Through market-leading categorization and core enrichment capabilities, Bud’s data intelligence solutions empower financial institutions around the globe with the data-driven insights they need to succeed.
Hosted by Product Manager Thomas Purton, the webinar covered customer-centric money management features and engagement through enrichment to rich personalization capabilities and truly targeted messaging. It's jam-packed with value you won’t want to miss.
[Sam: 00:03.5]
Great. Looks like we're doing fairly well on attendance, so we will kick things off. Just wanted to begin by saying good morning or good afternoon, depending on from where in the world you're calling in from and thanking you for joining our webinar today.
[Sam: 00:20.0]
My name's Sam, I'm VP of Sales here at Bud. I have the easy job today. I'm going to be moderating and reading out your questions. I'll hand you over to my colleague Thomas in just a moment. We really do encourage questions throughout this. Please post them in the Questions tab as part of the portal and I'll be reading those out as and when.
[Sam: 00:40.9]
So, Thomas, over to you.
[Thomas: 00:43.9]
Thanks, Sam. Hey, everyone. So, as Sam mentioned, I'm going to be the one leading this webinar today and hopefully answering all of your questions as well. My name is Thomas. I'm one of the product managers here at Bud and have been heading up our Engage Proposition for about the last two years now.
[Thomas: 01:03.9]
So, as I mentioned, thanks for signing up and coming along. This webinar is for those of you who maybe already have your own customer transaction data, or perhaps you're intending to aggregate customer data. So, depending on the market that you're in, this may be via open banking, it may be via PDF ingestion or via screen scraping.
[Thomas: 01:25.1]
And hopefully you're wondering, how can you turn this data into personalized insights in order to provide a better experience for you and your customers? So what are we going to cover today by the end of this webinar? My goal is to provide you with a number of use cases and examples of how you can use transactional data in order to enhance your customer experience.
[Thomas: 01:47.2]
We're going to be talking about things such as the importance of transaction data, how you can start building personalized customer journeys once you've enriched your transactions, and then almost most importantly, kind of personalization and how you can use this in marketing to help increase, cross, sell and upsell to your existing customers.
[Thomas: 02:10.6]
So, before we start talking about personalized journeys, I just want to take some time at the start to really talk about data enrichment. Now, data enrichment is super important at Bud. It's the foundation for all of our more complex features.
[Thomas: 02:25.8]
And so everything that we build is kind of dependent and reliant on these core enrichments that we place on each transaction. So just to give you an example of what I mean by transaction enrichment, if we take an example transaction, you can see that we've got a description, we've got an amount and we've got a credit debit. Indicator.
[Thomas: 02:47.4]
Now, depending on the source of the data that you have, you may also have things such as MCC codes or bank transaction codes. Now, what we do at Bud is we take all of these variables and we run them through our enrichment models. The transaction description is particularly important because it appears in almost all transactions and therefore there is typically a lot of information contained within it.
[Thomas: 03:11.1]
So for descriptions, what we have is a specific entities model. Now, what our entities model allows us to do is essentially tokenize this description. So, as we can see here, this raw transaction description has been tokenized into all of these different tokens.
[Thomas: 03:29.2]
Each token is then assigned a specific type. So this may be to do with the processor, it may be to do with the merchant, it may be to do with the location, it may be some sort of identifier, whether that's a date, a time, a store id.
[Thomas: 03:44.7]
And then once we have all of this information, what we're able to do is then start adding enrichments to our transactions. So, on an individual transaction level, we focus on four core enrichments. First of all, we've got category, so as you can see here, so what we do from a category perspective is using Bud's own taxonomy, we assign both an L1, so a level one and an L2 level two category to that transaction.
[Thomas: 04:13.8]
So, depending on where you are, whether you're based in North America, you're based in the uk, Europe, in Oceana, you may have slightly different taxonomies because we localize those to the local market. But, yeah, for each transaction we will provide an L1 and an L2 category.
[Thomas: 04:33.4]
Just before I move on to the other enrichment types. Sam, I don't know if we have any early questions about transaction enrichment, about categories.
[Sam: 04:42.3]
There's been a couple that have come in related to a question around accuracy, and I think there's another one here that is related to, I guess, our differentiators in somewhat of a crowded market.
[Thomas: 04:55.8]
Sure. So the accuracy obviously depends on the market. So in the UK is where we see the highest accuracy. So we're looking at 98 of all transactions categorized in UK at an L1 level. At an L2, because it's that more detailed, it's slightly lower, around the 97% mark.
[Thomas: 05:17.3]
And then in the US, we're looking at 97 actually for L1s and 95 for L2s. Now, if you're wondering what the kind of split of the number of catches categories across L1s and L2s are, we typically have around 20 to 25 categories.
[Thomas: 05:34.1]
At an L1 level, and then below that we have about 120 categories at an L2. So depending on the use case and depending on how detailed you want to get, you can obviously show for a PFM use case, it's more about the L1s. And then if you're thinking about lending or affordability, you may want to get more focused on an L2 in terms of how we differentiate ourselves in the market.
[Thomas: 05:57.9]
It's a great question. As the market matures more, what we've started to see is more and more companies start to focus on transaction enrichment, something that Bud has been doing from the start. Where we think we probably differentiate ourselves is when speaking with clients and kind of their experience with our competitors.
[Thomas: 06:19.4]
What we see quite often is what clients typically have is a rules based approach to categorization. So if they see that a transaction of a specific merchant, then that will dictate what the category may be. This works for a lot of merchants and it does have its benefits.
[Thomas: 06:38.7]
However, it also has its limitations. As merchants start selling more and more products, there's obviously different ways that you could categorize a transaction, right? So for instance, if we look at Bud, if we see transactions coming from gas stations, so if we see that it's a small transaction under $10 and it's from a gas station, we may categorize that as a convenience store.
