How Computers Could Care More, and Why It Is Critical for Machine Learning Marketing?
We all know about Big Data — megabytes, gigabytes and even terabytes of data, which is crucial for analytics. Remarkably, even a couple of megabytes of unsorted and unorganized data can bring a human being to its knees. That is exactly the problem that marketers face today — General Metrics of Facebook Advertising campaigns already look like this.
If you add up details, it will transform into a kilometer-long table, that you will be unable to master. To reinforce the understanding, let us remind you how Google Analytics’ small report looks like — the thing that the majority of us see every day.
Can you already use this data to increase conversion rate and ROI to accomplish your marketing objectives? Probably, no. At the same time, computers are made to work with data tables, and they will be significantly better than humans in working with marketing data. Even your cell phone can browse through the table faster than you can even think about it.
What we need to computers to do is to find consistent patterns in marketing data, which is the first step of every analytic. For instance, if we found that males in Boston converse to Captain’s ads worse than any other audience segments, we would exclude them from the campaign and save money on CPM. If we revealed that kittens on the landing page increase sales more than penguins, penguins would be thrown out into the trash bin.
Finding and using data patterns and connections between variables is the primary task of marketing analytics, which takes 80% of average marketer’s work routine. If we could use computers for it, we would save this time for planning, creating ads and strategies — the most value-adding activity in marketing.
So, if we order the computer to find patterns, it is enough for machine-powered analytics, isn’t it? Unfortunately, data is much more troublesome, and that is why machine learning has been revolutionary in marketing.
What Is Wrong with Data Patterns and Why Machine Learning Marketing Analytics Crave Intelligence?
If the data patterns analysis was so easy, then data scientists would not be hired by Google and Facebook as quick as new iPhones are being sold out on preorder. Captain remembers many hilarious cases where mathematically-true patterns between variables appeared to be nonsensical.
For instance, let us take a computer, connect it to Wikipedia and allow extracting any data to find the cause of the global warming. The machine makes some noise with its fans, loads for a while and draws this graph.
Indeed, when we check pirates’ quantity and average Earth’s temperature by alone, they look like related — when there were many sea bandits, the temperature was lower. We may even retest these results because, with the emergence of Somali pirates, the global temperature decreased, which now fuels Pastafarian’s excitement.
However, it is entirely nonsensical — pirates have no means to affect temperature. The computer cannot get it, though. Without an understanding of data, all statistically significant patterns seem legit to software.
However, let us give a second chance to the computer and ask it to analyze the consequences of Internet Explorer failure in web browser market. Few seconds pass, and the computer alarms — critical coincidence is found!
Unless poor UI design drives people crazy and violent, these findings are nonsensical again. It appears that successful automated analytics requires the understanding of what is being analyzed, and what are its possible consequences.
By the way, the lack of machine understanding is the reason for why online marketing dashboards are useless. On dashboards, marketers dig through charts with global warming pirates and deadly Internet Explorers to find valuable insights — still wasting 80% of their time on data.
That is high time for data science to step in