating a continuous stream of content throughout their day.
In that continuous stream, mobile is the constant. Driven by real-time signals and delivered across every screen, mobile-first data provides the foundation for true omnichannel engagement.
The lines may be blurring. But the direction is becoming clearer.
This post was written by Meital Goldberg, Product Manager at Start.io.
As our video business continues to grow, we’re expanding our ad quality capabilities to deliver stronger coverage across video environments.
Our video response scanning integration enables us to scan video responses in near real-time, improving our ability to detect and address quality issues as they occur. With increased visibility into video traffic, we can more effectively validate creatives, ensure compliance, and provide greater confidence to both demand and supply partners.
This functionality builds on the foundation of our existing display and video scanning solutions, broadening our coverage as our video business grows. Together, these capabilities create a more unified and comprehensive approach to ad quality across formats.
By continuing to evolve our solutions alongside the growth of video, we’re helping ensure a more consistent, reliable, and high-quality experience for partners and users alike.
This post was written by Or Oren, Product Manager at Start.io.
We’re excited to introduce Ad Unit configuration support in our Bidding SDK- giving publishers greater control over how ad placements are defined and managed within their apps.
With this update, publishers can configure multiple ad units, each representing a specific placement in the app. The SDK can then request bids and render ads based on defined parameters such as placement ID, ad format, and targeting settings. This allows for more precise alignment between in-app placements and the demand sources competing in the auction.
By improving the mapping between placements and demand, this enhancement helps drive more efficient auctions and more accurate reporting, giving publishers better visibility into performance across their inventory.
In addition, Ad Unit configuration introduces greater flexibility for monetization strategies. Publishers can now easily configure different ad units for formats such as interstitial, rewarded video, and banner, allowing them to tailor demand and optimization strategies for each placement.
Overall, this update provides publishers with more granular control, improved transparency, and new opportunities to maximize the value of their in-app inventory.
This post was written by Galith Arroche, Director of Strategic Planning at Start.io.
For more than a decade, games have dominated the mobile monetization equation. From puzzle hits to massive multiplayer titles, gaming consistently drove the largest share of in-app revenue.
But 2025 marked a turning point.
According to the Sensor Tower State of Mobile 2026 report, consumers spent more money in non-game apps than in games for the first time last year. The shift signals something deeper than a single year of market fluctuation. It reflects a fundamental evolution in how people use their phones and where developers are finding new revenue opportunities.
Apps Overtake Games
According to the report, consumers spent approximately $85.6 billion in non-game apps in 2025, a 21% year-over-year increase. Just five years ago, spending in non-game apps was less than half that amount. Today, it has grown to 2.8 times its size in 2020.
By comparison, spending in mobile games reached $81.8 billion in 2025, representing only 1% year-over-year growth.
The difference between the two categories is striking. While games remain a massive and resilient segment of the mobile economy, revenue growth is increasingly happening elsewhere.
Generative AI played a major role in accelerating this shift. Sensor Tower estimates that GenAI apps alone added $3.5 billion in revenue growth between 2024 and 2025. But AI is only part of the story. Other categories also expanded rapidly, including streaming platforms and social apps. For example, apps focused on movies and television content generated $2.2 billion in additional revenue growth year over year.
In short, the mobile ecosystem is diversifying. The center of gravity is moving beyond gaming into everyday utility, creativity, and entertainment.
Engagement Is Expanding Beyond Gaming
Revenue is only part of the story. The same report shows that mobile engagement is also growing within non-game experiences.
According to Sensor Tower, consumers spent nearly 2.5 trillion hours on social media apps across iOS and Android in 2025, representing a 5% year-over-year increase. For the average mobile user, that translates to more than 90 minutes per day spent on social platforms alone.
Social messaging apps continue to command enormous attention as well, with 901 billion hours spent globally, while browsers accounted for 477 billion hours. OTT streaming apps generated more than 100 billion hours of user engagement during the year.
By comparison, worldwide mobile users spent 444.6 billion hours playing mobile games in 2025, an increase of just 0.9% year over year.
