I was on a sales/informational call focused on big data in the public sector this past week when the subject of nudges happened to come up. Given that the people on the call were deeply interested in policy development, that shouldn’t have come as a surprise. To anyone interested in public policy, the idea of “nudging” is probably familiar. I personally owe my familiarity with the idea to David McBride, who pointed me to the book “Nudge: Improving Decisions about Health, Wealth, and Happiness” by Richard Thaler and Cass Sunstein three or four years back. Thaler and Sunstein pioneered the concept and have helped make choice architecture a focus of modern public policy debate.
I was particularly excited to have nudges come up on a sales call (a first!) not least because I think it has deep professional interest for anyone doing digital analytics.
The basic idea behind nudging is simple. In lots of public policy (and private sector) scenarios, the way in which choices are presented or selected can have a profound impact on the type of decisions people make. The goal of nudging is to create an architecture that, without sacrificing freedom of choice, helps people make good decisions.
Thaler and Sunstein introduce the idea of choice architecture in their book with a real-world example from a school cafeteria. By re-arranging the order and display of items in a school lunch buffet, a school was able to significantly shift the consumption of certain foods (by as much as 25%). In other words, without removing ANY choices, a different layout of a buffet line could help students choose healthier foods. The choice architecture, in this case, is the layout of the foods.
Thaler and Sunstein argue that “nudges” – systems of choice that are designed to encourage good decisions – are ideologically neutral. Both men are clearly committed to the idea that it’s better to let people choose than to force them to make certain decisions. But they also are committed to the idea that in the real-world, it’s possible to help people choose better. They present a range of examples – more than few of which have become standard public policy – from the world of finance (nudges to drive better default allocations of your retirement portfolio, disclosure for loan fees, and automated allocations for people to save more), health (organ donations), charitable giving (scaled giving programs) and many more.
There are people who object to the idea of choice architecture as “paternalistic” and as someone who is probably closest to Libertarian in philosophy, I understand the impulse. But the objection is largely misguided.
Thaler and Sunstein argue (I think correctly) that setting a choice architecture is inevitable. In other words, you have to layout foods in a buffet in some order. You have to show stocks and mutual funds in some fashion when people are building a retirement portfolio. You have to provide some mechanism for letting people choose whether or not they wish to donate organs.
Take, for example the idea of having people select stocks and mutual funds for a 401K. You can punt on the problem by forcing people to insert symbols one at a time and providing no lists to choose from. But this simply makes them dependent on some other system’s choice architecture – wherever they go to next to find symbols. You could randomly list the stocks and funds, thereby changing the order for every user. But while this would systematically avoid any bias, it gives each and every individual a biased list (there would be strong bias toward the top of list) and has the additional problem of giving nearly EVERY user a set that is biased in an extremely sub-optimal fashion. Bottom line: in many situations there is no avoiding the creation of a choice architecture and attempts to avoid creating a choice architecture will almost always result in extremely poor AND extremely biased frameworks.
Given that some form of choice architecture is inevitable, it’s a matter of deciding what principles will drive that architecture. To Thaler and Sunstein, nudges are features of choice architecture that help people choose what they actually want to do. That’s important, because it is easy to imagine choice systems that are more like “shoves” than “nudges”. If, for example, the manager of that buffet simply replaced all meat and dessert items with Kale and Arugula salad, it might well increase the healthy eating choices (at least a little) of the diners. But it wouldn’t be a nudge. Choice, in that situation, hasn’t been preserved.
At my daughter’s school, they’ve made an aggressive effort to only provide healthy eating choices in their cafeteria. Along with poor execution, this drives many kids to eat nothing or to simply bring snacks from home. The creation of this type of “black market” or widespread avoidance behavior is generally clear evidence that you’ve missed the mark in building your choice architecture.
From all this, it should be clear why the concept of nudges is interesting to policy wonks. It’s a concept that is as appealing to Stephen Harper as it might be to Barack Obama. Conservative and liberal can find common ground in making public sector systems better.
So why is this interesting from a big data, digital analytics perspective?
I think there are two reasons – the first is more about big data and the second more about digital analytics. Both are intriguing.
From a big data perspective, the construction of a choice architecture is largely a data driven exercise. To find places where nudging is desirable, you might look for systems where people’s actual behavior falls significantly short of their intended goals. There are a broad swath of activities (from retirement planning to health insurance to organ donation to charitable giving) where large gaps between intent and action persist. Finding these systems is exactly the kind of analysis that the Customer Intelligence System I’ve written about this year supports.
Problem identification is just the beginning, however. The choice architect is quite often faced with the problem of how to build proper defaults. Creating these “easy” choices is key ingredient in most nudges. But the right default solution is rarely obvious. Analytics provides a way to identify likely optimums based on actual choice behaviors. This is a paradigm case of big data analysis. It almost always requires large amounts of actual behavioral data, supplementation with unstructured VoC data, and complex analysis to identify likely optimums.
Since the folks on my call were policy oriented, I think it’s this type of big data use they had in mind. But to me, the application of “nudge” and choice architecture based thinking is directly applicable to digital analytics.
I’m thinking most of my readers can see why without even a nudge on my part.
It should be pretty obvious that a Website (every Website – whether public or private) instantiates a choice architecture. In the navigational layout, the placement of content, the selection of links, and the options for fallback navigation (e.g. search), a Website makes countless nudges and more than few shoves. In my writing on Topology analysis, I showed how the internal structure of a Website defeats traditional statistical analysis of correlation and I compared Websites to a road network – a combination of highways, boulevards, avenues, streets, lanes and alleys with easy access between some points and nearly impossible navigation paths between others.
I still think that’s apt, but it’s also apparent that what’s captured in the Website is a physical manifestion of a choice architecture.
Thinking about a Website from the choice architect’s perspective has a couple of real benefits. First, it puts us in the frame of mind of thinking about whether the created access paths instantiate proper nudges. Many times, they don’t. And that’s largely because Websites (particularly public Websites) tend to get designed by committee and without a clear focus on constituent choice.
Second (and here the concept extends Thaler and Sunstein’s work), I think it forces the choice architect to think about segmentation and personalization. No site is more admired or frequently cited than Amazon, and rightly so. From a choice architect’s perspective, they’ve solved a massive problem with an elegant nudge system. Their problem – how to present millions of possible book choices to a consumer – had previously been solved either by choice reduction (showing only best-sellers) or zero-guidance (internal search). Their recommendation engine created a system of nudges that enhanced choice without restricting it. What’s compelling about the Amazon solution is that every single visitor gets a different set of nudges.
The Amazon experience isn’t unique; it extends into many digital problems. It should be obvious that the optimal choice architecture is almost never going to be static across a broad population. In the vast majority of public policy discussions, the focus is on developing a single choice architecture for all constituents. These single architecture solutions will never be anywhere near optimal.
Digital provides a means of delivering flexible, personalized choice architectural solutions that are simply unavailable in traditional public policy. Digital can deliver solutions that are highly tailored to deliver optimal choice systems to every individual. That’s new ground in public policy and it’s a re-thinking of why digital is important in today’s public sector world and how it can transform the solution set available to the creator of public policy.
It’s also the place where the big data theme and the digital analytics theme come together. Big data provides the mechanism for creating data-driven, highly-personalized nudges. Digital analytics provides the mechanism for optimizing and incorporating those nudges into the customer experience.
Much like the concept of nudging itself, bringing these two disciplines together is a win-win that we can all agree on.
