As cognitive computing platforms are becoming increasingly familiar in the marketplace, a usual conversation I have with clients revolves around the following two questions:
1. Where is the differentiation in cognitive computing platforms?
2. Am I ready for cognitive computing?
That second question is the one I pose after we’ve discussed the benefits of cognitive computing, or applying machine learning to data.
Associated with this latter question is the level of insight an analyst, a marketer, or a publicist, is able to work with. So I want to briefly touch on levels of insights as a topic to determine how ready an organization may be for the investment (time, people) of taking on cognitive computing as a resource.
Insights are, in large part, subjective. Take for instance the identification of a new competitor in the market. X competitor appears to be active in MENA. To one observer this may be considered as a meaningful insight, as they now have context for a competitive landscape. We’ll tag this example as a Level 1 (competitor) and level 2 (where they compete) insight. Now, this would beg the question, what would be a level 3, or a level 4… a level 7 be?
This is where deep learning can take you. Level 3 could be identification of the product being sold; product X. Level 4 could be an understanding of the conversations driving need for the product. Level 5 may address stated vs actual needs being addressed by product X, and so on. One platform can deliver various levels of insight that traditionally would require several vendors and research methods, but can one analyst act on these levels of insight? Can your publicist react to a real time discussion about how your product is inferior to another because the ingredients may cause skin irritation on two different marketing channels and with enough context to appear informed and inform others while doing so? Do you have capable copywriters with an editorial strategy to address product and industry trends in real time, with content approvals processes in place, to distribute content and engage audiences on various channels?
What we arrive at is the idea that an insight needs to be actionable, and while that’s a term widely used, I don’t think it’s understood in its entirety. It’s not actionable merely because it provides deeper levels of context, it should be actionable because you have a process designed, perhaps a governance strategy, and a digital toolset to help you act on it; to write the editorial response through various channels at once, set an email alert that drops to a colleague in a different department and geography the moment something is flagged as a significant insight from forums and blogs in a specific geography in southeast Asia, etc.
Other things to consider:
· Do I have capable analysts that will know where to look for insights?
· Do I need a data scientist to run a classification engine and teach the machine to bucket accurate associations to topics of interest?
· Lastly, do I have a good partner (vendor/agency) for this?
The simple answer to understanding whether your team or broader organization is prepared for machine learning or DL, lies in your ability to act on the insights derived from these practices. Otherwise, it may be costly, time consuming, and leave you with a so what reaction to something that allows you to dive a little deeper into data.
The upside? The learning’s I’ve seen with DL can be remarkable. Imagine using only one or two data sources (twitter, and TV/Radio) to get to a level 7-type insight, in real time?