David Wasserman’s 2014 article (Senate Control Could Come Down To Whole Foods vs. Cracker Barrel) of using restaurants as a proxy measure for the people who live near them is relatable: I used to rely on Trader Joe’s to identify neighborhoods I’d enjoy living in when I was moving to a new city. I got the idea from a friend who mixed credit card transaction and GIS data to sell maps to companies looking to find ideal locations to open new stores. I guess I like jokes on my food labels and the people who do, too.
The classic example of a proxy measure is per capita gross domestic product to measure quality of life. It makes sense because money per person up to a certain amount is correlated to happiness and implies relative access to food, housing, education, medical care and luxury. Yet if you cut down the redwood forests, or start a war, or rebuild after a natural disaster, GDP goes up independent or counter to effects on quality of life. In recognition of the deficiencies of GDP-focused policies, Bhutan created a Gross National Happiness Index. And from OPHI’s summary on the GNH Index I definitely want to explore this for a data science project at CUNY’s MS in Data Science program.
I recently applied for a job at a PR firm in their Global Intelligence Unit, and they shared they have so much information related to how many people view their advertisements, how many viewers click on their advertisements, down to which pixel they clicked, but can not definitely say whether a purchase was completed as a result of that click. Instead they use clicking on an advertisement as a proxy measure for making a purchase. This is a good proxy in that they need some metrics to report back performance of an advertising campaign to a client; However it’s not complete because it doesn’t show actual purchases. A surreptitious campaign could generate a lot of clicks of tricked, unmotivated buyers, but then the normal proxy measure (say 10% of clicks) would be way too high and the PR firm would be overselling their impact on sales. As a side note, I find a lot of promise in the pixel-based metrics. You could provide advertisements with several choices and use which choice the person clicked to generate data about the individual clicker.
When I think of a proxy measure with application for my personal life it would be the number of times you independently reach out to a friend as a proxy for how much you value that friend. I’d like to build an app that tracks how often I call and text people and have it create an affinity network of the people I care about. I could then add people to the affinity map, or say whether I would like to be closer or less involved with them, and the app would update it’s dashboard with suggestions of who to reach out to. Hopefully I could build in a metric for not being a stalker if the friend never texts you back but I think it would be great way to remember to reach out to the people who aren’t in your immediate sphere of affection. I’d name the app Foxy Measures. Can you picture the cute animated fox chat bot measuring your emotional propinquities?