The real problem with marketing intelligence isn’t insight, it’s execution. How can AI help?

For many organizations, the martech stack comprises many different applications and platforms; some play a major role in marketing, while others fill niche requirements. In all cases, there is data that helps marketing teams understand how their strategies are performing. The problem is that most of the data is siloed and disconnected, resulting in fragmented marketing intelligence, and the more tools marketing adds, the less reliable that intelligence becomes.

The NinjaCat & UserEvidence report, The Next Phase of Marketing Intelligence: AI Maturity Across the Analyze, Optimize, Act Lifecycle, shared insights around what’s working and what’s not in marketing intelligence, and how AI is the answer most organizations need to not only produce accurate insights, but act on them quickly.

The insights shared in this report focus on the core loop that marketing teams use to understand and act on performance data:

• Analyze: Consolidate, validate, and interpret data across channels to understand what works and what doesn’t
• Optimize: Identify what to change based on the insights gathered and benchmarks measured.
• Act: Execute these changes live across platforms, and measure the impact for future refinement.

So where, according to this study, does the loop break down? 

Data access isn’t the bottleneck

Marketing and advertising teams spend a lot of marketing budget on paid media; an average of $26.2 million per year across channels, and most use point solutions to get performance insights. What this study of over 500 marketing leaders discovered is that there is fragmentation, but it’s not due to the lack of access to data (as many might believe). Instead, fragmentation occurs because marketing teams can’t act on performance data quickly or across systems.

Marketing teams use an average of eight martech/adtech platforms and have another three tools to identify performance issues and opportunities and make the necessary changes.

As for what the top paid media channels are:

• Paid social - 75%
• Paid search - 69%
• Email marketing - 59%
• Programmatic display/video - 52%
• Audio/Podcasting advertising - 50%

If access to the data they need to make decisions isn’t the problem, then what is? The report offers what you might call the yin and yang of performance marketing:

• Teams have trusted benchmarks to compare against and are satisfied that they can quickly analyze marketing performance data.
• There are serious issues related to aggregation and normalization, as well as to building reports and dashboards.

The breakdown occurs in managing fragmented data across systems and in the quality of that data when it’s brought together for analysis. Only 37% reported having a single, unified source of truth that covers all channels and campaigns.

Much of this work to bring all this data together is manual and slows teams down in reaching the right decisions and acting on them. In this study, 73% said they spent a significant amount of time reconciling inconsistent data before it’s analyzed, and 45% said that the reporting process is manual and requires a lot of effort, taking up to five days turnaround. As a result, teams are often making decisions using stale data. 

How AI improves performance analysis

Like other marketing strategies and programs, AI has the potential to transform marketing intelligence, and many of the marketing leaders surveyed for this report agree. But the question is, is there more AI can help with? And the answer is yes.

Some of the top use cases noted in the report include:

• Identifying optimization opportunities
• Generating written performance summaries
• Detecting anomalies or significant performance changes
• Ingesting and unifying data

There were also more advanced use cases mentioned, like orchestrating multi-step workflows around marketing tools and teams, but these are rare because most teams are not at the level of AI maturity necessary for this level of orchestration.

The report suggests that AI initiatives are augmenting analysis rather than truly transforming execution. And there are a few reasons for this.
AI maturity still lags

One of the primary reasons marketing teams are still lagging with AI is because they are using the AI capabilities in the individual platforms they use (57%), or isolated AI point solutions not connected to performance data (19%). Only 16% have a centralized AI/ Machine Learning layer and only nine percent are piloting or using AI agents that act across systems and workflows.

What are the biggest barriers to using AI successfully? 

The biggest barrier noted is the ability to integrate AI tools with platforms and data. Data privacy and security was also high on the list as was limited internal AI use and a lack of trust in AI-driven recommendations.

What happens all too often is marketing teams bolt AI onto existing processes instead of redesigning workflows with AI at the core. When they don’t step back and re-think their processes and how using AI strategically can make things better, AI won’t have the impact necessary.

The biggest potential for using AI in marketing intelligence is that it can help shift the focus from reporting to decisioning. AI can do the basic work of bringing data together, analyzing and summarizing it, and suggesting improvements. Then, it can help implement those recommendations. It’s a multi-step workflow that connects with multiple tools upstream and downstream to analyze, decide, and act on insights.

A roadmap to bring AI into the loop

It’s fine to point out all the barriers to leveraging AI in performance marketing, but what does success look like? The report offers a roadmap, and it’s a pretty standard roadmap for implementing AI into any aspect of marketing (or sales and service for that matter).

The first step is to consolidate your tech stack, and invest in a unified data layer, but this is probably one of the hardest things to do. With so many different marketing tools supporting paid media, including tools that activate the data (like marketing automation or email marketing), the work required to bring that data together can take a lot of work.

Some marketing teams look for platform solution that provide this unified data layer (like a customer data platform), but there are still bespoke AI-based applications that standalone and will need their data integrated with the platform data.

The second step is about assessing where your team is at today, including how they work and the tools they use, and coming up with a strategy to address those limitations. To me, this should be the first step, because until you understand how you are working today and the issues with your current tools, you don’t know how to best consolidate your tech stack. You might also determine in your assessment step that a certain tool is no longer required. Better to know that first, then consolidate its data and find out you didn’t need to do that work.

The last two steps are about putting guardrails in place for data rules and access controls, monitoring, measurement and ownership, and training your team to supervise AI and prove its working. 

My take

Marketing teams that think strategically about how use AI are the ones who will get the most benefit from it. But support needs to come from high up, because using AI requires rethinking how things are done today and too many leaders don’t get that.

Most marketing processes are designed for humans to do the bulk of the work, so there are a lot of manual steps built into those processes. But when you bring in AI, it takes on a lot of that work, leaving the humans to spend more time on reviewing and approving, as well as strategic thinking and execution. However, do you really need a human in the loop for every step in a process, or every process? That’s the question we are at now and thinking about taking the human out of the loop requires truly rethinking where AI really fits in the process.

While most organizations are figuring out the best way to incorporate AI in marketing intelligence, the ones that have already figured it out (and a small percentage have) are at the next step of leveraging AI agents to perform most of the work without the need for a human in the loop.