There’s a lot of hype about predictive analytics, which, as most experts agree, will undoubtedly transform online marketing. The whole point of predictive analytics is to predict the likelihood of an outcome through the use of data science, so its most promising application from a marketing standpoint is in the realm of performance marketing, as these efforts are tracked and measured to a resulting sale or customer acquisition.
There are several ways predictive analytics is changing the performance marketing game. Here, we’ll explore how it enables advertisers to make smarter bidding decisions, how it allows brands to monetize their non-converting users by showing ads to these consumers and finally, how it’s leading to the demise of the black box model that ad networks rely on to maximize profits.
Until recently, predictive analytics has been primarily applied to online advertising to identify those consumers with the intent to transact in order to target them through impressions-based media (display, video, social ads). Because of that, most advertisers think they are already leveraging predictive analytics for their performance marketing campaigns in the best way possible. However, predictive analytics is also extremely valuable when buying performance media on a cost-per-click basis, when you already know that the consumer has strong transactional intent.
The misconception is that since these customers have an obvious intent to transact, there’s no need for data science to predict this. This couldn’t be further from the truth. Sure, all customers comparing mortgage rates on Zillow are in the market for a mortgage, but that in no way means they are equally likely to buy a mortgage today – and it certainly doesn’t mean they are equally likely to buy from mortgage lender A vs. mortgage lender B. This is where predictive analytics comes in. Through predictive analytics, we can determine the likelihood that a given consumer who just searched for a mortgage rate in Phoenix will ultimately convert into a funded loan, as well as the expected value from that loan to lender A. Lender A can then determine exactly what to bid for a click from that consumer in order to maximize its return on ad spend (ROAS).
In addition to helping advertisers make better bidding decisions on performance media, predictive analytics also has the potential to create an entirely new category of media publishers – the brand publisher. In this instance, we’re defining “brand” as any company trying to sell a consumer their product or service. Let’s consider how a brand that sells car auto insurance uses predictive analytics to become a brand publisher and sell performance advertising. When a consumer goes to the company website, he or she is likely searching for information or a quote on car insurance. Naturally, the company would like to sell a policy to every consumer that visits its site; however, for all brands, the vast majority of consumers who search for a product or service on their sites don’t actually buy anything. With predictive analytics, brands can now predict both the likelihood that a consumer will convert as well as the expected value of that sale. For shoppers who have a low probability of converting into a sale, company can now display comparison offers from other car insurance providers in the form of native search results, with little to no risk of losing a potential sale. Thus, predictive analytics enables brands to generate a significant new revenue stream from non-converting shoppers (which can then be used to fund marketing to attract new shoppers), while also improving the user experience by providing their shoppers with a seamless comparison experience, with full confidence that they are protecting their primary revenue steam. In insurance, as well as travel and e-commerce, we are seeing the emergence of brand publishers who are leveraging predictive analytics to change the landscape of performance media in their markets.
Finally, the proliferation of predictive analytics will be the beginning of the end of the black box ad network. Ad networks are not transparent with their data because they rely on a black box, average pricing model to maximize their own profits. If an advertiser has transparent access to all relevant data about a consumer and can apply predictive analytics to bid precisely what they want based on the expected ROAS for that specific consumer, the ad network model of packaging and selling a large bundle of inventory based on just one average ROAS falls apart. To use an overly simplistic example, imagine that an advertiser is willing to pay $100 per sale and that an ad network is able to deliver $1M of performance media that converts, on average, at a rate of $100 per sale. All good, right? Not really. Imagine that by leveraging predictive analytics and transparent access to relevant bidding data, the advertiser can bid the exact right amount (sometimes zero) for each consumer based on the algorithmically-predicted likelihood of a sale and the expected value of that sale for that consumer. The advertiser may be able to buy just $500K of this media, not lose a single sale, and thus acquire customers for only $50. The advertiser had been willing to pay up to $100 per sale and may even have been satisfied under the old model, but is obviously far better off in the latter scenario. This is an overly-simplified example, but you get the point that there are enormous potential benefits advertisers can reap by using predictive analytics to de-average their media buying. The problem is that the ad network is now far worse off because it doesn’t get to make its industry-standard 30-40% margins on the $500K of non-performing or underperforming media that the advertiser didn’t buy. Unfortunately for ad networks, the increased use of predictive analytics is causing advertisers to demand transparency and access to more granular bidding data, which is slowly but surely tearing down the black box model these networks have long relied on to maximize profits.
If you think predictive analytics is impactful now – just wait. The aforementioned three points are just the tip of the iceberg. The technology that powers predictive analytics, the access that marketers have to valuable data to feed these models, and the availability of in-house data science capabilities is only getting better, and the effects that predictive analytics will have on the entire performance marketing industry will be unprecedented.