[Thomas: 07:05.4]
And then if we see that it's a substantial spend, so 40$50, then we know, okay, this is actually now more likely to be gas or petrol, you know. So this is just one example of how using this rules based approach would just always set the transaction as gas.
[Thomas: 07:24.7]
Whereas with this more nuanced approach that we use at Bud, we can kind of get more granular and really understand how transaction amounts, how processes all affect what the category could be for a transaction.
[Thomas: 07:41.6]
Great. So as we start to look at more of our enrichments. So then next up, in a nice segue, we also have merchants. So what we do from a merchant perspective is we take any merchant tokens that we have. In this example, it may be the star and the buck.
[Thomas: 07:59.2]
And then what we're able to do is cross reference this against Bud's proprietary merchant database, which then allows us to kind of match that to a specific merchant and provide more information about that merchant.
[Thomas: 08:14.9]
So that includes things like a clean merchant name, a merchant logo, and a website. And so what this allows you to very easily do is then start incorporating these pieces into your banking app. Now, you may have seen online that sometimes we talk about merchants, sometimes we talk about brands just to really illustrate the difference between the two.
[Thomas: 08:39.7]
So any place where a transaction could take place could be a merchant. So when we're thinking about our long tail of merchants, so where there's typically lower coverage, so this is those mom and pop independent stores.
[Thomas: 08:55.5]
It could be a corner store that would still be a merchant and would still have an entry in our database. Brands are where we see kind of more than three locations for a single merchant. So if you're thinking about someone like a McDonald's, a Subway, a Wendy's, a Shell, the places which have multiple, multiple locations, those are really what we think of as brands.
[Sam: 09:21.5]
Thomas, just a quick question regarding regionality and where we can identify merchants.
[Thomas: 09:29.2]
Sure. So in the UK, we have around about 20,000 merchants within our merchant database. In the US we're looking at around 40 to 50,000 brands across those. I will also touch on it a bit later on.
[Thomas: 09:46.4]
But across locations for those brands, we're looking around the 15, 16 million mark. So, you know, 40 to 50,000 brands and then about 16 million locations for those brands. The third type of enrichment that we do at Bud is around regularity.
[Thomas: 10:05.5]
So here what we do is we take this transaction description, we clean it. So we take out any specific identifiers that may be to do with a specific date or a specific time. We then compared this transaction to all other transactions for a given customer.
[Thomas: 10:23.5]
And then what this allows us to do is start building up groups of transactions. And so then once we've got these groups of transactions, we then start looking at regularity. So this may be something that happens on a daily basis, a weekly basis, monthly, annually.
[Thomas: 10:38.9]
But we also take into consideration things like public holidays, weekends, leap years to kind of understand where that regularity happens. Finally, as I previously mentioned, we have locations. So when we're able to find a match for a merchant, what we're also able to do is then start taking any, any other identifiers which we may have about a transaction's location.
[Thomas: 11:06.0]
This could be a town, could be a city, could be a state, could be a. It could be a postal code, could be a zip code, or it could also be something like a store id. And then what we do is we take this information, we take the information that we have about the merchant, and again, we try to match it against Our list of 15, 16 million locations to then be able to return an exact street address and coordinates for that transaction.
[Sam: 11:38.8]
Thomas, just quickly, another one, I guess, related to a crowded market. I've heard a number of vendors use the same location database. How can you differentiate?
[Thomas: 11:52.1]
Sure. So in terms of locations, particularly in North America and Europe, there's typically two vendors that we see most people using. That's either someone like a Foursquare or a Safe Graph. Now, whilst the underlying database that transaction enrichment companies may be matching against, where we see superiority and where we think we differentiate in the market, is around the underlying tokenization of transactions and our ability to match to merchants.
[Thomas: 12:24.9]
Without these, it kind of limits the amount of locations that you can actually find forward transactions. And therefore, if even though they're trying to match against the same database, they're not going to be able to do as good of a job. You may also be wondering how we handle online transactions.
[Thomas: 12:45.1]
So, thankfully, the days of Blockbuster are over. You no longer need to go to a store. So if you have things like a Netflix transaction or maybe a Spotify transaction, instead of just saying, oh, we can't find a location for this, what we'll also be able to do is add an extra layer of enrichment to that transaction.
[Thomas: 13:06.5]
So the fact that we know that the merchant is someone like a Netflix or someone like Spotify, we can identify that this is an online transaction and therefore, instead of just saying no location, we can say, hey, these are the transactions which this customer is making online.
[Thomas: 13:26.0]
Just before I move over, I was just wondering if there are any other questions that we've had about enrichment. If not, I can move on.
[Sam: 13:32.7]
No, that's all looking good, Thomas.
[Thomas: 13:35.0]
Nice. As you can see here, by running these enrichments over transactions, what we can do is turn a finance app from looking something like this to something like this. Now, just by doing this, you're already adding a tonne of value for your customers.
[Thomas: 13:52.8]
It cannot be understated that the number one and number two reasons for logging into finance apps is to check your account balance and to view recent transactions. So, first and foremost, we want to make this as easy as possible for customers to understand. By adding these types of enrichments to a transaction, you're making it easier for your customers to review their transactions and importantly, recognize them.
[Thomas: 14:17.3]
One of the top reasons for customer disputes is when customers see transactions that they don't recognize. By adding these enrichments and context to your transactions, you're therefore making it less likely for your customers not to recognize a transaction and therefore less likely to raise a dispute.
