What is the most representative way to measure the impact of a digital signage deployment?”
I suppose we should take a “current” perspective since technology may allow for more possibilities in the future. I suppose we could contemplate a world where face-detection algorithms can be combined with social data, which can be cross-referenced with big data sources to ultimately conclude if someone, in fact, bought in to your product based on the sign they saw at the corner of Las Vegas Blvd and 5th. If we could do this, we could get the real metric (direct sales) produced. However, short of the future of big data, social integration and politically tricky privacy issues, I suppose we should stick with more primitive and “current” capabilities.
In my world, we keep a lot of big data at very detailed levels. I recommend starting there. You’ll need at least some kind of history (the longer the time period, the better) to compare. Don’t let I.T. tell you that they need to purge historical data for cost savings or risk mitigation, because, if you let that happen, you’ll find yourself unable to measure lift when looking for long-term trends. You can certainly survey people, but watch out for sketchy memories, personal biases and multiple interpretations (our last election certainly shows some weaknesses in surveying). It’s really better to look at objective results. Did sales or impressions rise?
With our interactive signage, we are fortunate enough to collect enough data points to actually triangulate direct sales attribution in some cases. However, I know this is not the norm as most digital signage isn’t directly connected to such data points. I think, for most cases, we’re looking at sales lift for the products we’re advertising and correlating it by time and location of the signage. If “sales” seems too detached, go for “impressions.” For example, check for a lift in “visitors to your web site” from a specific digital signage location. Check for quantities of that product sold since you put it on the signage, and check for times it was sold and relate that to times it was displayed. Test different timing/product scenarios and track it. One of these methods, when related to something (hopefully history), is going to show either sustained business (just as important as increased business sometimes) or lifted business.
And then, when you start seeing the results you want, think about algorithmically making timing/product placement adjustments for real-time optimization!