We’re building systems that make it possible for computers to do what people would normally have done. And that stretches across our entire value creation process. Automation is an area in which TNS is leading the way within the market research industry.
TNS collects information, which it transfers into useful marketing information that clients can use to make strategic management decisions. The market research company is making use of automation at each step of its value creation process, resulting in more relevant surveys, a much smaller margin of error, a reduction in cost and all-round greater efficiency.
The process begins with the actual collection of data – making decisions about what to collect and how to collect it. It is possible to set up an automated process to pull required information from a website such as Twitter, by having a computer monitor all activity on the site and dumping any mention of a particular brand into a data store. The important thing about that kind of information is that you don’t have to ask questions to get it – it’s just there; you just need to have a system to collect it.
Of course, the more traditional side of data collection continues to revolve around asking people questions. There are all sorts of ways in which human beings would traditionally have been involved. You’ve got to know which people to ask, you have to script the questions. Increasingly, all of that is being automated. In terms of knowing which people to ask, the example of China’s population is useful. By speaking to the right 2 000 people, it is possible to predict the behaviour of the 1.4 billion-strong population within 1%.
A significant sampling challenge, but one that can be overcome by programming computers to select the right respondents. In addition, by making use of computers, surveys themselves are becoming smarter. This is because each question can be dependent on the answer given to the previous question, resulting in more efficient, more relevant surveys that get to the heart of each respondent’s context. Hofmeyr points out that this would be impossible to do with a regular survey, in the hands of a human being, as the calculations required are too complicated to be done manually, and would certainly be subject to human error.
Data collection is only the first step in TNS’s value creation process, however. Probably the area where the market research industry has been most focused when it comes to automation has been the transformation of data into useful information for the marketer. Simple automation here means that instead of requiring someone to look at the data, turn it into more meaningful tables and use these to develop marketer-friendly visualisations, the automation goes direct from data to visualisations. This is a lot cheaper and requires fewer people.
While a human must still specify which numbers should go into the visualisations, it is then up to the computer to pull the numbers and put them into a graphical layout. But where automation really takes off is when the automation doesn’t just stop with visualisations. The computer write actual, grammatically correct reports – without human intervention. You can either have quite a rigid system where there are no people involved at all and it just goes straight to the graphical layout and then to the client, or you can have an interactive system where it’s really easy to modify what you’re seeing and to call for different written reports.
The automation of the entire market research process is exciting. However, the development of software that writes actual, grammatically correct reports based on data collected, is certainly the most ground-breaking innovation – and one that he believes TNS is making use of particularly effectively.
While there are three companies that turn data into basic reports – Narrative Science, Automated Insights and Intellection Software – the software that TNS uses was developed by Intellection Software, which I co-founded along with several young South Africans. It takes the numbers, draws the pictures, looks at the pictures and instead of having a marketer or market researcher figure out what it all means and write the report with marketing implications, the computer does it automatically. The software is also not rigid, but flexible and intelligent, meaning that if the data changes, the computer picks this up and produces a new report automatically. “The beauty of it is that machines do a better job than human beings because you can get an expert to write the software, whereas manual market research reports are variable be- cause some people are really good at it and others are pretty ordinary. This removes the ordinary.
However, while automation removes the human from several steps in the market research process, it is not the case that humans are not needed at all – they simply need more training and a higher skill set. At the very start of the process, people with marketing and market research understanding are needed to write the programming. Further down the line, the human touch is also needed to understand the marketing implications the software provides. This is because the machine will highlight all the consequences of what the data is saying, it will outline the different strategies that are open to you, but it won’t tell you which strategy you should take.
Jan Hofmeyr is chief research officer of behaviour change at TNS.
