Open Banking has moved from compliance to commercialisation, with PSD2 establishing an important baseline. According to Stephen Whitehouse, head of payments for Europe at Oliver Wyman, PSD3 and the Financial Data Access Act (FiDA) will further drive this movement, increasing the scope of data involved and thereby enabling new business models and use cases.

​ 

Attribution in the modern marketing age can be confusing. But the pressure on marketing teams to “prove what’s working” never goes away. 

Traditionally, marketers had certain data we could always rely on, but the data pool we can pull from seems to be growing and shrinking at the same time. Between privacy constraints, zero-click searches, AI Overviews, and channel-walled gardens, marketers are flying blind in more ways than they realize. Attribution has always been an imperfect science. And in 2025, it’s gone from fuzzy to fragmented.

If you’re planning marketing budgets and trying to defend where your spend is going, there’s no need to freak out. Marketing attribution is possible. It doesn’t look like it used to, though. And if you’re still only relying on touch-based models or last-click reports, you might be measuring the wrong things entirely.

Let’s break down where attribution is failing, what’s making it harder, and what forward-looking marketers are doing to close the gap.

Key Takeaway

  • Attribution challenges have multiplied due to AI, automation, and privacy shifts.
  • Walled gardens, offline sales, and dark social are major blind spots, and they often overlap.
  • Deterministic, touch-based attribution is giving way to modeled and probabilistic methods.
  • AI isn’t just the problem, it’s also part of the solution.
  • You don’t need perfect data. You need data that helps you make better decisions.

The New Face of Attribution

Attribution used to be about stitching together clicks. Now, we’re lucky if we get clicks at all thanks to zero-click search.

Today’s buyers bounce between different platforms on multiple devices and AI-curated content. They’re influenced by ads on a connected TV or product mentions in a ChatGPT thread, and neither of those leaves a clean digital trail.

Meanwhile, ad platforms like Meta and Google have leaned hard into automation. That means fewer transparent levers to optimize and more “black box” performance metrics. According to NP Digital analysis, there are over 90% fewer optimization permutations in Google and Meta Ads today compared to 2023. So yes, marketing attribution is back. But the infrastructure around it seems more broken than ever.

A graphic explaining the collapse of optimization levers.

Finding Marketing Blindspots

Unfortunately, the reality is that attribution blind spots don’t come with a warning light. You may be staring directly at your dashboard and not notice traffic is piling up in areas you’re not tracking. And the amount of potential blindspots is growing.

Here are the big ones:

  • Walled Gardens: Platforms like Google, Meta, and Amazon are all powerful, but have become much more mysterious as search evolves. You’re renting their space, but if you don’t play by their rules, you may not get complete visibility.
  • Offline Sales: Leads turn into deals in CRMs, call centers, or retail. They may have started as a click, but the customer journey ends at a brick-and-mortar location or an entirely different platform than the original click.
  • Cross-Device Journeys: That ad someone saw on mobile might convert from their phone, but they could just as easily become a sale on their desktop or smart TV.
  • Building Awareness: Upper funnel spend (like digital out-of-home (OOH) or video) gets undervalued because it rarely leads to a direct conversion.
  • Dark Social: Private sharing (think WhatsApp, SMS, Signal) shows up in attribution models as “direct”, but it’s not.
  • LLM Traffic: People are discovering brands via large language models, and those referrals are often invisible in GA4.

To make matters worse, these blind spots can stack. Before you know it, you find yourself in a nightmare marketing scenario where you’re not just missing one data signal, you’re missing combinations of them, making optimization even harder.

A graphic that explains how multiple marketing blindspots can pile up.

New Attribution Trends and Technology

You can keep up with all of this. It just requires a switch in perspective. Marketers should evaluate their campaigns using a combination of modeled attribution and traditional touch-based metrics. You may never fully connect every dot, and that’s okay. The goal isn’t perfection, just enough clarity to defend marketing budget allocations.

Modern marketers are using these tools:

  • Incrementality testing: Geo holdouts and lift studies to isolate what’s actually moving the needle.
  • MMM (Marketing Mix Modeling): Especially useful for larger budgets or mixed channel strategies.
  • Correlation analysis: Pre/post testing, contextual lift, and even proxy signals like brand search volume.
  • Unified first-party data: Clean, consistent CRM and web data feeding both your models and your platforms.

