Not often do we find digital marketing agencies embracing complex technologies such as machine learning and personality insight awareness to assist in generating new sales opportunities. Well, not until now anyway.
We exploited our cross-functional teams to trial new innovative technologies such as machine learning
The technology team understood the future of cognitive technology whilst the marketing team understood the concepts in appealing to consumers and ensuring the message is being conveyed correctly.
But what if we could bridge the two disciplines together, and try something new. What would be the results, and can we realistically embrace the technology to make a true impact on our customers' marketing efforts and ROI? So we conducted a test.
Test: Utilising Machine Learning, Personality Insights and Facebook Ad Campaigns To Create Something Special
Part 1: What Data Can We Provide
To get started, we needed to categorise how our customers speak to us and what their personalities reveal. Is there an essence of emotional attributes during sales enquiries or were they more excited when talking to us about their needs? These questions are important to understand so we can ensure we are speaking to them in a way that appeals to their personality traits, making them more comfortable that we are actually listening and creating trust.
So we grabbed as much data as possible across a range of customers, grouped them and using IBM Watson cognitive services, we fed it into the machine. What we received was 20+ personality characteristics allowing us to average them out and produce a top 4 list (which we graphed), resulting in;
Emotion - They were speaking to us from the heart, considering we help clients across a range of services this made sense, as majority of the conversations we had were in regards to the businesses they own.
Openness - Our customers were open with us about their expectations and what they were looking for. Majority of the discussions we analysed the client was upfront with us and expected this in return.
Structure - We learnt our customers are talking to us in an organised and structured manner. Usually getting straight to the point and expecting this in return.
Practicality - Clients would often refer to previous experiences or methods they use to conduct business, often avoiding theories or ideas.
Part 2: Build A Benchmark
We reviewed a set of existing Facebook ads (that have been boosted over 3 days), and took note of the performance. For this experiment, we focused on the most recent ad run before performing this test. We took note of the reach, reactions, likes, comments and shares, this would be our benchmark which we will attempt to improve upon using cognitive technologies.
Part 3: The First Machine Assisted Ad
Using the personality traits analysed in part 1 of this test series, we recreated a new ad that would appeal to the top 4 personality traits identified, using different string of recommended text. We worded the ad based on the feedback provided by IBM Watson personality insights and posted to Facebook. We boosted the ad and let it run over the same time period. It’s important to note here that the intent of the ad was to produce awareness as we did not add any call-to-actions.
Part 4: Human and Machine Assisted Ad Development
The increase in brand awareness speaks for itself, next we needed to encourage commenting on the post in the attempt to generate leads and reduce emphasis on reaction. We setup our next ad, followed the same recommendations as personality insights provided but added an element of “excitement” to encourage the customer to either comment or share the post. By encouraging the customer to comment, it gave us the chance to discuss the product, answer questions and direct them to appropriate marketing content.
Summary: We Thought It Was Interesting
The end result speaks for itself. We were running broad campaigns across the social channels that were producing little return and in most cases appeared to be ignored. Utilising machine learning technology including personality insights, we were able to discover more about our customers and how they speak to us, making the assumption about how they would like to be spoken to in return. We tested the theory against social advertising and learnt how to increase brand awareness vs. engagement. I want go into the sales pipeline and how much it benefited from this experiment, but whilst experimenting with these theories, we saw 6+ leads enter our marketing / CRM engine within the first 4 hours at go live for ad3.
And here's the kicker… we did this all on a train ride home within 30 minutes.