In the late '90s this client foresaw the coming era of ATMs, gas pumps, and countless other devices in our environment with little screens and printers. They wanted to develop the infrastructure with which advertisers could publish content to this network of screens and printers. The first application was to be advertisements offering coupons through the printer.
They discovered that this led directly to serious interface questions they didn't know how to answer. Who were the people who would be buying publishing rights on the network? With thousands, even millions, of screens and printers on the network, how could they meaningfully identify which boxes they were publishing to?
The solution
In our research, we met with the media planners who would add this system as a new advertising publishing medium to their vocabulary. This revealed serious flaws in our client's business model. For example, our client was assuming a pay-per-click model comparable to web advertising, but this would mean that a successful coupon campaign could have an unpredictable, uncontrollable spike in costs that would destroy media planners' budgeting.
But we also discovered a pleasant surprise, media planners' strong interest in customers' locations, from which they could interpolate customers' demographics. Since we knew the location of each box, we could use standard geographic demographic data to determine demographically who would be exposed to an ad.
As a result, I designed the media planners' system around identifying the target demographic and then optimizing the number of people in that demographic which a campaign could reach within its budget.
I also created some novel reporting displays making it easier for users to see the patterns within the numbers.
What's cool
With boxes showing ads all day every day, and some boxes at some times more valuable than others, we needed a way to set appropriate prices. An auction model wouldn't work; our media buyer persona needed to be able to make a whole set of purchases within her budget quickly and easily. I came up with a model for what units the system would sell, plus a way for their prices to float to an optimized level without disrupting short-term pricing stability.
