Nearly all of the discussion of the "big data" that we are told can be amassed and analyzed by monitoring consumers' online behavior seems to center on the possibility of using the results to predict what people are going to want at a particular time and place, with the goal of launching an offer that will intercept them under precisely those circumstances. Perhaps this will work, if it is attempted. To date, in our experience, its principal use is to set a flag when someone purchases a pogo stick online, and then start displaying ads for pogo sticks on the greatest possible number of sites visited by the hapless customer.
We are told that the increasing capabilities of digital imaging systems and pattern recognition software can bring a new dimension to the collection of this data by encouraging people to interact with technology at the point of sale. Pilot projects for doing this with vending machines have received considerable publicity.
We think it also could be put to good use by taking a new look at the age-old problem of finding out just what a vending patron does when approaching a machine.
We have mentioned this problem before, and it covers a number of behaviors. First, vending often has been thought of as making "impulse" sales, although many people who work or study in a location where vending is the most convenient source of food and beverages certainly are not buying on impulse. Related questions include whether patrons know what they want to buy before they approach the machine; whether they will buy something else if they don't find their first choice; and, generally, whether they can be induced to approach the machine in the first place.
There has been surprisingly little research done on these subjects. A memorable one was conducted nearly four decades ago by two graduate students at Purdue University who proposed to conduct brief interviews with a large number of vending patrons. The late Dr. Joseph Cioch, who supervised their studies in dietetics, worked with the Indiana Vending Council to expedite the project, and the results were hailed by IVC and the National Automatic Merchandising Association as breaking new ground in market research.
Several decades later, Nabisco (later acquired by Kraft Foods) also used multiple short interviews in its pioneering investigation of vendible snack categories. Again, the results of the research attracted widespread industry interest.
This approach is by far the best way to find out how people interact with vending machines in their environment, but it is very labor-intensive. Industry market researchers put it to good use, but they are looking for specific information. An astute client is willing to pay the cost; it is more difficult to raise money for general industry research.
This is where we think the new interactive tools, such as the ability to analyze facial expressions and interpret gestures, can make a very positive contribution to understanding how customers relate to vending machines as retail outlets. The information collected would be specific to the machine's location, and also could be compared with data sets collected in other sites to look for patterns characteristic of particular location types -- and of consumers away from home.
When a vending machine is available, how many people walk right by it and how many stop? Can moving the machine, or changing its display illumination, or adding some kind of signage increase the number who will stop and look? As people approach the machine, where are their eyes directed?
This kind of information, collected automatically at many locations, also would provide detailed information for addressing the differing theories about product placement that have been floating around for decades. Do people "shop" a glassfront machine starting at the upper left, or in a "T," scanning the top row and then sweeping down the center, or in a "U" pattern, or something else altogether? Is it better to keep core products in the same predictable slots, or to move them around so customers will look for them and, perhaps, see other items that they hadn't considered? (We suspect that different patrons use different methods, and the same patron may use different ones depending on circumstances, but we just don't know.)
The folk wisdom related to arranging machine displays may seem less relevant as operators make better use of automated data collection and sales analysis to identify the fastest- and slowest-selling products. As valuable as these tools are, they have nothing to say about factors affecting patrons' perceptions that may contribute to the appeal of a fast-moving item (or, for that matter, detract from the potential appeal of a slow-moving one).
We think that there are excellent reasons for building this kind of analytical capability into the software developed for interpreting behavioral cues at the point of sale. The ongoing delivery of information that operators can use to improve overall sales would add persisting value to a technology that already offers the prospect of enhancing the appeal of vending to product suppliers, to everyone's benefit.