In other words, the categories attracting the most attention on mobile today are overwhelmingly non-game experiences.
AI is also rapidly becoming a larger part of that time-spent mix. Sensor Tower reports that gen AI assistant apps ranked among the top 10 subgenres globally by time spent in 2025, with 37.4 billion hours spent, representing a 426% year-over-year increase. AI companion apps, while still emerging, are gaining traction as well, recording 68% growth in time spent.
Taken together, these trends help explain why monetization patterns are shifting. As users spend more time across social platforms, streaming services, productivity tools, and AI assistants, those categories are naturally capturing a larger share of mobile spending. These shifts also have implications for how brands should be thinking about their mobile ad spending.
What Start.io’s Data Shows
Start.io’s own mobile audience insights reflect similar momentum across non-game categories. Analysis of app download data shows significant year-over-year growth in several non-gaming verticals during 2025, including:
- Sports apps: +31%
- Food and drink apps: +24%
- Books and reference apps: +17%
- Entertainment apps: +13%
These trends reinforce the idea that mobile users are spending more time engaging with apps that help them follow passions, manage daily life, and consume content.
The rise of sports apps, for example, reflects the growing habit of second-screen viewing and real-time engagement around live events. Food and drink apps continue to benefit from delivery services, restaurant discovery, and digital loyalty programs. Meanwhile, books, reference, and entertainment apps are capturing audiences looking for education and streaming content directly on their phones.
In other words, the smartphone is becoming less of a gaming device and more of a hub for daily experiences. This shift opens new opportunities for developers and advertisers alike.
A New Phase for the App Economy
Games are not disappearing from the mobile ecosystem. They remain a powerful driver of engagement and revenue.
But the mobile economy is clearly entering a new phase. Growth is now coming from a broader mix of categories, from AI assistants to streaming platforms to everyday utility apps.
For advertisers and developers, the takeaway is simple: If you’re focusing solely on games with your mobile plans, you’re missing a growing opportunity within other categories. The platforms capturing attention today are the ones embedded in users’ daily lives.
Julien Gardès, VP International Exchange & Business Development, at TripleLift
🎧 Audio version is available exclusively on Spotify:
Briefly introduce yourself, your role, and the pivotal moment in your career that led you to AdTech.
I’m Julien Gardès, and I have been with TripleLift for six years. My current role is VP of International Exchange Optimization & Strategic Projects. This means I oversee all supply-side aspects of new business, including publisher development, open auction optimization, and curation. The pivotal moment that led me to adtech was a desire for deeper technical understanding. I started my career in online advertising at a performance network, which was very media-oriented. However, I was frustrated by my inability to technically comprehend how things worked in the backend. Seeking a new challenge, I decided to move into a more technical field and subsequently worked for an ad server, so that’s how I ended up working in adtech.
What is an underrated tool, habit, or soft skill that has become indispensable to your daily work?
I strive to maintain a zero-inbox policy, as a cluttered email inbox can be a significant source of stress for me. To achieve this goal, it is essential that I remain highly organized and ensure timely delivery of all tasks. The key is to address and act upon items promptly, avoiding the tendency to leave things untouched for an extended period.
What is one common misconception about programmatic or AdTech that you find yourself constantly correcting?
For too long, I’ve had to counter the persistent myth that programmatic advertising is only applicable to low-quality inventory. This is simply not true. Programmatic technology has evolved significantly, allowing for real-time transaction of campaigns across premium, high-quality, and even shoppable inventory. It’s crucial to continuously educate the market and demonstrate that programmatic is a sophisticated, efficient, and effective channel for all levels of advertising investment.
What are your platform’s three biggest priorities right now, and why do they matter for the broader ecosystem?
Our three biggest priorities right now, and why they matter, are:
- Deliver unmatched performance through creative and custom solutions. This is crucial for driving measurable results for our advertisers and ensuring the value of our platform.
- Operate smarter, faster, and stronger through AI and Automation. We’re investing heavily in building the first agentic creative SSP, through TL Spark – our agentic intelligence layer that interprets campaign goals, orchestrates decisions across the ecosystem, and continuously reallocates resources toward measurable outcomes. This efficiency and intelligence are vital for scaling and maintaining a competitive edge.
