The net present value of market research investment is silently, and profoundly changing.
The Status Quo
Market research has existed as an industry since the 1920s. Over the last 100 years, there have been many innovations. For most of that time, however, the intellectual property (IP) produced by market research has remained relatively unchanged.
We traditionally think the outputs are valuable from market research projects: questionnaires, discussion guides, data files, transcripts, reports, charts, executive summaries, descriptive analysis and dashboards. Clients traditionally own the created IP they pay for (i.e. pretty much everything above). If organisations buy subscription insights platforms (e.g. Software as a Service) they often come with templated content and reports which are then licensed by the client, although they can also be used to generate new outputs, as described above.
The IP generated from market research is valuable but remains static knowledge: that is, the value created relates to the circumstances at a specific time. The work has a ‘use by’ date. This knowledge (mostly) has a shelf life, and its value degrades over time. This is why we don’t see a thriving secondary market for syndicated research reports from 2010.
Enter ‘AI models’
Market research certainly yields value, but we now live in an increasingly AI enabled world, where AI has forged new ways to create value. AI generates a ‘model’, which is an algorithm or set of capabilities that delivers insights or functionality for human interaction and engagement. The model learns from data, it is dynamic, and it improves each time it is used. The difference here to the status quo described above should be clear: AI models are dynamic and age well, whilst human generated research is static, bound to context and only as good as its researchers.
Our conversational AI technology EVE (Evolved Verbatim Engine) is an example of AI technology that produces dynamically learning models. These models work across projects, delivering better insights over time. The more EVE talks to humans, the better she gets at having conversations from deep learning powered text analytics. This in turn means more insights generated from project to project. This becomes a significant strategic asset.
Our technology is one example of a broader principle. Imagine the competitive edge for a client organisation that develops a superior AI model in its category? Or the edge a research vendor gains by applying its own AI powered insights models across client projects? These models will become the true value of insights that emerge from data rather than the actual data layer that sits below. Given the data and the model are intrinsically linked, however, have we paused enough to think about who really owns the model in this world?
The long game
The current legalities and IP rules around this concept are not surprisingly, still quite murky and can differ across different countries. The key concepts include authorship versus ownership, ownership of source data, and the level of connectedness between the source data and the model.
For example, questions arise about ownership: if a model is trained with a data set: does the data set owner, by default, have some proprietorship of the model because it is created from their IP and hence is an abstraction of it? Is it the programmers of the code that developed the software that did the heavy lifting? Do the vast number of human subjects that enabled all this by providing their own data, deserve at least a share of the IP?
Similar important questions also manifest when considered through the lens of our industry structure – buyers serviced by suppliers. When a client buys access to an existing model, what exactly are they buying? And who gets what out of the bargain? For SaaS products, buyers license the technology, but do they own the generated models as well as the data? Do you own the training model or does the platform? Who can leverage it in the future?
Overall, the underlying question is about value creation. How is value ultimately created and who owns it now and in the future?
These are just some of the important questions we think need to be asked by clients and vendors alike.
The net present value of knowledge – changing on multiple levels
Consider the net present value of the insights yielded to your organisation right now from current research practices. Now, consider the net present value of a model you train over a few years, its value compounding as it gets more efficient and intelligent over time. Evidently, it is much more important to own the model and have choices of about how to license it even across your vendor base. This is even more critical if the model uses “your” data even if it is created through someone else’s technology. For research vendors using tools that can create models, suddenly this concept becomes critical. Traditional research consultancies don’t have to own the generating technology if they can develop their own defensible IP using someone else’s technology. For example, predictive analytical AI focused on finance, healthcare or retail sectors. Now consider what can be achievedin non or pre-competitive contexts such as one of the Alzheimers research collaboratives.
We know these business models are emerging because we have created a platform powered by learning algorithms, and we see our research partners using them in this way. Our partner research agencies develop their own models using EVE. They focus on specific industries and their own experience/expertise to build valuable IP. Our clients are also doing the same when they license. We see an opportunity to sub-license the use of these models to others – thereby retaining the model IP while multiplying its value over time. The two systems work together within a common framework and platform, one in the near term, one amplifying value over time.
We are seeing new value like this enabled by AI that was not possible a few years ago. Organisations need to be clear on what they are buying, smart about how to generate value and how to leverage a competitive edge from it.