[Thomas: 14:33.6]
Not only is this a better experience for your customer, but it also saves you operational costs when dealing with customer disputes. And so now that we have these enrichments on an individual transaction level, what we can also start to do is build up what we call customer level enrichments.
[Thomas: 14:52.1]
These include things like identifying a customer's characteristics. Are they a parent, do they own a home, what's their income, what are their spending habits, do they frequently travel? Or maybe they live paycheck to paycheck. And these in turn help us to build out customer journeys.
[Thomas: 15:12.9]
Now, for those of you who may be unfamiliar with budget, we're an API based solution, which means we're providing all of the information that I'm about to show you in our demo. But ultimately, clients have the freedom to design the interactions in their own brand and style.
[Thomas: 15:29.3]
I just want to make that clear from the outset
[Thomas: 15:35.1]
in terms of what we can do to help clients when they're thinking about what the visual style should look like for their financial app, whatever it may be. We do act on a consultative basis, so with a number of clients, we give them advice on how they can provide a personalized experience within their application.
[Thomas: 15:57.5]
We feel like we can do this because we provide, we've run this and done this with a number of projects alongside financial institutions and fintechs over the years, as well as for those of you who may not be aware, but did actually start out as our own personal finance management app here in the UK.
[Thomas: 16:13.5]
So we know a thing or two about how to deliver insights to customers and kind of the results that you can expect. Now, one of the first things that we always talk to our clients about is the importance of user journeys, often with PFM solutions.
[Thomas: 16:29.4]
So personal finance management, for those of you unfamiliar, what we see is a collection of features buried within a sea of menus. This can be overwhelming and confusing for customers and it also leaves a client wondering why they have low adoption rates.
[Thomas: 16:44.9]
Why aren't customers using these features that I'm putting in the app? What we always suggest is to think about the customer journey that you expect someone to go on and shape the experience around this. Our framework for creating these journeys is a three step process.
[Thomas: 17:00.6]
First of all, we have enrichment, which is everything that I've just spoken about and the foundation for everything else that we're about to do. Next, we have education. So this is how you display a customer's data back to them and educate them on their financial situation and then finally you have action.
[Thomas: 17:19.4]
So now that your customer has that context, what is the next step that they should be taking? And this is something that we try to do again and again at Bud. So whilst I'm walking you through a few examples of journeys that you can build with Bud, I want you to have this three step process in the back of your mind before I swap to demo.
[Thomas: 17:38.1]
I just want to pause for a second. Are there any other questions?
[Sam: 17:42.3]
We've got a few questions in Thomas. So question one. Do you use the update category to understand and update the main categorization when enough people have suggested the same new category?
[Thomas: 17:57.4]
Yeah. So how recategorization works at Bud is we have a set endpoint. So as I mentioned, our accuracy is around 97, 96% depending on whether you're using L1s or L2s. So inevitably there will be times when a customer may want to recategorize a transaction.
[Thomas: 18:16.5]
Now what we can do and what that endpoint allows us to do is first of all recategorize what a transaction is being listed as. But then secondly, you can also recategorize similar transactions at the same time. So if you're wanting to recategorize something like your Netflix subscription as fitness, let's say you don't need to then go through and update every single one of your Netflix subscriptions.
[Thomas: 18:42.7]
What the Bud platform will do is understand what are the similar transactions to this transaction that you want to recategorize and do them all at the same time. And this will also work going forwards. So that kind of learns how the customer wants to have their transactions categorized.
[Thomas: 18:58.5]
And going forward for that customer, that rule will be put in place and it will update and categorize those Netflix subscriptions as gyms and fitness. Now all of the categorizations that we get submitted at Bud get sent through and stored within what we call our data lakehouse.
[Thomas: 19:17.7]
Now these then go through a process of being manually reviewed by someone within our data team and then they can ultimately decide whether this is a valid recategorization or a recategorization that we want to reject. When working with machine learning models and AI, you always need to have some sort of supervision for it.
[Thomas: 19:39.8]
Because for instance, let's say we have customers recategorizing transactions that we just fundamentally don't agree with. So this may be someone such as revenue categorizing gambling transactions as entertainment. We don't want the model to then start learning that and picking that up for the rest of our customers.
[Thomas: 19:57.5]
So all of the recategorization that you see first of all affects the individual customers transactions and then it gets observed and checked before we spin out to the rest of our customers.
[Sam: 20:12.7]
Great. And another question here related to categorization: How does your categorization of business transactions compare to your consumer categorization?
[Thomas: 20:22.3]
Yeah, so we have some examples and some experience of doing business categorization. It's something that we have a few clients using it for. Typically, the accuracy of the business categorization depends on the type of business.
[Thomas: 20:40.1]
To give you an example. So what category a transaction may go into very much depends on the context of the business itself. So, for example, someone buying flowers from a personal perspective may categorize a transaction as personal care.
[Thomas: 20:59.1]
Someone within a hotel who's buying some bouquets of flowers may categorize it as operational costs. And then a florist may then actually categorize that as I'm just taking in stock. So what we see with the accuracy of business transactions, it very much depends on the business itself.
[Thomas: 21:19.4]
We do have customers who are using Bud's categorization models for their business. However, we don't currently offer individual taxonomies for individual businesses.
[Sam: 21:33.6]
Great. And I'm not sure if you want to answer this one or go through the demo, but we've got one here that says, what help does Bud provide in actually creating these personalized user journeys that you've referenced here?
[Thomas: 21:45.7]
Yeah, I think I'll address that as we go through the demo. It's something that we act as on a consultative basis. So as you'll see in the demo, and what I've actually spent a lot of my morning doing as well is talking with clients and explaining to them how they can build up these user journeys.