The best strategies blend these methods based on spend level, complexity, and conversion volume. Leveraging AI in your marketing efforts is one of the best ways to automate this research as much as possible and maximize the benefit of these tactics. 

AI and Blind Spots

Some marketers may feel like AI is eroding attribution. While that could be true, the technology is also helping to rebuild it.

Here’s how AI is stepping in:

  • Generative AI: LLMs like ChatGPT are now discovery platforms. They drive traffic, but don’t always identify themselves unless you tag them.
  • AI coworkers: Agentic AI simulates user behavior, tests messaging, and can even help set up GA4 tracking automatically.
  • Machine learning models: Used in MMMs and platform attribution to refine forecasts, assign contribution, and make predictions.

Still, only 55% of marketers trust AI-generated insights, according to CoSchedule. The key is to treat AI as an assistant, not the authority. Use it to speed up testing and build models, but validate with your own data.

A graphic that explains how to introduce GenAI into reporting workflows.

Analytics platforms like Adobe Analytics are also making steps to better capture attribution from AI tools. In October they released a new referrer type called “Conversational AI Tools” to segment out traffic from ChatGPT and other LLMs from the other channels marketers have historically monitored.

Closing The Gap With Attribution Strategies

So, how do you go from blind spots to better planning? You don’t need perfect clarity. You need consistent signals and a smarter strategy.

Here are some ways marketers are closing attribution gaps:

  1. Clean your first-party data: Data from internal sources like your website and CRM needs to be trustworthy. These are your most important sources of truth.
  2. Use multipliers: Adjust performance based on geo lift or experiment results. Not every click counts equally.
  3. Invite questions: Models are approximations. Encourage teams to challenge them and make improvements as time goes on.
  4. Survey your customers: Ask where they heard about you. It’s old school, but incredibly effective for context.
  5. Use offer codes and landing pages: Even if not perfect, they create new signals across dark social or offline.
  6. Track “AI Referrers”: Create custom =channels in your web analytics, including in GA4, to segment out performance from LLM-driven traffic.

Linking Attribution To Business Outcomes

Attribution and business outcomes go hand-in-hand. Understanding where your most profitable leads originate is essential to growing any business, regardless of its size.

A graphic explaining savings attributed to fixing attribution.

You want to connect your data to actual decisions, such as forecasts, budgets, and resource allocation. But, with the marketing landscape changing so quickly and drastically, how do you know which metrics to follow?

Here are the metrics that matter now:

  • Total conversions and incremental conversions
  • Conversion value over time
  • Cost per incremental conversion
  • Spend thresholds by tactic
  • Directional change (old model vs. new)

Remember: even if your models aren’t perfect, if they get you closer to optimal spend, it’s working. Continuous improvement for your attribution strategy will get you closer and closer still.

A graphic explaining the value of continuous improvement for marketing attribution.

FAQs

What is a marketing attribution blind spot?

It’s any part of the customer journey you can’t track, like dark social shares, offline sales, or LLM referrals that may be influencing conversions without showing up in your data.

Can AI help with attribution?

Yes, but only if used smartly. AI can simulate behavior and identify patterns, but it’s not a silver bullet. Use it to complement your experiments and first-party data.

What’s the best attribution model?

There isn’t one. The most effective models mix touch-based data with testing and contextual clues. Choose based on your business size, channel mix, and data maturity.

Conclusion

When it comes to effective attribution, you just need to see enough to move forward.

Mastering this skill in the modern marketing world is less about getting the credit right and more about making smarter calls with what you can measure. The key is to stop chasing perfection and start building a system that helps you plan and adapt to the data you gather from your testing in real-time. Attribution isn’t the whole picture, but it remains the best tool we have to illuminate the path forward, including its blind spots.

Naturally, we can still learn from tried and true marketing methods. We may just have to think outside the box on how to apply them to today’s search environment and customer journey. It’s worth checking out our guides on which marketing campaigns drive the best impact and how to track your marketing ROI. Combining this extra knowledge with your new attribution perspective could be the secret sauce to put you ahead of the pack in 2026. 

​