- Lead the industry’s next-generation exchange by creating an infrastructure focused on signal sophistication and yield optimization. This will be key in combating open-web headwinds and ensuring a sustainable, high-quality exchange for publishers and advertisers alike.
Based on your experience, what is the #1 “best practice” most companies overlook when trying to succeed in programmatic?
I’m very pragmatic and for me it’s all about having common sense. Sometimes, people tend to overcomplicate concepts that are fundamentally simple. Ultimately, the core purpose of an SSP like TripleLift is to efficiently connect a buyer and a seller, ensuring that both parties’ criteria are mutually satisfied—a foundational truth we must always remember. It often feels as though we’ve allowed programmatic to become synonymous with overly complex ideas, which simply isn’t true.
Without giving away trade secrets, can you share a recent project or case study that highlights a creative solution to a tough AdTech problem?
Brands are becoming much more sophisticated in monetising their first-party inventory, but the real opportunity lies in recognising that on-site revenue is just the start. To truly scale their advertising businesses, brands must leverage off-site media. This means using valuable customer browsing and buying signals to drive measurable outcomes across the open web.
A recent example is how Philips Hue sought to expand its reach beyond Amazon and connect with New-to-Brand customers on the open internet, with the support of digital marketing agency Numberly. Numberly selected TripleLift for its advanced creative capabilities and access to premium publishers to ensure a seamless, premium ad experience consistent with the brand. Leveraging the Amazon Ads Native Responsive eCommerce format, combined with operational simplicity and a premium advertising experience, Philips Hue achieved 70% higher CTR, 60% new-to-brand sales, and 41% unique campaign reach.
AI Beyond the Hype: We know AI is everywhere. Where specifically do you see it providing the most practical value for identity and targeting today?
AI is already helping us to improve campaign insights, planning, execution and performance. It’s the engine that pulls the whole train. AI refines the targeting, optimizes the content delivery and helps tailor the creative. It’s making every single step more precise, faster, and easier to execute. When it comes to identity and targeting, AI provides the most practical value in helping create look-alike audiences from first- and third-party data insights, and optimizing campaign performance by identifying key behavioral patterns and predicting future user actions.
Which specific niche or sub-sector of AdTech would you bet on for the most rapid growth right now?
Attention is a critical factor for measuring reliable campaign outcomes, and attention-based marketing is gaining significant traction in the industry. As a result, campaigns are increasingly being optimized on both viewability and attention metrics. This represents an evolution in campaign strategy, as previously, optimization was primarily focused on viewability alone. The current approach involves a simultaneous focus on these two combined metrics to drive better performance.
For a recent Vodafone UK Home Broadband campaign, TripleLift used Amazon DSP and curated purchase intent data to target high-propensity switchers, driving awareness and engagement without distracting from surrounding content. The campaign surpassed benchmarks, delivering a lower CPM than Vodafone’s historical programmatic buys. An eye-tracking study confirmed high performance: 96% of those exposed saw the ad, which was noticed 3x faster than a standard display ad, with an average view time of 2.14 seconds, indicating strong engagement.
What is a “hidden gem” opportunity currently opening up for advertisers or publishers that most people aren’t talking about yet?
We’re noticing a strategic shift towards sell-side decisioning that moves the targeting intelligence from the buyer to the publisher, making the publisher’s first-party data and resulting curated segments highly valuable assets for targeted advertising initiatives.The TripleLift platform offers significant value by enabling publishers to leverage their first-party data within curation campaigns.
Share one book, podcast, or article that has fundamentally shifted your perspective on business or technology recently.
When I was a teenager, I got asked by a teacher to read “Zadig, or the book of fate” from French philosopher Voltaire. This book taught me that nothing comes to you by surprise and that whatever the journey is, you need to own it. Its most famous quote is “One Must Cultivate One’s Own Garden” which has always been my mantra. Learning, being curious, asking questions is a never ending discipline I try to apply myself everyday.
