Nudge and Sludge: A Conversation with Dilip Soman

by Dilip Soman

Last updated 20 July 2026

ETHOS Digital Issue 05, Nov 2019

A person in a white shirt sits on a grey chair against a split black and turquoise background.

1. In the context of behavioural insights and nudges, what is “sludge”?

The term “sludge” is new, but the concept goes back almost a hundred years, to psychologist William James, who wrote about how people’s ability to get things done is often either impeded or facilitated by what he called the context or the environment.¹ Little things that we don’t see or don’t expect to be relevant can impede or facilitate decision-making. The work done by Richard Thaler and Cass Sunstein on nudge theory had to do with trying to bring down these frictions and constraints.

Different kinds of impedance get in the way of people achieving their goals. For instance, there are procedural impedance, the way procedures are designed—to purchase products, cancel subscriptions, or apply for a welfare programme—can actually inhibit decision-making. There are also informational or perceptual obstacles—the way information is presented may be confusing, when attributes are shrouded or we don’t truly understand what needs to be done. Sometimes the process itself is efficient, but there are outcomes of the process that make people not want to engage. That is sludge.

In Canada, government welfare programmes are oftimes intended to help low-income Canadians—many are new immigrants who don’t know the language or don’t follow the same cultural norms, and they are reluctant to actually go to a welfare office to sign a bunch of forms to get cash. There are also examples in Latin America, where housing subsidies are based on how bad a state one’s house is in are often underclaimed—because people are embarrassed to say how bad their houses really are. These psychological barriers are also sludge.

Part of the problem is because there is a big empathy gap between people that design the process and the end user. Unless you have actually been in the situation of somebody who’s about to receive a subsidy, you can never anticipate the guilt, the embarrassment, the shame, the loss of face. So we need to be particularly sensitive to these different ways in which sludge might show up in a system.

Sometimes the process itself is efficient, but there are outcomes of the process that make people not want to engage.

2. Is the friction associated with sludge always unintentional?

There are distinctions we need to make. Is there something about the process that facilitates or inhibits decision-making: is there friction or no friction? Is it for a good or bad outcome?

If you are intervening to reduce friction and it’s for the greater good—say to help people file taxes on time, what Thaler and Sunstein were writing about in Nudge,² in 2008—that is nudging for good. But there’s also nudging for bad. Sometimes we find for-profit coporations using some of these ideas to get people to purchase things that they don’t really need, for instance.

Friction is a lot more interesting. I think a lot of people will say that friction in the system is always a bad thing, but I would actually disagree. There are some cases where friction is good. There are times when we want people to be thoughtful, just like how we often impose a cooling-off period, so people can sleep on a decision before it is finalised.

As we move to a cashless society and people use more mobile or digital payments, it becomes easier to pay and they tend to spend more. Now, what if everytime you tap your phone to make a payment, there was a 30-second pause and it says “You’ve already spent $650 this month, are you sure you want to make this purchase?”. This is friction, it’s impedance, but it’s for good. Likewise in negotiations and contracts, we often impose cooling off periods—a friction in the process that allows both parties to think deliberatively about the outcome and back-out if they feel they have reached a poor outcome.

Of course you get to the fourth instance where there is friction for bad—now this could either be deliberate or not. For instance, a magazine may make it very difficult to cancel a subscription. That’s deliberately imposing a friction to try and get people to not cancel, or to not stop spending.

But oftentimes, bad friction happens because we just haven’t thought about the processes in our systems. We just haven’t given thought to whether, for instance, a person might feel embarrassed, or whether our process takes four steps when it really only needs two steps. In the old days of the Palm Pilot device, there was a “Two-Tap” person whose actual job was to make sure there wasn’t anything you couldn’t do with the Palm Pilot in just two taps of the stylus.

I think we need more “Two-Tap” people in government. We need people to look at every single process and ask: do we actually need four steps, do we need to wait for a month, and so on. Often, with legacy systems, nobody has actually looked at the processes and their defaults for a long time, so we end up with a lot of sludge in the system.

Often, we don’t even see the friction, or we don’t realise that something is friction, or impedance as we call it. If the impedance actually helps people make better choices by making them mindful, that’s great. But when you design something to slow people down when it’s not in their interest as they see it, or when you don't even think of it as sludge, such as when you have a legacy system or different processes joined together, that’s when it becomes problematic. I think we need to be more aware about looking out for it.

