AI has a data problem. Not a technology problem.

AI is rapidly changing the way marketers analyse performance. The way marketers are analysing performance is looking...different in the age of AI. Campaign reports used to takedays to compile, now they can be summarised in seconds. Patterns can be detected across thousands of metrics while recommendations can be generated at a scale that would have been unimaginable a few years ago.

Yet despite all this progress, there is a growing risk that marketing teams become more confident in their decisions without becoming any better informed. The problem is no longer the technology – it’s the context.

The danger of credible answers

We all know that today's AI tools are remarkably capable. The new rising challenge isn’t that they can’t analyse data, it’s that they will analyse whatever data they are given. Most organisations are not accustomed to questioning large volumes of data critically, resulting in any datasets containing gaps, inconsistencies or missing business context. When that data is presented to an AI model, it’s unlikely to stop and say, "I do not have enough information to answer this."

Instead, it will do the best job it can with the information available and the answer generated is often very convincing. The answer may appear insightful,supported by charts, explanations and recommendations. It may even pass an initial sense check but if critical inputs are missing, the conclusion can still be fundamentally flawed.

This is where businesses can get caught out. The challenge used to be producing analysis, now it’s knowing whether that analysis should be trusted. As AI becomes more accessible, competitive advantage will not come from generating more recommendations. It will come from ensuring they are grounded in the right context.

Data tells you what happened. Context tells you what matters.

AI will explain the data it sees, but what it cannot automatically understand is what the business is trying to achieve.

Consider a simple question: "How do I increase average order value?". An AI model might reasonably suggest increasing prices by 10%. On the surface, that recommendation makes perfect sense since average order value would almost certainly increase in theory.

However, if the business operates in a highly price-sensitive market, the outcome could be lower conversion rates and lower total revenue. The recommendation is logical and data-driven but the recommendation may also be completely wrong. The missing ingredient is context.This understanding that a business competes heavily on price is not necessarily visible in a performance dashboard. Neither are commercial constraints, strategic priorities, customer expectations or market dynamics. Yet these factors fundamentally change what constitutes a good recommendation.

If organisations want AI to move beyond reporting and towards meaningful decision support, they must provide more than data. They must provide the commercial reality that surrounds that data.

Why metrics need meaning

Many conversations about AI focus on data access. If only the model can access enough reports, dashboards and metrics, surely it can deliver valuable insight. Unfortunately, it is not that simple. Metrics are not universal. A conversion rate that represents outstanding performance for one organisation may represent poor performance for another.

One business may care primarily about brand awareness. Another may care about ecommerce sales. A third may be focused on lead generation. Even when organisations use the same terminology, they are often measuring different things. 

What constitutes a conversion? What defines engagement?Which metrics genuinely matter? Which should be ignored? The answers vary significantly between businesses. In some cases, an organisation that sells products online may discover that catalogue requests, sample requests or consultation bookings are more valuable than online purchases. Looking only at eCommerce transactions would provide an incomplete picture of success.

This is where many AI initiatives struggle. Making metrics available is relatively easy, while making metrics understandable is considerably harder. For AI to analyse performance reliably, it needs to understand not only the numbers themselves, but also the business meaning behind those numbers.

Without that shared understanding, organisations risk receiving recommendations that sound intelligent but are based on assumptions that do not apply to their business.

The limits of channel thinking

Marketing has become increasingly specialised over the past two decades. SEO specialists optimise search visibility. PPC teams focus on paid acquisition. Social teams build engagement. Email marketers drive retention. Conversion optimisation teams improve onsite performance. Often these disciplines sit in separate teams, even in separate agencies.

The challenge is that customers do not experience marketing through channels - they experience a journey. When performance is measured in channel silos, it becomes very easy to optimise one outcome at the expense of another. Take the relationship between SEO and paid search: if organic visibility improves significantly for key terms, should paid advertising continue targeting the same keywords?

In some cases, absolutely. In others, budget could be redirected elsewhere. The correct answer depends on understanding how the channels interact. The same is true when increasing advertising spend. More investment may generate more traffic but lower conversion rates because a broader audience is being reached.

Viewed in isolation, a declining conversion rate appears concerning but viewed in context, it may indicate that an acquisition strategy is working exactly as intended. This is why cross-channel visibility matters. Without it, organisations can easily misinterpret performance. With it, they can understand the trade-offs and relationships that drive real business outcomes. AI can help uncover those relationships, but only if it has access to a coherent view of performance across the entire marketing ecosystem.

Why benchmarking changes the conversation

Even organisations with strong data foundations have limited visibility of what good performance looks like across their sector. As a result, expectations are often shaped by conference presentations, industry articles and isolated success stories.

The problem is that these examples rarely represent typical performance. They represent exceptional performance, which creates a dangerous tendency to benchmark against either unrealistic goals or insufficiently ambitious ones. A business may believe its underperforming when its actually ahead of the market.

Equally, it may believe its performing strongly when significant opportunities remain untapped. This is where benchmarking becomes valuable. A conversion rate on its own is simply a number but whencompared to relevant sector peers becomes insight. Understanding what good looks like within a specific market provides a more useful frame of reference than any isolated metric ever could.

This is also one area where agencies possess a unique advantage. Because they work across multiple sectors and organisations, they develop an understanding of performance ranges, patterns and expectations that individual businesses cannot easily acquire for themselves. 

The value is not simply in reporting what happened but in helping clients understand how their performance compares, where the opportunities lie and where investment is likely to deliver the greatest return.

The enduring role of human expertise

The rapid evolution of AI has prompted many discussions about the future role of marketing specialists. The reality is that data has never been the entire story. One example comes from LEGO: faced with declining sales, the company spent years simplifying products and experiences based on what appeared to be a logical interpretation of available evidence. The prevailing assumption was that shrinking attention spans demanded simpler products.

The breakthrough came when researchers spent time observing real customers in their homes. What they discovered was that customers derived enormous satisfaction from completing complex builds. The challenge was not the problem, it was the source of the value. The conclusion was counterintuitive.

Rather than simplifying products further, LEGO leaned into complexity and this decision helped change the company's trajectory. The lesson is important: AI is exceptionally proficient at finding patterns but humans remain uniquely good at identifying what is missing from the pattern. Sometimes the most valuable insight comes not from the data itself, but from understanding motivations, behaviours and experiences that were never captured in the dataset.

The future is therefore unlikely to an ultimatum between AI or human expertise. Its AI and human expertise. One accelerates analysis. The other provides judgement.

The next source of competitive advantage

Historically, much of digital marketing's competitive advantage has come from measuring, understanding and optimising performance. Over time, those capabilities have become commonplace. What was once a differentiator has become table stakes. AI will accelerate that transition.

The ability to analyse large datasets, identify trends and generate recommendations will become increasingly accessible to everyone. When that happens, the value of marketing organisations must evolve. Success will belong to those who combine analysis with context, perspective and judgement.

The organisations that thrive will be the ones that can connect business objectives to performance metrics, understand interactions across channels, benchmark against meaningful peers and apply human insight where data alone is insufficient. In short, they will be the organisations that understand that marketing performance is not just a data problem.

While AI will become exceptionally good at helping us understand what has happened, the most important decisions will still require us to understand why it matters. That is the difference between reporting and insight. And it is where the next generation of marketing advantage will be found. Get in touch with IDHL to find out how you can leverage it and where our technology fits into your business