This post was written by Omri Barnes, Chief Marketing Officer at Start.io.
In the ever-escalating coffee wars, it’s no longer just Starbucks vs. Dunkin’. Challenger brands are multiplying, each with a distinct identity and growth playbook. But who they are and why they’re gaining ground depends heavily on where you are.
In New York City, Blank Street has transformed from a 5×10 cart in Brooklyn into a mint-green cultural signal flare for Gen Z. Small footprints, smart tech, and drinks engineered for TikTok have helped it expand across NYC, London, Boston, and DC. Now Luckin Coffee, the Chinese giant with more than 20,000 stores globally,( soon to be more, if its acquisition of Blue Bottle goes through), has entered the U.S. market with its first New York locations, signaling that price-driven competition is heating up in urban cores. Similarly, Cotti Coffee, another value-focused Chinese chain, landed in Hawaii in 2024 and has since expanded to California and now New York.
Meanwhile, in less flashy but equally coffee-crazed markets, Scooter’s Coffee is quietly taking root. The drive-thru specialist now operates more than 900 stores across 32 states, with its highest concentration in Nebraska, where it has 126 locations. Its model is simple: convenience, consistency, and speed.
So what separates these markets? Start.io’s mobile-first coffee enthusiast segments reveal that the answer is not just taste, but demographics.
A Nationwide Snapshot: Young and Balanced
Across the U.S., Start.io identifies more than 10.4 million coffee enthusiasts within a population of roughly 329 million. Nationally, this audience skews young, with 50.9% falling into the 18–24 age bracket. Gender splits are almost perfectly even at 50% male and 50% female.
This is a highly mobile, digitally fluent group. From an OS perspective, roughly 63.5% use Android devices versus 36.5% on iOS. Income distribution is relatively broad, with 34.1% earning under $25,000 and 14.6% in the $100,000–$149,999 range. In short, the U.S. coffee enthusiast is young, balanced, and economically diverse.
For national brands or chains expanding across states, this profile suggests scale opportunities among Gen Z and young millennials, without leaning too heavily into a single gender or income cohort.
New York City: Urban, Male-Skewing, and Bifurcated by Income
In New York City, Start.io identifies 111,445 coffee enthusiasts within a population of just over 8.3 million. The demographic makeup shifts meaningfully.
Here, 43% of coffee enthusiasts are 25–34, and 42.1% are 18–24, making the audience slightly older than the national average. The gender split skews heavily male at 62.7%, compared to 37.3% female.
Income data tells an especially interesting story. About 33.2% of NYC coffee enthusiasts earn less than $25,000, while 20.1% fall into the $100,000–$149,999 range. That kind of polarization reflects the city itself: students and early-career workers alongside high-earning professionals.
For brands like Blank Street or Luckin, this mix rewards distinct value propositions. Trend-forward branding and culturally fluent marketing resonate with younger, urban consumers. At the same time, premium offerings or loyalty strategies can appeal to higher-income segments looking for convenience without sacrificing quality.
Nebraska: Younger, More Evenly Distributed, and Convenience-Driven
In Nebraska, Start.io identifies 38,087 coffee enthusiasts within a population of roughly 1.9 million. The audience skews even younger than the national average, with 56.8% in the 18–24 category.
Gender tilts slightly female at 53.1%, and income distribution is more evenly spread across brackets compared to NYC. While 27.1% earn under $25,000, meaningful portions fall into mid-tier income categories, creating a less polarized economic picture.
This aligns closely with Scooter’s growth model. In markets where driving is a daily necessity and convenience is paramount, drive-thru formats and straightforward value propositions thrive. Messaging around speed, reliability, and everyday indulgence can outperform trend-driven hype.
Why Granular Audience Data Matters
Coffee enthusiasts are notoriously on the go. They move between neighborhoods, campuses, offices, and retail hubs, often multiple times a day. Reaching them effectively requires more than broad demographic assumptions.