3. Aren’t most regulatory functions, such as quotas, restrictions or security measures, sludge or sludge-like?

They are impedance for sure, but I’m not convinced that they are all sludge, in the sense of being negative. For example, privacy protection requirements do have positive features. We want people to think about sharing data online, or to be careful about the kind of services they sign up to. They may not think that it is necessary, but it is for the greater good.

Now having said that, what I think is happening is we design system in silos: so when you look at somebody who is actually applying for a welfare programme in the US, for example, there’s one department that takes applications, there’s a second department that does the processing, there’s a third department that actually pays out, and so on. If you look at the healthcare system in many countries, it’s exactly the same story. Often, the consequence is sludge. We now have the data and the ability to track customer journeys—we can go back and see whether we really need some of these extra steps. When we already have citizen data, why do we need to ask for it again?

Another important thing to keep in mind is that, while process is an important element, communication is also an important part of it. I think we often have this notion in government (this is certainly true in Canada), that the more information you give people, the better off they are. As a result, we communicate in volume and not impact. It can look like government is trying to obscure information, but it’s not: the information is all there. We know people have now shifted in the way they communicate, yet we continue with our traditional way of communicating. We can end up demotivating people with too much information.

We often have this notion in government that the more information you give people, the better off they are. As a result, we communicate in volume and not impact.

4. What looks like a nudge to one group might look like sludge to another. How might differences in cultural norms and values affect the way we engage the public?

I grew up in India, which ranked high in bureaucracy. My parents and their generation, for example, have a notion that unless it’s a complicated process, it isn’t a robust process. So when they came to Canada and saw me file taxes, my dad used to always say “Are you sure you've done it right?” Because it seems too simple a process to him, without even a photocopy or scan of anything. So sometimes, cultural differences are just differences in norms or expectations.

Singapore is a lot more like Canada in terms of cultural diversity. In Canada, everything the government does needs to be in English and French, but there are many other languages as well, so everytime we think about a process, we try to think about it from the multilingual point of view. Communication doesn’t just mean putting out the same information in multiple languages—in different cultures there might be norms such as who in the family I should address the communication to. In trying to encourage some low-income Canadians to sign up for bank accounts, we have to think about whether we should encourage them to go to the bank, or have the banker go to them, because there are some communities where people don’t go to banks—their banking is informal, and they are not used to physically taking their cash and handing it to a stranger in a suit in a large building.

We have to make sure we are aware of the cultural sensitivities, it’s the little nuances in how people might react, that can make a difference—such as the potential for embarrassment, wanting to save face and so on. It takes time and discipline to do this right, but we have seen dramatic differences once we’ve done this.

We have two approaches to trying to develop and customise these sludge-free processes. One has to do with using large data sets to infer behavioural patterns across either geography, cultures or communities. If we have a data-rich environment with which to work, machine learning works really well. It doesn’t tell us why things are going on, but it does tell us that there’s something going on. It tells us where the hot spots are. Then we can go in and do more ethnographic or observational studies. That’s one approach.

The other approach is that there really is no substitute for essentially walking in the shoes, not just of citizens, but also of frontline service providers. For instance, we have started thinking about ways we can collect observational data from our frontline service agents. For example, in banking, we look at whether a family comes to meet a banker on their own or all at the same time; is it husband and wife together; who goes in first; how do they communicate and so on. These are not often scalable research projects, but in conjunction with the big data, we might notice that particular people are behaving differently from expected in some way, and it lets us build up a hypothesis. Then we try different approaches to see if it works. We start small and over time scale it up.

Machine learning doesn’t tell us why things are going on, but it does tell us that there’s something going on.

5. What’s an example of a well intended policy or product that had a different impact because the impact of sludge was not understood?

One example that jumps out is the Canada Learning Bond. To encourage low-income Canadians to send their children to schools, the government gave out money to cover textbook expenses. To receive this money, parents had to apply for a Social Insurance Number (SIN) as soon as their child was born, so they could get an account to receive the $2000.

When the policy was launched, the takeup rate was poor—under 50%. It turned out that in order to get a SIN, you would have to go and apply for it within the first three months of a child’s birth. Things have changed now—when a child is born, there is a set of forms that covers all the information needed, including an application for the relevant numbers and accounts, and now the forms are going online. That’s a last mile solution.

