If you have ever wondered what the proper use of AI in a marketing department should look like, consider the example of the recent rollout of a video newsletter about general engineering by IEEE GlobalSpec.
The first step IEEE GlobalSpec took was to build a propensity model (a statistical scorecard that predicts the behavior of a customer or prospect base) that examined all the users that have a chance to interact with graphical elements, explained Don Lesem, Vice President and Chief Design Officer at the organization. Not just video, he emphasized, but all users that might click on a GIF or look at tables or otherwise interact with images. “We did not worry about the title of the user — whether they were, for example, an electrical engineer or mechanical engineer,” he said.
From there the organization built the subscriber list and when the actual newsletter was sent out it generated three times the average open rate and three to four times the average click through rate, Lesem said.
Now this is what that example might have looked like if the marketing department was not well versed in AI technology, even if it had it deployed. First, the marketing department might approach the subject with preconceived notions such as only Millennials look at videos, Lesem said. So it would only use data about videos and the type of people who looked at them.
The marketing department might also look to add certain titles of the demographic such as engineers and possibly limit it to engineers who have been in their jobs for four years. It is also possible that this marketing department might rely on survey data that showed that, when asked, engineers answered “no” to whether they would look at a seven-minute video on robots.
This would be a mistake because IEEE GlobalSpec has found that engineers on its site have indeed spent more than seven minutes watching such a video even when indicating otherwise in a survey, Lesem said. In short, this fictional marketing department, is not looking at the data for what it is but rather is putting its own conclusions on it before it even ran an initial analysis.
Related Article: Why the Benefits of Artificial Intelligence Outweigh the Risks
Not Ready For AI
That is but one mistake an ill-prepared marketing department can make when attempting to use AI, Lesem said. There are in fact many such oversights, all of which show that it takes more than just merely implementing an AI-based application to have a successful marketing deployment. The truth is, there are many marketing departments that are simply not ready for an AI application no matter how much they want one — or could even benefit from one.
Niranjan Krishnan, head of Data Science and Innovation for Tiger Analytics, offered seven questions an organization or business unit must ask itself to see if it is ready for AI. If the answer to all seven is “yes” then an AI investment would be a good fit according to Krishnan.
- Does your company have business actions or decisions that are or can be driven by data?
- Are those business actions/decisions routine and repeatable?
- Are those actions/decisions time-consuming, expensive, inconsistent or error-prone?
- Are those actions/ decisions made by human operators and/or rudimentary business rules coded into a system?
- Do you have a stable IT and data infrastructure supporting your day-to-day operations?
- Is there a business leader who is willing to champion experimenting with AI solutions?
- Is it possible to assemble a small team (3 to 5 members) of data scientists, data engineers and business analysts, using in-house capacity, augmented by vendors if needed?
Even if you answered yes to all the following, beware, there are potential pitfalls. Here are some signs that a marketing department is not ready for AI, according to Lesem and his colleague, Christian Noe, senior director, product management and marketing at IEEE GlobalSpec.
Related Article: 8 Examples of Artificial Intelligence (AI) in the Workplace
How Does Your Organization Handle It's Current Data
Does your organization handle it's current data well? A marketing department that doesn’t handle the data that it already has well needs to understand what they have before adding AI to the mix.
By now most organizations have a plethora of data coming into their organization, Noe said. “If you’re not able to decipher the data that you have now, slapping AI on top of it will not solve your problems," he said. Another sign is that the data itself is not well defined or clean, Lesem said. Instead, it is in multiple repositories. In either of these scenarios all AI will do is get you a bad answer quicker, he said.
You Don’t Have the Proper Skills In-House
This is not (just) a matter of having a chief data scientist, Lesem said. An AI-ready marketing department will also have subject matter expertise, product expertise and domain expertise. Without such broad based knowledge it is too easy to fall into the trap of allowing AI to reinforce a pattern that you think you already see — such as when the fictional marketing department limited its propensity data to millennials. Another key person to have is someone who can analyze the data or the output and help guide decisions based off of that data. “Oftentimes we see marketing departments not understand what the metrics are actually telling them,” Noe said.
Lack Of Clearly Defined Goals
If your marketing department hasn’t clearly defined its goals, it's likely too soon to add in a layer of AI. Marketers need to understand what its expectations are and clearly define them, Noe said. “It needs to ask itself whether it is trying to better understand sales or market trends. User behavior with regards to interaction with a specific type of content (such as the video newsletter)."
AI-Led Marketing is Still a Long Way Off
Krishnan, likens a company or a department such as marketing run by AI to the Nest Learning Thermostat, which automatically fine tunes itself to every household’s needs. Most companies, he said, are far from getting anywhere close to that ideal. “A full-blown AI system that has attained maturity is like a gear system with three interlocking wheels: data processing, machine learning and business action.”
Data processing is usually the first to be successfully automated, but automating business action and machine learning can be fraught with risks, depending on the nature of the business, he said. A more realistic goal to aspire for is an efficient human-machine operating model with some level of automation of machine learning and business actions, according to Krishnan. That is a doable goal for marketing departments as well — if, that is, they have certain functions in alignment.