“In the insights industry, there is a real gap between what has been traditionally available and what we need today,” says Tim Warner, who heads PepsiCo’s consumer and market research…. Leading consumer goods companies want to upgrade decades-old techniques, such as consumer surveys … which are seen as too slow, too expensive and often incomplete. (FT)
With all the technological innovation out there for marketers, it’s hard to say why market research hasn’t caught up. Maybe it’s hard to break out of “just what we’ve always used.” Maybe some lucky salesperson from the old guard scored a two-decade-long contract.
Whatever the reason, brands are increasingly less willing to wait around for a research vendor to assemble the right panel, organize data, suggest how to use it… and pay a hefty sum for the privilege. Business decisions — from product to marketing — need to be data-driven to be responsible business decisions, and the process for gleaning and leveraging those data needs to be modernized.
Ad creative needs smart, fast, actionable research more than anything. Why? Because, as Nielsen Catalina documented, creative drives nearly 50% of advertising effectiveness, more than targeting, reach, and brand combined. With that much power, who wants to leave figuring out what’s most effective to “the old methods that were invented before the digital era?”
In audio marketing, the problem is actually compounded. More than just facing legacy measurement systems, many of the biggest companies in the world have yet to measure audio creative at all. In today’s “audio renaissance,” how is it possible that decisions about ads, sonic branding, voice and more are still often made by gut? Look no further than the most commonly referenced stats (e.g., smart speaker adoption is growing faster than the early days of smartphones; 65% of podcast listeners are inclined to buy a product advertised on the show; etc.) to understand just how unsustainable that is.
So, for marketers concerned about getting audio marketing right — and tired of hitching up the horse and buggy of traditional research to try to figure it out — there’s good news: Machine-learning measurement — that quickly and accurately predicts how people will respond to audio creative — is here. It’s called the Audio Effectiveness Platform.
What makes an audio effectiveness platform effective? Four things.
Merely moving from traditional research methods to a predictive platform is a big step as it is. But four critical components need to be present to fulfill the big promise.
1. The robot has to continually absorb tons of creative data
The smartest machine learning platform in the world can’t learn without analyzing a large and steady volume of audio assets, from voices to streaming ads to sonic logos. Creative needs to be assessed for characteristics like timbre, brand mentions, use of music and the like, then weighed against other assets in the system for its’ ability to drive the marketer’s desired outcome.
Let’s say a sneaker brand is assessing 50 voices for the one that will sound most “inspiring” to podcast listeners. The platform needs to be able to analyze those voices against a critical mass of other voices that have been gauged for their power to inspire podcast listeners.
Pandora’s Lauren Nagel describes a similar test that Pandora did on The Sonic Truth podcast.
2. It needs to know what “effective” is
“…market research was all about mitigating risk of the decisions that the business had already made,” says Stan Sthanunathan, who has led consumer insights at Unilever since 2013. “Today our role has changed to anticipating consumers’ desires….”
For the system to predict how effective audio creative will be, it first has to understand what “effective” audio marketing actually is. An aggregate of different research philosophies points to four key components. Effective audio marketing:
- grabs someone’s attention (engagement)
- connects with them (emotional resonance)
- is memorable (recall)
- persuades them to want to buy the product (purchase intent).
While “machine learning” generally describes how a platform gets smarter as it processes more data, we like to think of the audio version as “machine listening and learning,” or M-LAL. “Listening” describes the machine’s ability to hear and understand the assets, and “learning” means evaluating them in the context of everything ingested and analyzed previously.
Back to our example: let’s say our sneaker brand is also looking for the voices that are most likely to persuade consumers to buy its sneakers. To predict it accurately, the machine obviously needs to be listening to and comparing against thousands of other voice assets that have scored relatively high for purchase intent in podcasts.
3. It needs to keep getting smarter…with the help of people
While survey-based businesses are emblematic of “the old ways,” human response data is still important — just in the right context. Predictions need to be constantly validated and panels of people can help do it.
Back to the sneaker brand. Let’s say the platform, assessing a particular sponsor voice, predicts a relatively high score for “inspiring.” Without losing much time, the brand should be able to check those results against any custom audience segment it wants — say, several hundred in-market sneaker buyers in the midwest who listen to podcasts at least twice a month.
Once that data reinforces (or corrects) the prediction, those new learnings need to flow seamlessly back into the platform. That, in turn, makes the robot predict even more accurately next time.
4. It needs to produce a simple, standardized score quickly
Intelligence about audio creative needs to be robust, cost effective and fast. But ultimately, if it can’t actually be put to good use easily then what’s the point?
The most useful output of audio effectiveness analysis is a simple score that not only incorporates the most relevant data, but that adheres to a standard that makes comparing across the market easy.
One more time to our sneaker brand. With smart speakers as a key part of their strategy, they know the “voice of their brand” needs to resonate with listeners as much as, if not more than, their competitors. The platform needs to make that benchmarking easy.
A simple, universal score makes the relative value of their voice asset — and their smartest path forward — clear as a bell.
It’s unsurprising that audio, which is driving some of the most ubiquitous innovations in the modern era, is driving innovation in analytics as well. Technology at its best gravitates to where it’s needed the most. If one innovation ensures that millions of podcast listeners get ads for products they actually care about, another one needs to ensure that what they hear actually compels them to buy.
From the earliest days of radio, the smartest brands, media companies and others have always known just how resonant and powerful audio can be. Now, in the audio renaissance, the same types of leaders are embracing technology to reveal that power clearly and quickly. And it’s only getting smarter.