Validated primary and secondary research inside the cognitive marketing mix
There are high expectations towards behavioral data these days. Any move your customers make can be measured, right? The “when”, “where”, and “what products they buy in the same shopping cart” tell us nearly everything we need to know about them. We can contextualize POS data with journey data, trending news topics, and weather data. We can also track consumers across all digital channels and increasingly on offline channels by weaving usage eco-system nets around them. We know everything that happens — the “what”.
Yet, we don’t really know the “why” — like why it happens, unless we combine the sensor, tracking, POS, and contextual data with in-depth knowledge about the beliefs, problems, and perceptions of consumers. This includes knowing what the conscious or unconscious decisions are, which make the facts happen to allow us to track the “what”.
Our beliefs in abstract, statistical models are based on the assumption that we have a complete picture of the world. However, these complete pictures are almost impossible to achieve simply by looking at what happened, because they were predetermined by what was measured, and what was assumed to be important by the experts created the models.
Defining what is measured is called feature engineering in machine learning and cognitive computing, and is essentially a template or lens that is being applied to reality in order to filter out what’s important to achieve success for a certain task. Humans have these filters on various levels, ranging from sublevel senses that filter out noise unconsciously to high-level, cognitive filters that as we’ve learned over time tell us WHAT works well, helping us to achieve success. However, there are problems that these high level filters in human cognition introduce, which are:
- Learned by individuals, based on the entire context of their lives, ranging from the physical, to the experiential, to the emotional and beyond
- Highly situational in nature and not statically applied
- Only used for decision making — they do not define what we know about a situation. We know far more about a situation than we need for a specific decision: the features do not limit our capability to judge a situation
In cognitive computing, a data interpretation using a defined set of features significantly limits the ability to judge a situation and infer the right decision as it models a very narrow perspective of the world and of an individual customer in that world. Looking at the “what” with a predefined set of features is like ignoring 80% of what is there; it’s like looking at the world through the eyes of a horse with blinders.
At Market Logic, we firmly believe in the human interpreted “why” to complete the picture and context of the “what”. Your research agencies have many years of experience interpreting the “why”, because they look at every single problem through the trained eyes of human beings. They produce interpretations that are not limited by a predefined set of features. All you need is a system that helps you listen and understand without drowning in data.
Market Logic’s insights platform helps you to harvest both the insights from the “what” and the “why” with unprecedented ease and to produce your unique 360° picture of the problem you want to solve — as described by Harvard Business Review.
Originally published in Market Logic blog.