At times, it can feel like consumers will say one thing and do another. That’s because, every now and then, that’s exactly what they do.
A few years ago, Quartz published a piece declaring that “market research can no longer predict what consumers will like.” To prove this, it pointed to a Universal McCann study that had found a very low level of interest in the original iPhone in the US prior to its launch.
What this shows is that it can be very difficult for even the biggest global organisations to accurately forecast what consumers will do. Instead, consumer behaviour can often seem irrational and highly unpredictable. For businesses, this can lead to a whole host of uncertainty and, more often than not, mistakes – especially when it comes to indentifying prospects, the direction of marketing efforts and accurately forecasting ROI.
Dan Ariely delved into this interesting area in his 2008 book, Predictably Irrational: The Hidden Forces That Shape Our Decisions, which is summarised by Chris Yeh on the Book Outline Wiki. Among the specific examples Yeh pulls out are the following:
- When Williams-Sonoma introduced bread machines, sales were slow. When they added a “deluxe” version that was 50% more expensive, the original started flying off the shelves; the first bread machine now appeared to be a bargain.
- Tversky and Kahneman conducted the following experiment: when contemplating the purchase of a $25 pen, the majority of subjects would drive to another store 15 minutes away to save $7. When contemplating the purchase of a $455 suit, the majority of subjects would not drive to another store 15 minutes away to save $7. The amount saved and time involved are the same, but people make very different choices.
- Salvador Assael, the Pearl King, single-handedly created the market for black pearls, which were unknown in the industry before 1973. His first attempt to market the pearls was an utter failure – he didn’t sell a single pearl! So, he went to his friend Harry Winston, and had Winston put them in the window of his 5th Avenue store with an outrageous price tag attached. Then he ran full page ads in glossy magazines with black pearls next to diamonds, rubies, and emeralds. Soon, black pearls were considered precious.
The point we’re making here is that, not only is consumer behaviour less rational than we typically realise, but that many of the influencing factors are actually often hidden from view. What businesses need to do, therefore, is learn how they can bring these underlying factors to the forefront in order to gain a clearer understanding of what their customers will do and how to target them accordingly.
In the small-setting examples above, reasons for consumer behaviour became clear because people were looking for them. But what if you have vast amounts of data and have no idea what consumer behaviour cause-and-effect you might be looking for?
In a previous blog we’ve already outlined an approach to understanding and harnessing your customers’ behaviour by employing techniques such as segmentation, predictive analytics and customer acquisition modelling.
So, let’s get a bit more technical – how exactly do we go about finding patterns and trends within the data itself, and why is this important?
To begin with, we split data into what we call “training” and “evaluation” sets. Those are, as the names suggest, for training our machine learning algorithms and for comparing them against evaluative data, respectively.
We fit various forecast models to the training set – like ARIMA, exponential smoothing and Croston – and then evaluate each model by comparing the forecasts to the evaluation set. The models are able to find patterns and trends in the data that people simply couldn’t spot, or that we wouldn’t even consider as being significant possibilities. Using the best model found from the evaluations, we’re then able to make accurate forecasts about consumer behaviour based on various influencing factors.
Why do we do all this? Well, if you’re interested in a success story for accurately predicting customer behaviour, look no further than our work with leading online motoring service Regit.
We set to work by collating their user data, data from their marketing systems as well as data from external sources sich as the DVLA. Our AI-powered platform then ranked the company’s userbase on the likelihood of them being ready to change their car – allowing Regit’s call centre team to accurately prioritise the most high-potential leads. The business is now able to serve its customers in a more personalised and targeted way, with its call centre sales up by 27%.