[Thomas: 22:04.9]
Cool. So as we just shift to the demo here, one of the things that I want to start talking about and one of the most common features that we see on the market at the moment is around our left to spend feature. So you may know this about some, you may have heard of this as something auto referenced as balance after bills.
[Thomas: 22:27.4]
What this essentially allows you to do is provide two balances to a customer. So you can see, first and foremost, we're providing an available balance for customers. Customers. But then what you're also providing is a left to spend balance. Now, this is taking into consideration things such as any subscriptions which they may have coming up, any regular payments.
[Thomas: 22:49.2]
So if we know that their mortgage is coming out later in the month, we obviously don't want to be telling customers you've got a thousand dollars left if they've then got $800 that they need to spend on a mortgage. So the left to spend is kind of a way to tell a customer, here's your discretionary spend that you can do.
[Thomas: 23:08.5]
Now, as we click into this page. One of the things which I want to highlight is around the page hierarchy. So when we're designing pages and when we're advising our clients on this, what we always want to be thinking about are what are the jobs to be done and why are customers coming to this page.
[Thomas: 23:25.6]
In this example, from our research, the things which customers are most looking to understand is kind of how, how much money they have left first and foremost, which is why you see that at the top, and then how their balance is changing over time. So we can see, here's how much their bills, how much they spent on bills so far this month, and then here are any bills which are upcoming.
[Thomas: 23:49.3]
Now, one of the great things about Left to Spend is it's very easy to personalize to your customer using Bud's platform. So as well as our core enrichment, all of the APIs and features which I'm talking to you about are all modular, which means as a client, you can kind of pick and choose the information that you receive based on the use case that you're building towards.
[Thomas: 24:10.9]
So, for example, with Left to Spend, as well as utilizing things like a customer's balance, you can also utilize things such as regular transactions to show here are the bills which you spent so far this month. You can also look at future transactions.
[Thomas: 24:26.5]
So what are the bills which they have upcoming? And then if you want to take it one step further, something else you can do to personalize this feature to your customer is use information about their income here. So instead of just setting a cycle date as the end of the month and saying, we're tracking your left to spend month to month.
[Thomas: 24:47.5]
Typically, from speaking with customers, what we actually find is they're tracking it from payday to payday. Now, the majority of customers will be paid around the end of the month, but it could be the middle of the month. For some customers, it could be at the start of the month. It could be that they get paid multiple times a month.
[Thomas: 25:05.1]
What this feature allows us to do is use information that we get from Bud's income finder endpoint to understand, okay, when is the next time that this customer is going to be paid and then automatically setting that cycle date to this. This is just one of the examples of how, using Bud's platform, you can kind of personalize these features to your customers at scale.
[Thomas: 25:28.7]
Now, something else which I also want to talk about is one of the exciting features that we're building at the moment within Bud. So as we look here to this visualization, what we're Doing here is overlaying both the regular transactions that we see from a customer, but also the irregular transactions.
[Thomas: 25:45.1]
So when are they spending on groceries, when are they catching public transport, when are they filling up on gas? And what we can start to do is kind of understand, okay, what are the days which they're spending more money? It may be that they're a hybrid worker and they go into the office on Wednesday and therefore they're more likely to eat out.
[Thomas: 26:06.8]
Or it could be that they're more likely to go to a bar or a pub on Friday night. And therefore Fridays are typically a, a heavy spending day. And so what we're doing at Bud at the moment is we're understanding how we can use this data to not only show this for previous month and the current month, but also how we combine everything we know about the customer to kind of predict this going forward for the future months.
[Thomas: 26:31.2]
So how much money do we think a customer is going to spend on any given day for the next month? And when we're doing this, what we can then start to do is, is now that we've gone through the first two points, so now that we've enriched their transactions and now that we've educated the customer about their position, what's the action that we want them to take?
[Thomas: 26:52.0]
Now, in this example here, we can see that, okay, this customer has a positive balance and therefore what we're wanting them to do is move this money from a low interest account, so something like a checking account or a current account, into more of a high interest account, such as a savings account.
[Thomas: 27:10.8]
Now, not only are we personalizing this action to the customer, what we're also doing is personalizing the call to action. So what we can see here is we're saying, hey, we know that, you know, a savings account may be paying 5% interest, 4.5%.
[Thomas: 27:26.8]
And so by moving this money into the interest account, or, sorry, moving this money into the savings account, you will be earning an extra X amount per year in interest. So it's just about contextualizing the benefit to the customer.
[Sam: 27:45.2]
Thomas, we've got a couple of questions here. So one is how do these features improve user engagement?
[Thomas: 27:53.3]
Sure. So one of the key metrics that we start tracking, so at the start of every project that we do with clients, what we typically do is sit down with our client service team and understand what are the outcomes that they want to get from their customers.
[Thomas: 28:11.3]
So what are they looking to do? Are they looking to increase deposits? Are they looking to increase engagement? Do they want to increase daily active users, weekly active users customer retention from the clients that we've spoken to. By introducing these kind of features, they're increasing weekly active users, typically by around 20%.
[Thomas: 28:33.9]
One of our clients in the UK recently sent out a survey to all of their customers and said, hey, the insights that we're getting from Bud, how would you rate them? And 51% of customers rated these insights five stars. So what we're seeing from customers is by personalizing this to them, it's incredibly relevant and therefore they're a lot more likely to engage and keep on coming back to the app.
[Sam: 29:01.1]
Great. We have another one here around development. So does Bud support in the development process of an app or your clients expected to build this completely themselves?
[Thomas: 29:12.4]
Yeah, so we definitely support in the development process of the app. So what we have, as well as our API docs, we also have an extensive list of guides which contain recommended best practices, examples of how you can implement these kinds of situations.
