Measure what matters: how incrementality can double your ROAS

At Balkan eCommerce Summit 2026, Krzysztof Sobieszek delivered a highly practical session focused on one of the most important – and often misunderstood – topics in performance marketing: measurement and incrementality.

The presentation explored why traditional attribution models are becoming increasingly unreliable in an AI-driven advertising ecosystem and why eCommerce brands must rethink how they evaluate campaign performance, media efficiency, and real business impact.

Rather than focusing solely on platform-specific solutions, the session introduced a broader strategic framework for modern marketing measurement – combining attribution, experimentation, marketing mix modeling, and incrementality testing into a more holistic approach.

The central message was clear:

High-performing marketers do not simply optimize campaigns better – they measure business impact better.

Why Measurement Has Become More Important Than Ever

Krzysztof began by explaining that one of the key differences between average and top-performing agencies or marketers is their level of sophistication in measurement.

As advertising platforms increasingly automate campaign optimization through AI-driven systems, understanding what actually drives incremental business growth becomes far more critical.

Today’s advertising ecosystem is dominated by AI-powered campaign products such as:

• Advantage+ from Meta for Business
• Performance Max from Google Ads
• Automated optimization systems across marketplaces and retail media platforms

These systems are designed to maximize conversions at the lowest possible cost.

However, Krzysztof emphasized an important limitation:

AI algorithms optimize based on the signals they receive – not necessarily based on true business impact.

This creates a growing risk that advertisers may increasingly pay for conversions that would have happened even without advertising exposure.

The Core Problem: Paying for Customers You Would Have Acquired Anyway

One of the central themes of the presentation was the concept of incremental value.

Krzysztof revisited the well-known advertising quote:

“Half of the money I spend on advertising is wasted; the question is I don’t know which half.”

Despite being nearly a century old, he explained that the statement remains highly relevant in modern performance marketing.

The core challenge for eCommerce brands today is distinguishing between:

• Attributed conversions
  and
• Truly incremental conversions

A conversion may appear in reporting because a user clicked an ad, but that does not necessarily mean the ad caused the purchase.

Some customers:

• Already intended to buy
• Were already familiar with the brand
• Would have converted organically

In those cases, attribution systems may overstate the true impact of advertising activity.

As automation and AI optimization become more dominant, this problem becomes even more significant.

The Limitations of Last-Click Attribution

A major section of the session focused on the weaknesses of last-click attribution models.

According to research shared during the presentation:

• Nearly 80% of advertisers globally still rely primarily on last-click attribution
• Misattribution can lead to approximately 35% of ad spend being wasted

Last-click attribution tends to overvalue channels closest to conversion while undervaluing:

• Video engagement
• Awareness campaigns
• Discovery-driven interactions
• Upper-funnel influence

This issue is becoming even more relevant with younger consumer behavior patterns, particularly among Generation Zaudiences.

Krzysztof explained that Gen Z consumers are:

• Much more likely to discover products through video
• Less likely to click directly on ads
• More likely to convert after passive exposure and delayed consideration

As a result, traditional click-based measurement models increasingly fail to reflect real customer journeys.

What Incrementality Actually Means

The presentation then introduced the concept of incrementality in more detail.

Incrementality measures:
the actual additional business impact created by advertising.

To explain the concept, Krzysztof used a simple example:

• Two users see and click the same ad
• Both complete a purchase
• But one of them would have purchased even without the ad

In that case:

• Attribution systems count both conversions
• Incrementality recognizes only the conversion genuinely caused by advertising

The purpose of incrementality testing is to isolate true advertising impact from natural consumer behavior.

Control Groups & Randomized Experiments

To measure incrementality accurately, marketers use randomized experiments.

These experiments compare:

• An exposed group (users who see ads)
  with
• A control group (users who do not)

The difference in conversion behavior between the two groups reveals the true incremental effect generated by advertising.

Krzysztof emphasized that randomized experiments remain one of the most reliable ways to measure real advertising effectiveness.