Start.io’s Consumer Insights and Audiences Hub enables marketers to explore unlimited locations and segments, uncovering differences at the national, state, and city level. With tens of thousands of mobile-first segments and real-world behavioral insights, brands can move beyond generic targeting and activate campaigns instantly across leading DSPs.
The coffee wars may be global, but growth is local. And in a category this competitive, knowing exactly who your enthusiasts are, and how they differ by geography, can be the difference between blending in and standing out.
If you’re interested in learning more about Start.io audiences, email us at marketing@start.io or visit here.
This post was written by Daniela Shikhmakher, Data Engineer at Start.io.
LLMs are reshaping how organizations process and leverage unstructured text data. From intelligent chatbots to automated content generation and advanced analytics, LLMs are rapidly becoming fundamental components of modern data stacks. For data engineering teams, LLMs create a major opportunity – they allow us to reduce human-in-the-loop workflows and automate complex semantic tasks. On the other hand, we realize that the challenge of building production AI systems is not just about calling a model API, it’s about engineering repeatable and reliable infrastructure.
This article describes how we evolved from a single LLM-based classification project into building our own internal utils package and where generic abstraction helped us, and where it nearly slowed us down.
The Initial Use Case: A Single Classification System
We began with a focused goal: develop an LLM-based classification framework, which will automatically classify daily ingested data into internal audience segmentation taxonomies.
The architecture included:
- Data processing layer
- Caching mechanisms
- RAG (Retrieval-Augmented Generation) component
- Output parsing and validation
At first, everything was in a single repository with one configuration and one business problem, and it worked well.
The Pattern Emerges
Once there were more classification use cases, each project had different taxonomies, label data, business rules, and evaluation metrics. But they still have the same core requirements that demand a solution by an A LLM-based framework.
The Next Level of Abstraction
There are already tools that abstract across models and providers, such as LiteLLM and Amazon Bedrock. These tools address model abstraction by providing a unified interface for interacting with multiple LLMs. This allows developers to switch models or providers without rewriting large portions of their integration code. Our problem was different – We needed to abstract across classification systems.
The Decision: Build a Reusable Utilities Layer
We decided to build our own utilities package not to create a full framework, but to avoid repeating the same setup code across classification tasks. Before starting development, we asked what stays consistent across projects and decided to generalize only those stable parts.
What We Chose to Generalize
We identified reusable components that consistently appeared in every classification system.
1. Classification Mechanics – we needed to support 2 types of classification: single-class and multi-class.
2. Prompt Management – across projects, we required:
- System / user / assistant message formatting
- Dynamic few-shot / one-shot / zero-shot
- Injection into the user prompt
- Controlled prompt construction patterns
While prompt content varies by domain, the mechanics of building prompts are consistent.
3. LLM Response Parsing – we repeatedly implemented:
- Extracting structured classification from raw LLM responses
- Parsing single-class outputs (e.g., returning a single label ID)
- Parsing multi-class outputs (e.g., structured JSON objects)
- Result selection strategies for handling multiple classifications
- Output validation
Reliable structured output, like valid JSON or clean label IDs, is essential for all projects and should exist in a utils layer.
4. Cost Estimation and Observability – every project required cost and token usage tracking.
What We Explicitly Did Not Generalize
Equally important was defining what should remain at the project level.
We did not abstract:
- Label loading logic (each project sources labels differently)
- Classification taxonomies (categories, hierarchies, segmentation rules)
- Domain-specific workflows (offline batch processing vs. real-time applications)
These components evolve with business needs. Freezing them behind a generic interface would reduce flexibility.
When the Generic Code Started to Hurt
The first version of our utilities layer was helpful. Then we pushed abstraction too far.
- Configuration Explosion
We attempted to make everything configurable. The configuration object grew increasingly complex. Eventually, only the developers who built the system fully understood how to configure it correctly. When configuration becomes a domain language of its own, abstraction has gone too far. Instead of simplifying development, it increased the complexity. - Model and API Volatility
Between projects, new LLM models appeared, APIs evolved, and versions changed. When the abstraction layer is too generic, this volatility hits hard: frequent model and API changes break the shared code, force constant updates, and make the system fragile and difficult to maintain. - Each Task Needs Its Own Metrics
We first built one generic Test Grader, but it was brittle and hard to maintain because single-class and multi-class projects use very different metrics, and the evaluation also depends on the specific task, making one grader too complicated for all cases.