[Thomas: 29:33.0]
We also have a technical delivery team. So for each client that signs up with Bud, you will be assigned someone from this technical delivery team who's always there to answer questions and can kind of guide you through the process. So if you've got any specific questions, they can answer that.
[Thomas: 29:49.3]
If you tell them about the use case which you're wanting to build, they can guide you and tell you, okay, here's how I would build it if I was building on Bud's platform.
[Sam: 30:01.7]
Perfect. That's all for now, Thomas.
[Thomas: 30:05.0]
Right. As we can see here, obviously in this example the customer is doing great. They've got a positive end of cycle balance, but as we know, that's not always the case. So in an example where a customer may not be doing so well, what we can do is kind of again, tailor and personalize not only the call to action, but also the journey to that customer.
[Thomas: 30:28.8]
So in this example, when we see that the customer is going to be short of money, what we can do is one, recommend that they set up a budget. But secondly, what we can do here is say, hey, using our categories totals endpoint, we can understand that the discretionary categories where you're spending the majority of your money seems to be on eating out.
[Thomas: 30:49.8]
So if you're going to not have enough money left over, the budget that we recommend setting up is around eating out. So as a customer, as I click into this, firstly what I'm able to do is have this eating out pre selected. So what we're doing here is trying to personalize this journey at every step of the way for the customer.
[Thomas: 31:09.9]
And by doing this, what we're doing is just taking a little bit of friction out from each step where a customer may otherwise drop off. So again, as a customer, I don't need to think about, okay, where can I cut back? Or how much am I spending on what? I don't need to be tracking this stuff in a spreadsheet.
[Thomas: 31:27.3]
Instead, Bud can do all of this heavy lifting for you as I click through to this. What we can also do is do things like suggest, okay, here are amounts that you'd want to cut back on. So 5%, 10%. We can also display things and contextualize it, such as, currently you're spending 30% of your income on this.
[Thomas: 31:48.4]
Maybe you want to cut it back to 25% of your income. Or alternatively, if you see that the customer, for instance, is spending a lot of money on something like an Uber eats, what you can do is say, hey, last month you did six transactions with UberEats.
[Thomas: 32:04.9]
Instead, why don't we try to cut that down to 5? And by doing this, by knowing the amount but also the number of transactions, what you can kind of do is equate an individual action from the customer into what will the dollar amount be changed for that customer.
[Thomas: 32:23.6]
And so by setting up these budgets, of course, what this then does is gives you many more opportunities for touch points with that customer. So as they kind of using the app and going through the month, if you see that they are going off track or you see that they are doing well, this just gives you more examples and kind of positions where you can kind of re-engage with that customer, get them back into the app and say, hey, maybe we can see that you're not covering your budget this month.
[Thomas: 32:52.3]
Maybe you need to slow down. Or if they are doing well, you can give them signs of encouragement, or if they're doing exceptionally well, you can say, hey, why didn't you start setting up a savings goal for all the money that you're saving?
[Thomas: 33:08.0]
Any other questions at this point, Sam?
[Sam: 33:11.3]
Yeah, a couple. Thomas. Probably quite a timely position for this one. So all of this is reliant on actually getting the data in. How do you maximize the number of people who would look to connect their accounts?
[Sam: 33:27.3]
And there was a follow up question to that, and how would you influence them to connect more than one or other accounts beyond their main current accounts?
[Thomas: 33:37.1]
Yeah, they're both great, great questions. So in an ideal world, if you're a financial institution, it may be that you already have all your customer data, but For a lot of people in the market, it will be that you need your customer to kind of proactively share that information with you, whether that be via open banking, screen scraping, whatever that may be.
[Thomas: 34:00.3]
Now, when thinking about incentivizing customers, I think what we see a lot of our clients do is immediately think about what's everything that I could do with the data. But what we always like to think about is what is that value proposition for the customer, where is this value exchange?
[Thomas: 34:18.9]
And how can you provide as much use to your customer as possible? So instead of just thinking about Engage or thinking about personalization around how can we upsell products to our customers? What you can actually do and how you can see better results is if you think about, okay, how can we help customers to improve their financial standings?
[Thomas: 34:39.6]
When you target it like this, you kind of start building the proposal around the customer and saying, as a customer, what am I getting out of connecting my account? Because if the customer isn't getting anything, they're not likely to actually provide that information to you.
[Thomas: 34:55.2]
Depending on your use case, there's obviously other incentives that you can offer. We're speaking with clients from the reward space. So something which is popular to do there is to say, hey, if you connect your account, we can then start giving reward points for every time you shop at certain stores.
[Thomas: 35:14.3]
But what we can also do is take your historical transaction data and say, hey, we can see over the last six months you've maybe spent $200 at one of our stores. And so we can then proactively give you those points. So here's an incentive just for connecting your account equally if you're looking in the lending space or affordability.
[Thomas: 35:36.2]
We've seen clients offer situations where it's like if you connect your account, that obviously gives us more confidence and reduces our risk, and therefore we can give you a slightly better interest rate on that loan or whatever it may be if we've got this extra information from you.
[Sam: 35:54.7]
Great. Thanks, Thomas. No more questions from the audience at the moment, but maybe a question from our side that people can have a think about over the next 20 minutes and if they feel comfortable, maybe share some thoughts in the chat.
[Sam: 36:09.9]
I think the first one coming to mind is, is this something what we've shown you so far that you've had experience of maybe deploying yourself? And if so, do you have any experience you can share around that? And secondly, would be out of what we've shown you so far, what could you envisage providing real value if you were to utilize something like this.
[Sam: 36:30.3]
But also importantly, out of what we've shown you so far, which could you see not working and what are the reasons why? So for anyone in the audience that feels comfortable sharing, we can come to those at the end.