However, he also acknowledged several practical limitations:

• Experiments are not always easy to execute
• They can be resource-intensive
• They are not always suitable for continuous optimization
• They should not exist in isolation

Instead, incrementality testing should become part of a broader measurement ecosystem.

Building a Modern Measurement Ecosystem

One of the strongest strategic takeaways from the session was the importance of combining multiple measurement methodologies together rather than relying on a single reporting model.

According to Krzysztof, effective modern measurement should integrate:

• Attribution models
• Multi-touch attribution
• Marketing Mix Modeling (MMM)
• Incrementality experiments

Each component serves a different purpose:

Attribution Models

Provide channel-level visibility and operational optimization insights.

Multi-Touch Attribution

Helps distribute conversion value across multiple touchpoints in the customer journey.

Marketing Mix Modeling (MMM)

Provides broader business-level insights and long-term media impact analysis.

Randomized Experiments

Validate true incrementality and calibrate the broader measurement system.

Together, these systems create a more accurate understanding of real business impact.

Calibration: Aligning Reporting With Reality

Krzysztof also introduced the concept of calibration between attribution systems and experimental results.

For example:

• Attribution reports 200 conversions
• Incrementality testing shows 300 truly incremental conversions

Brands can then apply weighting and calibration models to adjust reporting closer to reality.

This process helps organizations:

• Reduce misattribution
• Improve budget allocation
• Optimize media investment more accurately

According to the examples shared during the presentation, brands implementing this type of calibrated measurement system can achieve:

• Up to 30% higher ROI
  through improved decision-making and more accurate attribution.

Experimentation Should Become a Continuous Process

A particularly important recommendation was that experimentation should not be treated as an occasional activity.

Instead, companies should build:

• Structured testing roadmaps
• Ongoing experimentation programs
• Annual testing plans aligned with media strategy

Krzysztof explained that top-performing advertisers increasingly plan:

• Media investment
  and
• Measurement experiments

as parallel strategic processes.

Organizations running more experiments consistently tend to achieve significantly stronger performance outcomes.

According to the data shared during the presentation:

• Companies actively conducting experiments are approximately 20% more effective overall.

Case Study: Beliani

One of the strongest examples presented was the transformation journey of Beliani.

The company evolved its measurement approach through several stages:

1. Moving beyond last-click attribution
2. Implementing multi-touch attribution
3. Adding Marketing Mix Modeling
4. Introducing randomized incrementality experiments

The analysis revealed that:

• 64% of conversions had previously been misattributed

After recalibrating their measurement ecosystem:

• Meta ROI increased by 20%
• Advertising investment doubled
• Overall business growth increased by 25%

The case study demonstrated how improved measurement can directly influence both profitability and scaling decisions.

The Future of Actionable Measurement

The session concluded with a look at the future of measurement and optimization inside AI-driven advertising ecosystems.

Krzysztof explained that platforms such as Meta are increasingly working toward:

• Conversion Lift solutions
• Value Optimization
• Profit Optimization
• Order-level profitability optimization

The next evolution will allow brands to optimize campaigns not only around:

• Conversion volume
  but also around:
• Profitability
• Customer value
• Incremental business outcomes

Future systems will increasingly integrate:

• External analytics
• Custom business events
• Profit data
• Internal attribution signals

directly into campaign optimization systems.

This will make measurement not only more accurate, but also directly actionable inside advertising algorithms themselves.

Key Takeaways from the Session

The presentation reinforced several critical principles for modern eCommerce marketers:

• AI-driven advertising increases the importance of proper measurement
• Attribution alone is no longer sufficient
• Last-click reporting creates major blind spots
• Incrementality reveals real business impact
• Experimentation should become a continuous strategic function
• Measurement systems must combine attribution, modeling, and experiments together
• Better measurement directly improves ROI and scaling decisions

Perhaps the strongest insight from the session was that while AI is rapidly automating campaign execution, human understanding of measurement and business impact remains a major competitive advantage.

As Krzysztof emphasized throughout the presentation:
marketing success is not only about generating conversions – it is about understanding which conversions truly matter.

The presentation is not available for sharing.