they give examples of alcohol in beer and sugar content of foods.
experimental specific:
alcohol lies in the vis spectra near what we would call yellow. beer is often yellow, but not due to its alcohol content. beer is yellow because of the malt used in it's creation. reflective vis. spec. cannot differentiate the two, and would always give a vastly false reading.
practical specific:
Glucose lies in the near ir (and has been known to for a long time), so this device could theoretically be used measure glucose levels in diabetic patients. a potentially multi billion dollar industry. So why would we worry about trivial things like sugar content in food (that could be obtained from reading the food's label), if we could instead be helping millions of people and making tons of money while doing it?
...Simply because it doesn't work well.
Sugar, in wine is often assessed transparently (absorption spectra, as reflectance is quite difficult to get clear signals due to surface shape, and air surface interface on incidence angles etc...) via polarization spectroscopy. there is visible spectroscopy, mass spectroscopy perpetrated spectroscopy... this does one, and in a narrow RF range. And in our studies and research, isolating one spectral signal from another was extremely difficult and required an enormous catalog of known solution variants to build our model.
This doesn't even mention the technological issues of calibrating, and certifying light sources, varying saturation time of your ccd array, warming up a light source (LED is a weak spectroscopy source as it is designed specifically to be narrow band).
these things come out once every few years and almost always fall flat. I'm sure kickstarter is littered with these things.
One use:
Could be used with things that have reagents added to expose the analyte more clearly, and then to assess the degree that analyte presents itself using the before reagent added as a baseline.