Now, how about a first mile solution: why do we even require people to apply for the SIN in the first place? The government does have some good reasons, such as verification of citizenship, but we could promote the idea of every child in Canada being born with a tentative SIN, subject to verification. The fact that the child already has a number I think creates the motivation for parents to actually go and get it verified. It’s an endowment effect: you feel you have it in front of you, but you have to do something to keep it.

Another example is an interesting backlash effect we have noticed about factsheets and product tables. Sometimes, a well-intentioned policy intended to give people more information may have an unintended effect of having people focusing on attributes and data that they shouldn’t rely on to make their decisions. Over the years, as more information is disclosed, the less people read it and so they may sign off on a piece of paper that says they have read and understood everything, without actually reading or understanding it. Then the complaints come.

There really is no substitute for essentially walking the shoes, not just of citizens, but also that of frontline service providers.

6. How can we strike a balance between having too much information and too little?

I think information needs to have three features. It needs to be organised and it needs to be customisable. The third feature—an important one that is often overlooked—is it needs to highlight its own relevance.

So when I read contract documents, I need to know why am I reading this. Previously, many documents are written by lawyers who don’t particularly care about organising the information as long as the details are there. We put a roadmap in. The information is organised, and can be customised if it’s digital, then people know what they are looking at, where to find information they need, and can decide whether some sections are only relevant say in the case of a disagreement. You can keep all the same information, but you allow people to select what parts of the information they want.

People don’t often know what they want, so we do need to also hold their hands in terms of what we think they may want to consider. We could gather preferences of the kinds of things people might be most concerned about, and then we can guide them to what they should look for. If we have a lot of information, we can do that better.

Information needs to be organised, customisable and highlight its own relevance.

7. Might the effort to design around or overcome sludge incur an additional cost burden on the policy or process, as a result of the additional research needed?

I actually think there’s a business case to be made for reducing sludge. Looking at government, just think about the number of people or employees handling and verifying processes—I think the costs of research and redesign would be more than compensated by savings if we no longer needed to carry out these procedures.

The same is true for companies. For example, research done on some banks in Australia shows that sludge can happen not in terms of too much communication but in terms of skewed communication. One intervention was that when a bank is doing advertising, instead of only saying what is good about a product, they say when it is good and when it is not good. In the short run, the banks lost money—they had fewer people buying their products. But in the long run, they realised that the people who did buy their products stayed with them longer, so in terms of net present value, actually the revenue of the company went up.

We know from marketing research that satisfying customers, delighting customers brings you more revenue for longer, so why not delight them?

8. When might we want to build in sludge to slow down processes, perhaps as a deterrent?

In general, overconsumption of anything is something you’d want to stop. I wrote a series of papers with a former PhD student, Amar Cheema on the notion of creating transaction costs to push people from a mindless into a mindful state of consumption.

We did an experiment where we divided a big bag of movie popcorn into six smaller packs and it turned out people did not eat as much—because after you finished each bag, you had to make an active choice. The same goes for gambling, with tokens in a bag of 100 or four bags of 25 each. Once the bag of 100 was opened, it was gone, whereas with the smaller bags, there was a pause between bags and nobody gambled everything.

We did a study in India, where there was a problem with savings because people were paid in cash, and they tend to spend it. So we gave people their cash earnings in four envelopes instead of a single bundle. There was a nice catchy slogan: something like “One a day, keep the fourth away” and because there was now a friction cost to opening each envelope, they did keep the fourth away. We also put a photograph of their children on the savings envelope, so every time they try to open it, they experienced the guilt of seeing their children’s savings being used. So this was psychological sludge, introduced to promote a more deliberate mindset.³

9. Might behavioural interventions backfire, if people feel that they are being manipulated? What are some pitfalls to bear in mind with thinking about applying nudges and sludges?

A lot of my work is consistent with Cass Sunstein’s research in this area. When you nudge me to help me accomplish my goal, and you are transparent with your nudge in this regard, it turns out it actually helps, and improves the efficacy of the nudge. But if you try and get people to do things that they don’t want to do, then we have the opposite problem right. There are many reasons why people don’t like nudges: for example, because sometimes you are nudging people to do things that they don’t believe in. Oftentimes we try and nudge people without motivating them about the importance of that behaviour. That’s a mistake.

