The premise of attribution is simple: Give credit where credit is due.
It’s how credit is determined where things get tricky. Marketers have long known that last-click attribution doesn’t adequately measure the effect that various single ad touch points contribute to a conversion. Media attribution companies solve for this through statistical modeling approaches, which assign relative value to the different parts of the media mix.
Of course, the value of these results is only as good as the model. And the reality is, there is no one right model. Each algorithm has pros and cons, and every application, whether it’s answering a business question or measuring marketing efficiency, is different. Banking on a single algorithm can be inaccurate and limiting.
The most effective solution to solving the attribution conundrum comes from leaders in predictive modeling. Rather than rely on a single algorithm to measure attribution, marketers can take an “ensemble” approach that uses multiple algorithms. By blending results from several models, the ensemble framework achieves higher accuracy than any single model on its own.
The ensemble method was recently popularized in a book by New Yorker columnist James Surowiecki called “The Wisdom of Crowds.” In short, the wisdom of the crowd is a statistical concept that states that multiple predictions from a “crowd” (whether of people or statistical models) will generally be more accurate than any single prediction on its own.
In an experiment at a county fair in England, NPR’s Planet Money posted a picture of a cow and asked people to guess its weight. While the 17,000 guesses were all over the place, the average guess was within five percent of the original weight. “There's something magical about it. It's not magic. It's just math, but it seems magical,” Surowiecki said.
Each person in Surowiecki’s cow example serves as a data point, providing input to the answer. When pulled together, they deliver the most accurate result. Similarly, an ensemble framework allows a data scientist to bring in models that each offer a unique point of analysis in order to achieve better, more meaningful data. On its own, each model might get pretty close to the right answer. Together, they hit the bullseye.
For a marketer running a broad combination of digital and offline campaigns across multiple channels, an ensemble method of attribution can increase predictive accuracy as much as 35% versus a single-algorithmic method. Depending on budget, that means big savings:
• $100 million budget: $25-$35 million potential savings
• $1 billion budget: $250-$300 million potential savings
• $600 billion, the entire global advertising industry: $15-$21 billion in cost savings
What does the ensemble method look like in practice? In short, it’s smart and adaptable. While a single-model methodology may work for a set amount of time or circumstances, the ensemble method adapts to changing market or business conditions. In the marketing world, those variables may be a new product, audience segment, consumer behavior, or media marketplace. An ensemble model is retrained to account for these changes, delivering attribution results that are more predictive and accurate over time.
In the end, that’s what marketers care about: better results. While single-model attribution technology can provide solid answers, a methodology that leverages multiple algorithms delivers the best of all worlds: the real wisdom of crowds.