[Thomas: 36:45.7]
Right? And so as we just kind of move on from that use case, so if we think of that as the end of the journey. Something else I just want to touch on is an example of what we've seen one of our clients do here in the uk. So in this example they're a subscription management app and so what they're looking to do is help customers get rid of and cancel dormant subscriptions.
[Thomas: 37:12.0]
So I'm sure all of you, or some of you may be in the same position as me that during COVID everything seemed to go subscription based. Suddenly you could order any type of subscription to your door, whether that's a cheese subscription, a new pair of shoes each month, whatever it may be.
[Thomas: 37:30.4]
What this has now left is a situation where many people have many subscriptions which typically they're not using or underutilizing. And so what this client has done in particular is they've used BUD to allow customers to connect their accounts.
[Thomas: 37:46.2]
And then using our subscription finder, what they can do is provide to customers a very clean list and say, okay, here are all of the subscriptions which are currently paying each month. And when they're coming out from here, what they saw was on average each customer had more than 10 subscriptions.
[Thomas: 38:05.1]
And then what they're allowing the customer to do without leaving the app, they're allowing their customers to cancel that subscription with them. And just by doing that, they're estimating that households are saving between 500 to 1000 pound per year from using their app.
[Thomas: 38:21.3]
So to put that in a dollar perspective, it's around 600 to $1,200 per year. Again, as you can see in this example, the client has followed our framework of one, enriching the customer transaction, two, educating the customer about their current subscriptions, their current financial position and then three, getting the customer to take action.
[Thomas: 38:45.8]
As we just click back now, one of the other things I just wanted to talk about is around kind of some of the out of the box insights which we can also provide to customers. So typically within our clients apps there's always a section which is talking about insights, or you may think of it as kind of a next best action for your customers.
[Thomas: 39:08.6]
So here there's a range of things which you can get customers to do and you can kind of tailor These to your use case. So in this first example here, we have a feature which you've previously seen called balance over time. So, so within the Left to spend demo, what you'd have seen is someone's balance and a nice little graph showing how that's changing over time.
[Thomas: 39:32.0]
But as well as doing that, what we also provide is information around the minimum account balance, the maximum account balance, and the number of days in overdraft for that customer for that period. And so what you can do with some of these is, for instance, if you see that they're spending a lot of time time in their overdraft, or you see through categorization that they're being charged money from their overdraft, you may want to suggest that they have a conversation with you about overdraft limits, or it may be that you want to speak with them about taking out a loan with you.
[Thomas: 40:05.1]
Equally, if you see that a customer's minimum balance isn't close to zero, and they do have money that's just sitting in their checking account month after month, what you can do is start how can you get customers to start putting that to better use?
[Thomas: 40:21.3]
So in this example, if that customer's got $2,000 sitting in their checking account and you know about it, well, now you can start saying, hey, why didn't you start moving this money into a higher interest savings account? Other examples of kind of insights that we can do out the box.
[Thomas: 40:38.4]
So here's an example of something around cannot cover bills. So maybe you don't want to go through the hassle of creating a left to spend feature. Instead, what you can have is this kind of insight which will notify your customers anytime that we see that the remaining account balance in an account is less than the total forecasted payments and forecasted income for that account.
[Thomas: 41:04.7]
So if we see that, hey, this customer is going to be short this month, we can proactively let them know instead of waiting for them to miss one of these payments. We can also do things around spending. So of course, with inflation, everything just seems to keep getting more and more expensive.
[Thomas: 41:24.9]
What we can do here is start kind of calling out these bill rises as we see them happening. So with regular transactions, if we see that a subscription has increased, we can make sure that's being surfaced and known to the customer. And finally, and I think one of the most important points is around kind of product personalization.
[Thomas: 41:46.6]
So here's a great low touch example of how you can kind of use all of the information that you know about a customer to start personalizing product recommendations to them. So if we take this as an example, we may know from customer characteristics that this customer is a renter.
[Thomas: 42:05.1]
We can see how much money they're paying in rent each month. We can see that coming out. We may also know that they're saving for a house. It may be that they've set up a savings goal and called it first time buyer home deposit, a specific address, or it may be that we're just making that assumption based on the fact that we can see that their savings account is rising from here.
[Thomas: 42:27.5]
What we know is there is a good chance that the customer is in the market for a mortgage. And so now what we want to do is show an offer to them. But again, if we think about personalization at every step, what we can do is start tailoring this and the messaging to them.
[Thomas: 42:43.1]
So in this example, what we've seen is, first of all, we know that, that we know what their income is, so we can say, hey, we reckon that we would offer you around a mortgage of X amount, so therefore you could start looking for a home around this value.
[Thomas: 43:00.0]
What we can also do is give them an example of kind of roughly how much this would be that they would need to pay back each month. And then what we can do is kind of contextualize this for the customer to say, hey, we can see that this is actually slightly less or the same as what you're paying in rent, and therefore this becomes so much more attractive to that customer.
[Sam: 43:23.9]
Thomas, we've got a question here. I'm actually surprised it's taken 44 minutes of a webinar for this term to be used, but the question is, could this not now all be done with Gen AI?
[Thomas: 43:39.0]
It's a great question and very timely. It's something we've certainly thought about it. Bud and Genai is something that we're definitely looking at how we can utilize within Bud. There's a number of blogs which have written about this recently, so I know we're not alone in our findings here.
[Thomas: 44:01.4]
What we've seen with Genai is it's great at kind of finding information and generating information from the Internet, but what it's not necessarily great at doing, or what it's not great at doing at a cheap scale, is looking through thousands upon thousands of customer transactions.