The Real Cost of Over-Generic Code
Over-abstraction can create real engineering problems. As our utilities layer became more generic, even small changes required navigating multiple layers of code, making debugging less transparent and demanding deep internal knowledge to understand system behavior. What was meant to speed up development ended up slowing it down, adding a minor feature felt heavier than it should have. In this case, generic code became a constraint rather than an enabler.
Our Solution
To solve the problems caused by over-generic code, we adopted a clear architectural principle: abstract the infrastructure and keep business logic concrete. Our shared utilities layer focuses only on stable, reusable components LLM interaction, prompt formatting, output parsing, multi-class handling, cost estimation, and logging. Everything that varies by project – like taxonomy definitions, domain workflows, and evaluation methods remains in the application itself. This separation allows us to reuse common infrastructure without forcing rigid, one-size-fits-all solutions, keeping each project flexible and maintainable.
Tools like LiteLLM and Amazon Bedrock make it easy to switch models and handle API differences. Our utils package solves classification framework reuse.
Both are abstraction layers – but at different levels. One operates at the model provider level. The other operates at the application infrastructure level.
Recognizing the correct abstraction axis is critical.
Abstract the wrong layer, and you build unnecessary complexity.
Abstract the right layer, and you eliminate duplication without sacrificing agility.
Conclusion
LLMs are transforming how organizations use unstructured data, but building LLM-based framework is not just about models – it’s about sustainable infrastructure. Just as tools like LiteLLM make it easy to switch between models, our utilities package lets us switch between classification tasks easily, without rewriting common components. Abstraction must be intentional: too little leads to duplication, too much slows innovation. The key is focusing on what’s truly stable and resisting the urge to generalize everything else.
Global programmatic ad spend reached $642 billion in 2025 and is expected to grow to nearly $800 billion by 2028. In the U.S., digital ad spending long ago surpassed traditional channels, with the vast majority of that spend going into mobile specifically.
While programmatic is still growing, the “how” is changing. The industry is moving away from an ecosystem optimized primarily for scale toward one optimized for quality, control, and adaptability. These five trends are shaping that shift:
1. Curation moves from “nice to have” to a default buying layer.
Curation is increasingly the mechanism buyers use to reduce waste, consolidate supply paths, and create more intentional media portfolios. Instead of buying inventory as a broad pool, curated marketplaces pair audience data with specific, transparent supply relationships. The value proposition is straightforward: fewer redundant paths, fewer unknowns, and more control over where budgets flow.
As more advertisers pressure the supply chain to prove its value, curated supply packages are becoming a practical answer to long-standing programmatic inefficiencies.
2. AI expands from optimization to segmentation and supply packaging.
AI’s role in programmatic continues to broaden, moving beyond bid optimization into audience construction, content-based segmentation, and smarter supply packaging. AI-driven approaches are increasingly used to interpret signals at scale and translate them into actionable segments and curated buying opportunities, particularly in mobile and omnichannel environments.
In other words: AI is not only helping decide what to bid. It is also shaping what “inventory quality” and “audience relevance” mean in the first place.
3. Measurement remains complicated, and expectations are recalibrating,
Despite robust programmatic tooling, holistic campaign measurement is still constrained by real-world realities: inconsistent identity signals, fragmented platforms, walled gardens, multi-touch journeys, and the growing share of impressions that do not produce a neat click-path to a conversion.
In 2026, sophisticated buyers are increasingly separating what they want to measure from what is realistically measurable, and then designing testing frameworks that match those constraints. The winners will be advertisers who treat measurement as an operating discipline, not a dashboard.
4. AI search reshapes format investments, creating new inventory gravity.
As AI search and conversational interfaces reduce click-through behavior on the open web, advertisers will rebalance investments toward environments where attention is still deep and measurable, including mobile, video, and in-app experiences. At the same time, new opportunities will emerge inside AI assistants and AI-powered search experiences, likely mixing sponsorship, commerce actions, and recommendation-style placements.