I think the motivation engine is separate from the process and communication engine. Motivation arises or should arise before the whole nudge/sludge thing. If I’m motivated to curb consumption, then the right approaches of sludging or impedance apply. If my motivation is to go get my welfare benefit that I’m entitled to, then that should be made easy for me.

We don’t think enough about the motivation engine. The gut response of any government, or for that matter any large organisation, to a problem is: let’s give people more information. Or let’s make them more ‘aware’. If there’s a programme and nobody takes it up, we advertise more, because we think it’s because nobody knows about it. Once we know what people’s motivations are, we can adjust to encourage the right behaviours. But often we jump in to provide information or incentives too soon.

Also, and this is not a message that most practitioners like to hear, but one size does not fit all. There is no rule of thumb that says facilitating is always better than impedance, but to me I think it is most helpful to think first about motivation separately, and then afterwards consider process and communication. Let's first ask ourselves: do we think people are motivated to make a behavioural change, and if not, how can we help them do that? Is the lack of motivation due to what they believe, is it due to cultural norms, or particular experiences, or is it a lack of awareness?

So these are the kinds of things we need to start asking first: where the motivation is going to come from and how do we make sure it exists. Then we think about whether, to get the appropriate outcome, we need people to become more mindful, or less. Do I need to add friction or subtract friction? Sometimes we want to make it easy, sometimes we need to get people think a bit more. And sometimes we have a third kind of situation—because one size doesn’t fit all—where people make decisions that are just right for them. There’s a lot of hetereogeneity in decision-making.

This is not a message that most practitioners like to hear, but one size does not fit all.

10. Governments can have a bias for action, and speed to solutions, because they are accountable to the public. How can governments be mindful when applying such interventions?

I often tell my practitioner friends that there is a tendency to think in terms of implementation details as opposed to conceptual ideas. For example, people will say: should I simplify forms or should I change the framing? Those are just ways of accomplishing a simple conceptual outcome: which might be to change perception or improve consumption of information and so on.

I think we need to start thinking along the lines of whether we want to help make people more mindful, or less. Are they overburdened with information or do they have too little information? This is actually a big challenge. Sometimes the best solution is not doing anything.

The fact is that different populations might require different treatments. Think about investing: the experts know too much, the novices have access to the same information but they don’t read it so they know too little. We have the same problem but the subgroups have different challenges, so for one group we need to simplify, and for the other group convey things in a friendlier format.

The challenge is that because one person’s sludge might not be another person’s sludge; sludge can be hard to detect. We need to develop “sludge googles” (or scorecards): to help us see sludge, and to help us notice how many steps we have to take, where the longest delays, potential liabilities, emotional touchpoints are.⁴

All of us as individuals in society have run into bad customer experiences—but when we ourselves design processes, we tend to undermine our own experiences. I think we need to be more sensitive to that. Everytime you have a problem as a consumer or as a citizen, just ask yourself, might other people have the same problem too?

ABOUT THE AUTHORS

Dilip Soman is a Canada Research Chair in Behavioural Science and Economics, and serves as a Director of the Behavioural Economics in Action Research Centre at Rotman (BEAR). His research is in the area of behavioural science and its applications to consumer wellbeing, marketing and policy. He is the author of The Last Mile (University of Toronto Press) and teaches a massive open online course (MOOC) "BE101X: Behavioural Economics in Action" on EdX. He was keynote speaker at CSC’s 2019 Behavioural Economics Symposium and spoke on “How Can Behavioural Insights Solve the Last Mile Problem in Service Delivery?"

NOTES

William James, The Principles of Psychology (New York: Henry Holt and Company, 1890).
Richard H. Thaler and Cass R. Sunstein, Nudge: Improving Decisions about Health, Wealth, and Happiness (Yale University Press, 2008).
Dilip Soman and Amar Cheema, “Earmarking and Partitioning: Increasing Saving by Low-Income Households,” Journal of Marketing Research Vol. VLVIII (November 2011), S14–S22, available at http://www-2.rotman.utoronto.ca/facbios/file/earmarking-jmrPP.pdf.
Dilip Soman, Daniel Cowen, Niketana Kannan and Bing Feng, “Seeing Sludge: Towards a Dashboard to Help Organizations Recognize Impedance to End-User Decisions and Action” Toronto, Canada: Behavioural Economics in Action at Rotman (BEAR) Report series, 24 September 2019, available at http://www.rotman.utoronto.ca/bear.

Back to ETHOS Issue page