One of the most annoying and inefficient tasks in life is going grocery shopping. I have to visit up to four stores every week to find everything for the family, and that eats up plenty of time, gas, and money.
Life would be great if we could buy groceries programmatically, right? I'm not talking about simply buying them online and having them delivered (i.e. FreshDirect), but rather buying them in the same way media is bought and sold. Set a menu in advance, combine it with first-party data from a "smart fridge." The items are scanned when they're put into the fridge and tracked when taken out, put back, or getting depleted. Throw in some family eating habits, maybe some third-party data from healthy eating studies, and the groceries arrive at the door in the exact amounts needed.
This would be more efficient and result in a lot less waste, which is great for the wallet and the environment. Although buying groceries programmatically would result in a family getting a lot of their preferred food and ingredients, it would result in the same meals week after week, month after month. This limits the ability to branch out and try new things. Without going to a store and looking at the stock, you're never aware of new flavors or alternatives. Online marketers are in great danger of falling into this exact trap -- being put into siloes through an overreliance on data and ceding too much control to automated algorithms.
So, while many have touted programmatic's advantages in targeting, data, performance and, above all else, efficiency, the vital human element is often forgotten. If you're using algorithms to buy groceries, you can get out of the rut by inputting new recipes.
Marketers need this too, in the form of a data expert who can make decisions on what to do, based on the data the marketer is inputting and generating with each subsequent campaign.
While programmatic leverages machine-based learning, a machine cannot make a gut decision. Humans can look at the results and say, "let's try this" to find new inventory elsewhere and produce slightly altered audience models that perform better. A high-end headphone manufacturer may be looking for males in their 30s who live in major cities and are shopping for Apple products. But machines don't know that Samsung is putting out a hot new phone, nor can they be taught to look for big updates to Android devices. These are two situations that will trigger a behavior change in the original audience. Rather than searching for Apple products, those 30-year-olds are suddenly paying attention to Samsung, and there are far fewer Apple-intender impressions available.
A human running the campaign can change the audience target slightly to reflect the potentially larger audience pool. It's like finding a new salad dressing you never knew existed. The same basic vegetables (or attributes) are underneath, but there is a new, slightly richer characteristic on top. Algorithms may suggest something new to replace a depleted item, but a human recommendation is the strongest endorsement, whether it's for food, media or audience targets.
For more proof of the dangers of trusting formulas too deeply, just consider the broadcast television network show model. Television networks put every single one of their shows through rigorous audience testing and focus groups. Shows that test high make it on the air and are promoted with multi-million dollar ad buys. This is data-powered decision making, but every year shows are cancelled mere weeks into the season at a huge loss to the network. The focus group data turned out to be worthless, and you have to wonder if anyone had a gut feeling and was afraid to challenge the data.
Algorithms can't do everything. They're good at helping a marketer repeat processes that are often tedious and time-consuming. But even the most cutting-edge machines are not good at exercising curiosity or understanding the potential for higher success based on outside factors. Marketers that are not using some human expertise to power their programmatic buying will be stuck eating the same inventory salad over and over.
Jon Schwartz is VP of sales at Netmining.
On Twitter? Follow iMedia Connection at @iMediaTweet.