Adam Goldberg is Chief Innovation Officer of ClearSaleing. AdExchanger.com asked Goldberg about the latest advancements and traction with effective attribution modeling in digital advertising.
AdExchanger.com: How does ClearSaleing approach attribution across channels and what are the coming milestones in cross channel attribution? Attribution modeling for an SEM direct response campaign seems pretty straightforward.
AG: Today, we have the ability to perform attribution across any online channel. We are able to account for the value of a click, a view-thru, an interact-thru, and video on a conversion. If we can track that an ad, a video, a direct visit, organic visit, etc., was involved in a conversion, we can attribute credit (profit, revenue, sales, engagement) across those activities in a statistically sound manner.
When it comes to attribution modeling specific to SEM direct response, one needs to be very careful here. It is rare that SEM is the only channel driving visits to a client’s site. If there are any other vehicles driving traffic and they are not accounted for in the attribution modeling, then the models are based on an incomplete data set, which will lead to decisions being made on bad data. Attribution cannot be done in a silo.
We are already capable of doing media mix modeling that incorporates not only the online world, but the offline world as well. As technology enables us to better track TV and mobile, we will be able to get more granular in our reporting and thus create even more effective attribution models that account for the entire advertising universe.
AdExchanger.com: On the video channel, do you think video's engagement metrics can work with search and/or display in such a way that a marketer will be able to spend appropriately on each channel?
Absolutely. Today, in the world of display, we have clicks, view-thrus, and interact-thrus. An interact-thru is when someone actually engages with the display ad by either simply mousing over it or perhaps by doing something more engaging within the ad itself. We can track these different types of interactions when it comes to display and provide different attribution weights to these actions when appropriate.
This is really the same for video. We have the ability to study how a channel performs when it is the first step of a path (introducer), when it is the last step of a path (closer) and when it is a step in the middle (influencer). By understanding the impact an ad vehicle has relative to its position in the path, we understand how much influence it had on a particular conversion, and then we can attribute the correct amount of credit. It becomes clear which types of ads generate the most profit. If attribution shows video to be a huge profit driver, regardless of where it most often occurs in a path, then allocating more spend to video makes sense. Attribution makes it easy for a marketer to put a value on the importance of ads on their bottom line or to whichever achievement metrics they have set.
AdExchanger.com: Are you seeing attribution helping the search marketer understand display today? Any trends you can share?
Definitely. Display was seen for a long time as a secondary advertising source relative to search because people rarely click on display. Even if they do click a display ad, it is rarely the last step of a path prior to conversion. Therefore, when a marketer looks at last click metrics, display looks like it is bleeding money.
By using attribution, our clients are seeing the real value of display. Since they are now able to measure display impact at a view-thru or interact-thru level vs. only on a last click method, they are able to see how display contributes to their goals. Display has proven to be great at driving incremental reach, at getting a brand and/or offer in front of new audiences. Clients see how display acts more often as an Introducer or Influencer vs. a Closer. In order to complete a conversion with a customer, you have to get into their consideration set. Display is perfect for starting the conversation. Our clients are moving more and more of their ad dollars into this channel based on the data they get from attribution.
AdExchanger.com: For ClearSaleing's business, do you think about providing attribution metrics that are different for brand awareness marketers versus direct response?
Ultimately, if branding is effective, we should see consumers that are exposed to brand messaging visiting our clients’ sites and converting. ClearSaleing’s ability to show a Purchase Path (chronological sequence of events a consumer clicked or saw en route to conversion) clearly shows how brand advertising led a consumer to visit their site and eventually buy. We, therefore, attribute credit to the brand ads. If a client of ours sees that some of their brands ads rarely result in a Purchase Path, then they would likely determine that those brand ads are not effective. No longer is it safe to assume that if X number of eyeballs saw your brand ad, then that means good things will happen down the line. We are tracking that line to definitively know if those eyeballs are turning into dollars.
AdExchanger.com: What about marrying offline attribution metrics to online? What might be a first mover in terms of metrics? Geo data overlaying in-store sales mapped against geo-targeting digital campaigns?
Today, we are tying offline sales (phone, CRM, bricks and mortar) to online advertising for quite a few of our clients. There are several techniques we use to do this. Simply put, if a consumer visits one of our client’s websites and converts offline, it initially appears as if there was no conversion. What we need to do is get that consumer that purchased through an offline channel back to the website through use of an email that is triggered when an offline sale occurs, or by having the consumer make use of the “My Account” tab many websites have. Once we get the offline consumer back online, we can see if there are cookies present and if so, we can then match the ads they used to get to the site to the offline order that took place.
When it comes to offline media we are able to account for this in a media mix model. Today, we cannot definitively say this consumer saw commercial X, then went online and made a purchase. But we can say that TV deserves X amount of credit for sales in aggregate. This data helps clients understand how much money to spend in each channel to get the greatest return on ad spend.
AdExchanger.com: What is the biggest misconception about attribution modeling? How do you and ClearSaleing address it?
There are many misconceptions out there. So many, that I dedicated an entire webinar to the subject of “Attribution Myths and Misconceptions”.
In this webinar, I highlight 12 misconceptions:
- I don’t have an attribution problem
- The last click is the chosen one
- There are no good methods for assigning attribution credit
- There are no good tools for attribution
- Attribution can be done with web analytics
- Attribution can be done in a silo
- Attribution is about buying the right mix of media
- Attribution pulls dollars away from search
- Path analysis is a waste of time
- A/B testing is effective for attribution
- Attribution Management takes too much time to be worth it
- Attribution Management is a silver bullet
If had to choose one as being the biggest misconception, it would be that there are no good methods for assigning attribution credit. We take our clients through a crawl, walk, and run approach to attribution. The crawl stage uses a very simple, yet highly effective attribution model. People always believe that if you do not have a perfect model, you cannot perform attribution. To combat that I use a quote from Voltaire, “The perfect is the enemy of the good.” The first model we implement for our clients is one that attributes credit evenly over the team of ads in the Purchase Path. This model is night and day better than a last click model. We have proven across every client we have that this model will increase their return on ad spend and shed a bright light on the importance of ads that introduce and influence. We also recommend, out of the gate, that our clients exclude giving credit to branded keywords that occur in the Closer position of a path. The logic there is that when a brand term is used at the end of a path, it is almost always the consumers chosen way to navigate back to the site. We want to give credit to the ads that made the sale occur, not to an ad that was just used for navigation. The Even Model with Exclusions on brand terms as closers is highly effective. It’s by no means perfect, but it is a far better, and more accurate method, than giving all the credit to the last ad clicked.
By John Ebbert