Do you think the challenges of attribution have become more difficult?
It’s always been an issue—it’s never been a simple response model. But now it’s a multidimensional challenge that has greater nuances and much larger supplies of data, which people haven’t gotten their arms around. The amount of data has increased, and the amount of effort it takes to process it is also up. The models were never very good to begin with, but now consumers have a lot more information at their disposal, and they use that information differently and communicate it differently with each other.
How could attribution models be improved to better respond to the changes?
The easiest way to directly attribute is on the transactional basis. What we don’t know in the big picture is what all the antecedents are. It’s incorrect to say, “I did this, so now I’m selling more, so that’s all I need.” There [were] many antecedents that led to the purchase intent [before] that call-to-action appeared. [For example,] what’s going on with the category overall? If I’m soap, and soap bar sales as a category is down and liquid soap is up, I have to ask if I’m down because the whole category is, or if my competitor has better advertising.
Is my word of mouth bad?
Are people saying something bad on social media? It’s never a simple stimulus response. If brands have that data, they know what it means if they see an uptick in new data. If you’ve been monitoring everything you do, then when you alter that, or change one of your levers and see a different response, then you know what caused it. Always be collecting the information, understanding what your brand means and how sensitive it is to promotions and new advertising. Attribution modeling is by no means easy.
What are the particular benefits of direct response marketing?
If it’s sold on TV only, then you know where the buyer heard about it. But if it has a retail component, maybe they saw it in the store, or a friend on Facebook talked about it, then they saw an ad on TV late at night and decided to buy it.
What kind of questions should marketers be asking?
What are the correct statistical models? Where does you timeline begin? You have to look at the business cycle and advertising cycle to set your parameters. If you have a brand that has been around a long time, you need to ask, “Were they predisposed to your brand? Were they newbies?” It’s a complex set of data.