1,000 dealers share their thoughts about chat, text and messaging in general...and how these communications pay off. SEE HOW
Every week, ad agencies and other vendors contact thousands of dealerships, pitching their digital advertising services that feature ‘real-time bidding’ or RTB. The term has become ubiquitous in the digital advertising space, the marquee feature of an ad delivery system. But there’s more to the RTB story than those agencies and vendors might not be telling the dealers they’re contacting. First, let’s define the term, understand how it can help or harm a dealership, and, finally, examine how “machine learning” fits into the digital advertising space.
The simplest and most accurate way to describe real-time bidding is in terms of the length of time between cause and effect – in the case of display ads, a shopper visiting a website causes a page to load; the resulting effect is content delivered to the shopper’s screen. Most sophisticated publishers and web properties – Car & Driver, YouTube, and The New York Times, to name a few – deliver display ads to visitors within the 20 milliseconds it takes a page to load. Instead of The New York Times queuing ads to serve to readers prior to them visiting the website, it submits an offer to digital advertisers to bid on for the ad space, millions of times per day, all in hyper-fast auctions. This is real-time bidding.
Although these droves of advertisers are constantly competing for the same shoppers’ attention in “real-time,” there’s one major issue: It doesn’t matter if the advertisers are bidding for shoppers’ attention in real-time if the advertisers don’t know who the shopper is.
For example, a BMW dealer running display ads that is accidentally serving advertisements to 13-year-olds using real-time bidding is obviously going to have a really tough time moving inventory.
Rather than just quickly – and blindly – serving ads to anyone online, it’s important to learn everything a dealer can about their shoppers, all within those 20 milliseconds. Dealers need to ask themselves what actions these shoppers have taken in the immediate past to indicate they are buyers and will respond to ads (this goes far beyond simply looking at web history). Which specific attributes of an ad predict the likelihood of shoppers visiting a dealer website?
This is why ‘machine learning’ must be part of a successful real-time bidding solution. Consider this analogy: RTB technology acts like a stock trader, collecting offers, making bids, and delivering ads based on the intelligence provided by his or her analyst (machine learning).
The machine learning component considers billions of probable outcomes and continuously decides how to optimize dealer ad campaigns every second of the day. If a particular topic on YouTube is generating high response rates within a dealer’s target market, and a shopper with an interest in the cars that dealer sells visits a related video – BOOM! The system seizes the opportunity and throttles the bid made on the ad space offered, resulting in dealers getting the most accurate ad to the most interested shopper.
Most ad tech companies are more or less throwing darts at a board when it comes to bidding and ad delivery (I have first-hand experience with this, coming to Dealer.com from a position within the largest newspaper publisher in the US). Their campaigns are set up to make consistent bids to achieve consistent CPMs (the ‘Cost Per Thousand’ times an ad is delivered). But this makes no sense. Why would any dealer care that it’s supposedly less expensive to deliver ads to people who are in no way interested in his or her inventory, parts and services?
The great irony is how much less of an investment needs to be made to run a display campaign the right way from the beginning. Machine learning allows campaigns to understand the entire shopper ecosystem and optimize ad delivery for the most impactful and profitable outcome (a website/VDP visit).
A dealership’s digital ad system needs to include both real-time bidding and powerful machine learning in order to identify in-market shoppers, competing advertiser bids, data that indicates which sites are most profitable, ad combinations (images/offers) that work best for each individual, and how to define success.
Anything less and dealers might as well join the competition already throwing millions of darts at a very big dartboard, and missing most of the time.