[Thomas: 44:21.4]
Now, one of the great things about gen AI or a ChatGPT is you can ask it almost any question and it will be able to provide you with an answer. Now, that's great, but it means that the amount of processing power that it needs to answer any one question is also very expensive.
[Thomas: 44:39.2]
All of the Bud models which we've been working on are hyper personalized and hyper fixed to customer transactions. So that is the only thing that they do. And by doing this we can reduce that processing cost because that's all that they're doing by something like 50 to 100 times.
[Thomas: 44:58.1]
So yes, you could probably replicate some of these things with Gen AI and you could probably come up with, you could get transaction categorization which is probably 50 to 70% accurate, but really optimizing it not only from a performance perspective but also from a cost perspective, we see that it's not really a comparison.
[Sam: 45:23.9]
Great. Thanks, Thomas.
[Thomas: 45:25.9]
Right, and so one of the kind of final things I just want to touch on before we just move over to the general Q and A is around kind of the ROI around personalization, so particularly for personalized product offers.
[Thomas: 45:41.5]
So at the back end of last year what we did is we ran some research through an independent third party organization and what they found was that customers are two times more likely to engage with product recommendations like this when it's personalized and contextualized to them.
[Thomas: 46:00.1]
More interestingly, if that customer was under the age of 44, it actually increases to three times more likely to put that in perspective of the real world. So one of our clients here in the uk, they're an independent credit broker and what they found from their own research is customers who are shown Bud insight, not necessarily trying to sell them a product, but any of the Bud insights are then 20% more likely to take out a loan with them.
[Thomas: 46:29.2]
So again, if you're just thinking about, okay, there's a cost to this, what ROI can I expect? Just think about, okay, what could you do if you were three times in your upsell rate or you're selling 20% more loans? That's the kind of ROI that you can expect.
[Sam: 46:48.3]
Great. We've got another question here, just Thomas, what organizations do you think would benefit most from adopting this personalized approach to customer engagement? So I guess we're maybe talking about kind of industries in that case.
[Thomas: 47:02.9]
So first and foremost I think financial institutions. So anyone who's already got this data available to them is obviously at a massive advantage. Beyond this, I think anyone who really wants to deeply understand what customers are doing with their money.
[Thomas: 47:24.3]
So that may be someone who is doing some form of income verification. So that could be lenders, it could be brokers, it could be a mortgage broker, for instance. We've also seen examples from the reward space.
[Thomas: 47:40.2]
So anyone who's looking to understand kind of customer loyalty, instead of just creating something like a loyalty card where a customer can scan and get points by going down the open banking approach or going down the transaction enrichment approach, not only can you understand how much money a customer is spending with you, you can understand if your customer is spending money with your competitors.
[Thomas: 48:04.1]
And what does your share of wallet look like?
[Sam: 48:10.3]
Great, we've got another one here. Would you be concerned by customers using Genai LLMs given its tendency to be overconfident on answers and the likelihood of giving potentially different answers given the same data input?
[Thomas: 48:24.4]
It's definitely a concern that we see often in the industry. One of the terms which we always like to think about when we think about interacting with Genai is around explainability. So the work that Bud is currently doing at the moment, looking at how we can incorporate Genai within the Bud platform, we're always thinking about, okay, anytime that a model is giving us an answer, we always need to be able to fact check it in the case that it's wrong.
[Thomas: 48:57.1]
What we always want to be able to do is go back to the source of that information and say, how has the model come up with this answer? I think people who are just taking something like a ChatGPT or Google's Bard and just layering it over the top of transactional data are going to get into these tricky situations where, yes, the model has hallucinated and made something up, or it could have been that it's just got a simple calculation wrong.
[Sam: 49:30.5]
Thanks, Thomas, a really good one here. Why haven't more businesses adopted this personalized approach?
[Thomas: 49:38.9]
I think there's kind of two things which businesses run into roadblocks for. One, is kind of understanding a clear roi? By no means is it a small project. There's definitely time which clients need to spend on getting personalization right.
[Thomas: 50:00.4]
And I think the second one is around kind of a belief that it actually works. I think what people are typically not necessarily scared, but maybe reluctant to do as well in the industry is if they've got an edge, which is helping their customers, they don't want to necessarily be shouting about it, right?
[Thomas: 50:23.2]
They don't want to be giving away all the answers to their, to their competitors. And therefore anytime so that they are using personalization to increase conversion rates or increasing product upsell, they're not in a position necessarily where they want to be telling everyone else about it.
[Thomas: 50:38.9]
So whilst it's not widely shouted about, I mean, if you know the players in the industry if you've been on their websites, you've seen who they're working with, you can see that there's definitely traction building in the industry and you can see that people definitely are getting results.
[Sam: 50:56.3]
Great. And then kind of, I guess, semi relate to that on the other side. Which of your current clients have best utilized these features and why?
[Thomas: 51:05.8]
Yeah, great questions. I think one of the first ones to call out is around First Direct. So they've been a big supporter of Bud for a long time. So all of the transactions which happen on the FirstDirect app all go through Bud's categorization to kind of understand and that's how they've kind of built their app from the ground up there.
[Thomas: 51:32.4]
I think two other companies worth calling out are Zopa bank in the US and also Totally Money. What they've both started to be able to see from Zopa's perspective, initially they were looking at, okay, how can they kind of understand and identify vulnerable customers, particularly before they may default on a payment or miss a payment?
[Thomas: 51:56.1]
And they've seen really promising results there. By using Engage and then Totally Money around, okay, they may not already have all of this customer information available to them, but by providing something like a Balance after Bills feature, they've managed to incentivize customers to connect their accounts.
