Building a Budgeting Tool That Works


In the previous article on budgeting you’ll have learned that the way we have been taught to think about budgeting isn’t necessarily the way which will make budgeting easier, or even effective for us. Oops. An overview of your expenses is not enough to change behaviour, yet this is where most apps start: they tell you how much you have spent already. However, research has shown that there is no value to just throwing information at people. It doesn’t do anything. Okay, let’s get more detailed then: a lot of apps have been coded to pick apart your expenses and divide them into categories. Or you can do this yourself. Or both (works with overrides). Now you know how much you spent on groceries, and that maybe you should spent a bit less. Is that useful? Not if you don’t have insight into your ordinary or exceptional expenditures, as both Sussman and Alter (2012) and Huebner, Fleisch and Ilic (2020) have shown. Rather than showing you what you have spent, should you be shown how much you have left? Lord, no, that’s even worse. Research by Pocheptsova Ghosh & Huang (2020) showed that the presence of a balance works as an anchor. The mind hooks itself on this number and somehow treats it as a goal to beat. If you know you have $300 left in your current account, you will spent $300, minimum. I say minimum, because Pocheptsova Ghosh & Huang found that most people overspent, because they couldn’t accurately recall how much they had already spent and how much was actually left, because people have a tendency to underestimate. Should you then split your money? Have multiple accounts? Yes and no. If you simply have more balances (of different accounts) you don’t solve the above issue. However, if you need to save $100 out of the previously mentioned $300, it’s probably best if you move it away from your spending account, into a savings account. Just don’t confuse the two accounts. It would actually be better if you could not look at that savings account at all. This is exactly the reason that Monzo decided to let you hide certain “pots” of money, so you’re not tempted to spend money that has a different purpose. Clever.



What we have so far is different pots of money – what can be spent and what needs to be saved – and the idea that some of it should be hidden. Fair. Do we need anything more? Well, unless you want to create your own feedback report and analyse your own progress you’re going to need a lot more. Despite categories not being the most useful mechanism, most people do seem to like them. Especially the category “take away coffee” or “take away lunch” seems to make people realize that these smaller expenditures do add up! In addition to these categories you will also need to have a category measure for ordinary and exceptional expenses, as discussed in last week’s blogpost. When it comes to accurate budgeting, that’s where it’s at! Now I’m sure that most apps will let you classify whether the expense was ordinary or exceptional. However, that seems like a lot of work, and, as we’ve found out from Sussman and Alter (2012), we often use extremely narrow bracketing for exceptional expenditures, which leads to overspending in that category, because we don’t account for all of the expenses that are, technically speaking, exceptional. It would be nice if the app could figure out what is going on here. And now we’ve moved into the domain of AI. It’s not surprising that AI has piped up in the fintech space (it’s literally less surprising than Santa not being real – who still has a working chimney these days?! – also sorry for ruining Santa for you). Through the application of AI you take a lot away from human bias, error and guess work and let the machine run the numbers for you. In the case of budgeting tools, the AI can figure out from your past expenses what’s going: how much do you spend, on what, how often, where, (maybe even with who?), when/at what times (hour, day, week in the month, month, seasonal trends etc.). The more information the app has, the more it can tell you. And it will tell you (assuming you’ve had notifications enabled). Picture this: you’ve had an app for years. It knows when you’re going Christmas shopping, the likelihood of you buying a gym membership in the new year which you’ll cancel again after 3 months, the additional Valentine’s day expenses, the fact that your mom’s birthday is in March and you will send a gift over (price + shipping). Plus, it obviously knows all the ordinary expenses: rent, insurance, groceries (estimated average), transport, bills etc. This app knows you, it tracks everything. But now what? We know that throwing information at people in the hope that their behaviour changes doesn’t work. Information isn’t enough. Somehow, the answer to this non-question was thought to be giving people more information, but then we ran into the information paradox where people faced with too much information turned into the proverbial ostrich and did absolutely nothing with all the information available to them. Just as the mighty ostrich lays its head on the sand, the mighty homo sapien lays its head on a pillow, hoping that next day, somehow, their financial situation will be better.


We have now gone into the nitty gritty of building a budget app, but have we maybe gone so deep into the trees that we can no longer see the forest? Is this not getting a bit too complex? What is the point of budgeting? In its very essence, it’s knowing what goes in, and what goes out. But maybe even more basic: it’s knowing whether you can afford certain things, and still stay on track. Realistically, you just want to be able to ask your budgeting tool whether you can afford go out to dinner tonight. There are apps out there who literally let you do this. One app that comes to mind is Cleo (not sponsored, wish it was). You can literally interact with Cleo, who is essentially an AI bot with insight into your finances in a way that only AI could manage it. The interactions aren’t stunted either: the feedback isn’t given in a way which makes you think you’re completely alone either. As the home site reveals, Cleo can praise or roast you, and does so through GIFs as well. Yes, this is a money app targeting the younger generations, but who says it can’t work for someone who’s near retirement age? The app, like many others, relies on a lot of automation as well: to save it has auto-save options, rounding spending options, as well as a “swear jar” (you get fined for spending on things you wanted to avoid spending on). It has goals you can set, updates on that goal and the aforementioned roasts if you’re not doing too well. I’m not saying Cleo is the be-all-end-all of finance apps, it’s just an example. And they don’t get it right either. They still use a categorical spending system (clothes, groceries etc,) as well as having separate accounts showing how much money you have saved in different jars (although these can be hidden, so there is advancement!). Additionally, I have to mention Cleo also has credit card options and “cash spotting” but I haven’t researched those options yet, so my opinion on those remains non-existent.


 

Although I fully believe in actively knowing what goes in and out, and going from there, I think life has become a little too complex for most people to actually manage this properly themselves. What we’ve seen so far is that the way we think about budgeting is not the way we actually budget, or should be budgeting, which is just unfortunate all-round. Given the counterintuitivity of the thing, I think we might as well outsource it to apps that track us much better than we could ever track ourselves. Ones that let us set our goals, give us feedback in a way which isn’t just information overload and hold us accountable for our actions, preferably with a GIF. These apps already exist and I’m keen to see what they come up with next. Let me know if you use such an app and what your experiences are with it. I’d also love to hear from people who have completely different ways of budgeting, or who are maybe building their own apps or platforms. Please do reach out!




Behavioural Science