This trend matters for programmatic because it changes where demand pressure accumulates. If fewer sessions begin and end on publisher pages, more competition will concentrate in channels that still offer high-intent and high-attention moments, and the industry will need new norms for what “good inventory” looks like in answer-first experiences.
5. Agentic buying changes execution, but it does not erase strategy.
Agentic workflows are starting to influence how programmatic campaigns are planned, optimized, and governed. Protocols and tools built for agentic demand aim to make buying more automated, but the big shift is not “hands-off marketing.” It is faster iteration and more adaptive decisioning within guardrails.
In practice, agentic buying is likely to accelerate bid and budget adjustments, reduce time-to-test, and make optimization feel more continuous. But advertisers will still need clear goals, clear constraints, and clear definitions of “good outcomes,” especially as inventory quality and measurement remain uneven across channels. The IAB’s release of its agentic roadmap was an important step toward helping agentic buying to scale with the establishment of standards.
Want to learn more about the evolving role of programmatic advertising in the modern digital ecosystem? Download our comprehensive “2026 Programmatic Advertising Guide for Media Buyers” here.
To learn more about how Start.io can help supercharge your programmatic advertising, contact us today at marketing@start.io.
Nathalie Liscia Bellaiche, Director Publishers Partnership Southern Europe/MEA at Criteo
🎧 Audio version is available exclusively on Spotify:
What is one common misconception about programmatic or AdTech that you find yourself constantly correcting?
A persistent myth in the industry – and one I actively challenge – is the belief that programmatic inevitably leads to standardization and a loss of control for publishers. In reality, programmatic does not commoditize inventory by nature- it only does so when implemented without strategy or governance.
Programmatic is complex and cannot be improvised: it’s not just a matter of simply “plugging in pipes,” but of building and maintaining real expertise. To be effective, a publisher needs to activate their first-party data and audiences, strategically manage the inventory across open auctions, private deals, and guaranteed programmatic, and measure performance with precision.
When this myth persists, publishers often underinvest in expertise and governance, which directly limits their ability to differentiate and create value.
When approached with intent and discipline, programmatic becomes a powerful lever for control, transparency, and value creation, enabling publishers to build differentiated and sustainable monetization models.
What are Criteo’s three biggest priorities right now, and why do they matter for the broader ecosystem?
Criteo has undergone a profound transformation, evolving from a single-product retargeting company into a global player operating across channels and devices. Today, Criteo is helping shape the Commerce Media landscape by connecting brands, retailers, publishers, and consumers through data-driven, privacy-first solutions.
In this ecosystem, the real priority is to restore meaning and value for every stakeholder. Technology should serve concrete business outcomes — not just campaign or click optimization, but real, incremental value creation.
It is critical that all parties benefit fairly from the ecosystem, while putting the consumer, the customer experience, and the purchase journey at the center of everything we build. Data plays a central role in this strategy, particularly commerce data, which enables relevance, measurement, and trust across the entire value chain. Measuring true incrementality is essential to ensure that performance reflects real business impact rather than surface-level efficiency.
Based on your experience, what is the #1 “best practice” most companies overlook when trying to succeed in programmatic?
One of the most important lessons I’ve learned over the years is that sustainable success in programmatic is driven by collaboration, not by isolated optimization from each individual player. Too often, stakeholders focus solely on their own KPIs, which ultimately diminishes the overall value of the ecosystem.
Programmatic works best when it is treated as a shared operating system rather than a zero-sum marketplace. True performance emerges when publishers, platforms, and advertisers adopt shared standards and unified practices that break down silos and enable interoperability, transparency, and greater operational efficiency. This requires ongoing dialogue and a willingness to co-build, rather than simply transact.
Another critical, yet often underestimated, factor is data quality and granularity. Activating the right data in a responsible and transparent way—while respecting the purchase journey and prioritizing relevance over volume—is essential. Combined with clear, granular reporting, this enables more strategic campaign optimization.