[Thomas: 52:16.0]
Then by connecting their accounts they've got a lot more information around affordability and then that obviously helps them to create better offers for a customer as well as kind of providing value back to that customer. And so as I mentioned earlier, what TotallyMoney have seen is they're seeing a 20% increase in people taking out loans after they've connected an account.
[Sam: 52:41.6]
Excellent. Thomas, we've got questions flooding in here, so seven minutes left and we'll try and get through as many as possible. Probably just worth noting for everyone on the call that we will be putting a link in the chat if you wish to understand a little bit more about what we've shown you today.
[Sam: 52:58.1]
If you fill in those details, we can provide you with further information. We'll also be sending out the video and information associated with this webinar via email, so do look out for that in the chat. One of the questions here, which I know you and I talk about a lot with clients, what are the benefits of organizations using Bud or someone else's enrichments and insights instead of building it out in house?
[Thomas: 53:25.5]
Yeah, it's a great question, something that we speak about often with typically larger enterprise clients. So the first thing I'd say is not to kind of underestimate the complexity of what we're doing at Bud, our transactional models have been trained on billions of transactions across different industries, across different markets.
[Thomas: 53:51.5]
And the other thing to talk about is kind of this maintenance aspect. Now, as you, I expect most people are aware of with tech development is not something that's a one and done. It's not something that you can just create a model and then say, okay, great, this is done.
[Thomas: 54:10.8]
Things like merchant identification, location enrichment, these are ever changing and kind of a living, breathing databases which constantly need to be kept up to date. What we also see is as payment networks change, kind of the information that we receive and how different merchants may be shown within transaction descriptions are also constantly changing.
[Thomas: 54:36.5]
So one of the big costs to Baden, something that we spend a lot of time doing is making sure that our model is kept up to date as much as possible. I think we're on something like the 88th version of it. So by no means is it something that you can just spin up a small team, work on this, and then just leave it running for years and years.
[Thomas: 54:56.6]
There's this ongoing maintenance cost that you need to factor in.
[Sam: 55:01.0]
Absolutely. We'll try and fit in two more. Do you provide any aggregated reports about my customers?
[Thomas: 55:08.9]
It's a great question. So when we're thinking about Engage, we're typically thinking about an individual customer level. However, something Bud's most recent product, if you haven't seen it already, is called Drive. Now this is really looking at these portfolio level insights, so one step even higher.
[Thomas: 55:29.8]
So instead of an individual transaction, instead of a customer's transactions, we're looking across customers. So this really is around, okay, can I see all my customers transactions increasing in certain areas? What are the kind of segments that I can build out to from my customer base?
[Thomas: 55:48.8]
Is there sections of people who maybe write for a home? Are these customers? I want to see customers who are maybe living paycheck to paycheck if you're someone who's not a financial institution. So in the reward space we see people who want to see things such as, where are my customers spending money with my competitors.
[Thomas: 56:10.0]
So if you have a physical retail presence, you may want to see, okay, I want to see all the locations where my customers are spending money with X, Y and Z Merchant and then that can influence things such as, okay, where should I be opening my next store?
[Thomas: 56:26.2]
It's not something we do with Engage, but it's definitely something that we do within Drive. If you are interested, I would definitely suggest getting in Touch.
[Sam: 56:34.9]
Fantastic. The final one that's just come in. Would a good use case for a bank be to suggest moving payment dates to have better cash flow? Could this be used, for example, to help move first direct credit card payments when the customer changes pay dates?
[Thomas: 56:51.7]
Yeah, I think that would be a great example of personalization. Essentially, when we think of personalization, we think about what are the things that we can do to kind of remove friction and remove that thought from a customer's mind.
[Thomas: 57:08.7]
So instead of them having to think about, okay, my pay date has changed, so I want to change when I'm paying off my credit card. If these are personalized recommendations that you can surface to that customer and say, hey, we've noticed that your pay dates changed, so do you want to change the state?
[Thomas: 57:24.3]
And then the customer just simply has to tap. Yes. That's a great example of how you can increase personalization, which is not just going to delight the customer in that exact moment. What it's also going to do is build trust with that customer and say, you know, I trust this relationship with this bank, I trust that they have my best interests at heart and I trust that they're always going to be looking out for me.
[Sam: 57:51.8]
Excellent. So we're getting really close to time here. Just to remind everyone, we've put a link in the chat to get more information and get in contact with us. Thomas, I'm going to throw one last final curveball question at you, which just to finish things off, so I guess in 20, 30 seconds, if the audience were to join this webinar again in a year's time, kind of, where do you think we're going to be as an organization?
[Sam: 58:17.4]
What do you think we'd be showing them?
[Thomas: 58:19.7]
Yeah, it's a great question. As I've kind of mentioned in a few of these answers, one of the really exciting things that we're working on at the moment is around kind of integrating Engage into kind of large language models and how we can kind of create this personalized customer assistant for customers.
[Thomas: 58:41.3]
So not only will it be able to do things like answer simple FAQs, set up payments, it may be able to do things like provide personalized recommendations for products. Or as an example, if you may say to the chatbot, whatever it may be, hey, I'm looking to take out a credit card.
[Thomas: 59:02.0]
What it can do is kind of understand your current spending and say, okay, here are examples, or here is the best credit card that we think for you based on xyz. So again, it's kind of giving that personalization to the customer while also trying to reduce operational costs.
[Sam: 59:19.5]
Amazing. Perfect. We're at time. So all that leaves us is to thank everyone for their attendance today. We really appreciate the questions. If you want more information, there's a link in the chat. And yeah, thank you very much for attending.
[Sam: 59:35.2]
And thank you, Thomas.
[Thomas: 59:37.1]
Thanks, everyone.