What If Reporting Was Always Broken? Navigating the Space Between Verification and Forecasting

Are Businesses Addicted to the Illusion of Certainty?

Over the last couple of weeks, I have been unable to stop thinking about a brilliant episode of The Rest is Politics Leading, where Alistair and Rory spoke to James Manyika about AI and technological development. James described how a quantum computer recently completed a calculation in five minutes that would take the world’s most powerful supercomputer 250 septillion years. Astonishing, yes—but Rory pointed out the fascinating catch: verifying this result would also require 250 septillion years, making it impossible to truly confirm. This means we must trust without traditional verification, relying instead on patterns and probabilities.

Reflecting on this, I couldn’t help but see a parallel in marketing. We constantly produce reports, dashboards, and attribution models, each aiming to quantify and prove the impact of our decisions. But how often do we actually stop to question whether these reports genuinely capture business performance? Perhaps we love reports precisely because they provide comforting illusions of certainty, numbers neatly packaged and ready to defend our decisions when questioned.

AI presents us with a compelling alternative: sophisticated, system-wide forecasting. Unlike reporting, which rigidly accounts for past actions, forecasting models systems holistically, embracing complexity and interconnected effects. It feels less definitive, yet it arguably aligns better with how marketing genuinely works, capturing the long-term value that the work from Ehrenberg-Bass through Binet & Field and beyond have consistently shown.

Yet, does embracing forecasting mean abandoning reports entirely, or can we create a balanced model that uses both effectively?

The Myth of Verifiable Reporting in Marketing

For decades, marketers have leaned heavily on data sources to measure performance, often treating these reports as definitive even when riddled with limitations. Digital marketing, in particular, produces extremely granular reports, impressions, conversions, and ROAS that appear precise. But does this flood of detailed data really illuminate consumer intent and behaviour, or does it just create an illusion of clarity?

Marketing research panels and studies can provide broad snapshots yet often struggle to capture diverse consumer segments accurately. Marketing Mix Modelling (MMM), despite its well-earned status as a gold standard, depends on historical aggregated data and regression models that inherently smooth over individual-level insights.

Despite known shortcomings, these reports remain trusted, often because they are easily explained and defended in meetings. It feels reassuring to point to a report and confidently say, “This is what happened,” even if the reality is far more nuanced and complex.

This reliance on measurable data echoes the McNamara Fallacy, the tendency to focus only on quantifiable aspects, ignoring crucial but less tangible factors. Reports often prioritise fast-moving metrics over difficult-to-measure messy low-frequency effects. But just because something is hard to quantify doesn't mean it isn't critical. Reports give us cover, yet they rarely offer insights into the full complexity of business performance. Forecasting, however, can simulate entire systems, leveraging machine learning methods such as Bayesian inference and neural networks to reveal more meaningful pictures of likely outcomes.

Forecasting: Expanding Options, Not Issuing Orders

Forecasting itself is not new; marketers have relied on forecasts for decades. The difference now is how AI technology drastically simplifies and enhances the process. Traditional forecasting often required painstaking manual collection of historical sales data, followed by simple time-series or regression analysis. In contrast, AI-driven forecasting rapidly integrates vast, diverse datasets from multiple real-time sources, continuously updating as fresh data emerges. Techniques like ensemble models, gradient boosting methods (such as XGBoost), and deep learning algorithms can rapidly identify complex relationships and forecast a range of scenarios simultaneously, significantly streamlining the entire forecasting process. Moreover, synthetic data—artificially generated datasets that replicate real-world behaviours—can now be placed directly on marketers' desktops. This enables real-time querying of forecasts, creating actionable insights in minutes that would previously have required weeks of data collection and manual analysis.

However, AI forecasts are only as good as the quality of the data they receive. Poor-quality or incomplete datasets can lead AI models to produce inaccurate or misleading outcomes. Just as overly detailed reports can falsely suggest certainty through granular data, sophisticated AI forecasts can similarly create a deceptive illusion of precision if the underlying data isn't robust. Marketers must maintain rigorous standards for data validation, collection, and management, because no amount of advanced technology can overcome fundamentally flawed inputs. Despite these challenges, when implemented carefully, AI-driven forecasting offers compelling practical benefits.

Take advanced predictive analytics in retail, for example. Instead of delivering simple forecasts based solely on historical sales, these tools rapidly integrate diverse data like consumer behaviours, competitor actions, and real-time market trends. While organisational inertia, skill gaps, and unrealistic expectations often challenge adoption, the reward is a clearer view of multiple realistic outcomes, enabling marketers to choose strategically.

Similarly, generative AI makes large-scale creative testing feasible. Previously slow and expensive manual testing is now replaced by rapid simulations of hundreds of creative iterations. This provides marketers with numerous promising scenarios, rather than limiting them to one prescribed approach.

AI-powered scenario simulations in media planning further enhance strategic choice. By modelling various budget allocations, media mixes, and targeting strategies, marketers gain a richer context and clearer foresight into potential outcomes, ultimately expanding strategic possibilities rather than restricting them.

Could embracing AI as a provider of informed options rather than rigid orders help marketers navigate the uncertainty inherent in strategic decisions more effectively?

The Verification Problem: Can We Trust What We Can’t Check?

If we shift towards forecasts, how do we cultivate trust without the traditional safety net of verification? Marketing has navigated similar challenges before. Econometrics, initially viewed sceptically for its complexity, is now widely trusted due to decades of consistent performance and refinement. Similarly, investments in long-term brand building have always required some faith in future outcomes, relying on models and benchmarks without immediate proof.

AI forecasting operates in the same uncertain space. Here, validation involves advanced methods like cross-validation, scenario backcasting, and sensitivity analysis rather than simple retrospective checks. Marketers must remain actively engaged in verifying model assumptions, data inputs, and methodological transparency. Trust grows from consistent predictive accuracy, transparency of methods, and thoughtful validation processes over time, reflecting that reliability is more meaningful than immediate explainability alone.

Conclusion

We’re not yet in a world where forecasts have fully replaced reports, but perhaps that's not the goal. Instead, marketers should equip themselves and their teams with critical skills, such as analytical thinking, data literacy, and strategic judgement alongside transparency standards and rigorous processes to use both effectively. Our real challenge may not be deciding whether to trust reports or forecasts more, but learning how to wisely integrate them both, making better-informed decisions in an increasingly complex world. Moreover, this balance between reporting and forecasting will soon be faced by multiple departments across the business, not just marketing. However, marketers may have a head start due to their longstanding experience in effectively communicating complex, long-term, and often unobvious results.

However a note of caution, while AI may streamline and enrich our reporting and forecasting, we must be careful not to end up drowning in five times as many reports as before, or else we'll soon find ourselves needing another AI simply to review the first AI's reports. Otherwise, we risk turning every marketer into a full-time professional report editor.