When collaboration, standards, data activation, and transparency come together, programmatic moves beyond being a transactional tool and becomes a true growth driver.
Please share a recent project or case study that highlights a creative solution to a tough AdTech problem?
Recently, we have explored how AI can transform advertising creation in a concrete and measurable way. We leverage generative AI tools to create and enhance visual and interactive content at scale.
For example, we have combined several techniques: text-to-image generation to produce compelling visuals, background enhancement and expansion to adapt visuals to different contexts, and the creation of personalized recommendations directly integrated into the creatives. This approach enables the production of advertising assets that are more relevant, engaging, and quickly deployable, while meeting the specific needs of each audience.
This has also helped reduce time-to-market while improving engagement and contextual relevance. In short, AI has become a true driver of creativity, capable of amplifying the impact of advertising messages and boosting audience engagement.
We know AI is everywhere. Where specifically do you see it providing the most practical value for identity and targeting today?
Today, the most practical value of AI for identity and targeting lies in its ability to identify and target consumer intent. Rather than simply capturing attention, AI can detect, from subtle and varied signals, where a user is in their purchase journey and deliver the right message, at the right time, on the right channel — all while respecting privacy. This approach makes targeting more precise, useful, and outcome-driven, rather than purely impression-based.
AI also plays a critical role in creative activation. It does not replace human creativity, but amplifies it at scale: by adapting messages, formats, and visuals to context, timing, and audience, AI enables relevant, engaging, and less intrusive experiences. Brands can deliver hyper-personalized content or conversational formats that strengthen relationships with their audiences.
Strong governance and human oversight remain essential to ensure these systems are deployed responsibly.
Finally, the rise of agentic AI provides a complementary lever for product discovery and decision-making. It does not replace existing channels — search, social, retail media, or the open web — but enhances them, guiding consumers across an increasingly non-linear journey. In this environment, the quality of commerce and intent data becomes critical: these signals allow AI systems to make reliable, relevant, and contextually appropriate recommendations at every stage of the journey.
Which specific niche or sub-sector of AdTech would you bet on for the most rapid growth right now?
Rather than betting on a specific niche or sub-sector, I would focus today on programmatic at a global scale. If I had to choose a growth engine right now, it would be omnichannel programmatic. We have been talking about 360° strategies for years, but they have rarely been fully operational: touchpoints remain siloed, activation and measurement are not fully synchronized, and it is difficult to connect each interaction to real business value for advertisers.
What is changing now is not the vision, but the conditions to finally execute it: stronger first-party and commerce data, improved cross-channel measurement, and AI-driven orchestration make true omnichannel activation achievable rather than aspirational.
The real turning point will come when we finally succeed in aligning all channels — social, CTV, retail media, and open web — around a coherent approach, where data, targeting, and measurement work together seamlessly. When this vision becomes reality, we move beyond simple impression or click arbitrage and deliver performance that is truly measurable across the consumer journey, while providing greater transparency and value for all ecosystem participants.
Omnichannel is not just a strategy — today, it is the most powerful lever to accelerate programmatic growth at a global scale.
The Influencer: Share one book, podcast, or article that has fundamentally shifted your perspective on business or technology recently.
Recently, rather than focusing on a single book or podcast, I’ve been deliberately broadening my perspective by exploring content across the entire technology and AI landscape — not just within digital advertising, media, or adtech.
What’s becoming increasingly clear is that we are at a true inflection point. The advances we’re seeing in AI feel less like an incremental evolution and more like a clear “before and after” for business and technology.
Following voices across research, startups, and policy has shifted my mindset from optimizing within an ecosystem to questioning it altogether. The most valuable insights come from connecting dots across disciplines, because AI’s impact will be systemic, not siloed.
Is there anything else you’d like to share?
Maybe just one thing. After many years in this industry, what still motivates me most is learning, especially from people with different perspectives. AdTech moves fast, and none of us has all the answers. Staying curious, listening carefully, and being open to questioning our own certainties has been essential for me, and I think it will matter even more in the years ahead.
