Best Practices for Testing New Data Sources | Prince Matharu | Ep. 281

Hello everybody, Adam Parks here with another episode of Receivables Podcast. Today I'm here with a fellow deep data nerd. Prince joining us from TEC to talk about data experimentation. And the reason I ask asked Prince to join me for the conversation today is because in his role he has an opportunity to build data experiments and to participate in them across a variety of products, organizations and different variables, which gives him A really broad

point of view and perspective of what does it mean to build a good data experiment that's built for success and not just setting things up for failure. So Prince, thank you so much for joining me today. I really appreciate you coming on and sharing your insight. Thank you for having me on. So for anyone who has not been as lucky as me to get to know you a little, could you tell everyone a little about yourself and how you got to the seat that you're in today?

Certainly. So my journey started in the early two thousand. I was I'm a travel addict, so I was in Australia backpacking and a friend of mine had an interview, so she couldn't attend, so she sent me to GE Money for for an interview for a role collecting outstanding debt, which I had no exposure to previously, but that was my start. And then I was finishing my MBA at the time in Australia

I saw an opportunity to be at an organization that had leadership programs. So I enrolled in a leadership program at GE, fast forward. I worked there for numerous years, managed a big book of bad debt. both commercial and consumer, ran a team of probably forty collectors, other managers underneath, and then I moved to the US with that experience and found TEC, which has been my home now for 10 years.

We essentially my my job at TEC is to identify how data can transform many aspects, including rate of return. and overall uh revenue commitments with with good data and that's the problem. I solved on a day to day basis. So excited for our discussion today. I'm glad to have you here. Now TEC I know does a lot of things even beyond the data. Could you give our audience an understanding of kind of a holistic ecosystem that is TEC?

Yes. So TEC has a few umbrellas. One that we're perhaps most renowned for is our professional services where we help implementations, integrations. We have a deep, deep talent pool, probably the deepest talent pool across multi software reach. So any system, particularly in the third party collections arm industry, or vertical rather,

So we've got broad coverage from a talent pool standpoint. So that's our professional services arm. Then we have T E C Analytics, which is the division that I run and manage. And that has again a very unique positioning because we have a flagship product which is supplemented with white s white glove consulting But the technology that we built is a marketplace of vendors, a quick portal, if you will, where any client who is generally a debt servicer can Access.

up to hundred and twenty different data points. So and data products. So we have a vast coverage technologically speaking from end to end. So any company that has any need to append at any stage of the consumer life cycle, because if you think of a general consumer life cycle begins with exclusionary scrubs like bankruptcy, disease, military, then it moves on to enrichment phase, which is most commonly your phone append, your address append.

all the way to verified and specialty scrubs, which would be appending socials, appending date of birth. verified places of employment. So that technology is combined with deep comprehensive reporting. So we capture first hand the results of implementing these scrubs and then tweaking and and managing efficiently as to where the data is being profitable and where it isn't and right sizing the relationships. So that falls under our analytics division.

And then we have TC Solutions, which is a tech advisory wing, if you will. So again, we have very, very talented colleagues of mine that have vast and deep understanding of how to optimize digital strategies, telephony systems, procuring systems of record. and qualifying each product based on the needs of the client. So that's that falls under our advisory service named T E C Solutions. And then most recently, as of

April, May last year, we acquired Latitude, which was formerly owned by Genesis. Now we have probably one of the best systems of record, particularly for our scope, which is receivable management. So that is a part of T E C umbrella as well and we're dedicated to growing that and and making sure that that continues to service not only just

The third party market, the first party market as well. Lot a lot of different things under one umbrella, but today I want to focus on the analytics piece because I feel like that's it's one of those areas where everybody's looking to do more with lack.

But when we set out on the journey of a new data experiment, and we're going to test this new piece of data here. So often do we not define what a successful experiment is upfront. And then the goalposts continually get moved throughout the process. So we're kind of setting ourselves up for failure on minute one because we're never going to reach the goalpost if the goalpost keeps moving.

Now, I know you've been through this quite a few times, and we were talking about the cycle of hypothesize, experiment, measure, and report, right? And having that life cycle of a data test. When an organization comes to you and says, okay, I'm ready to go test this new piece of data. What advice do you have for them? And how do you help them structure that experiment so they can really understand the impact of that isolated piece of data?

Yeah. So the very first thing that I I discuss with my my prospects or clients is a little bit of myth busting, which is generally the perception tends to be And and it's no one's fault, but I think we've we've commoditized data, especially in our vertical, where we have some common myths that all data is created equal, and if a company is diligently seeking

advancements and investing in digital strategies, they need to equally emphasize a multi threaded, multi vendor strategy for data as well. Because the myth is that if they're buying f uh data from a single source, that's gonna be enough. So that's generally my f my my beginning set point. And then from there it follows the formula that you mentioned, beginning with the goal setting. And the goal setting would be dependent upon

the use case at hand. So it's going to look different for a first party bank. versus a third party servicer and we can discuss about what those specifics are. But though that's my starting point, that we need to create an exist take existing data, which is often a challenge because I would say seventy percent of the clients that we work with either have very minimal reporting

or no reporting. Meaning even in this day and age of AI advancement, I still on a day to day basis deal with clients that could not tell, and these are clients servicing bad debt. in particular could not tell us what the right party contact rate is, where is the the diminishing return from calling'cause we're focused on calling, but we couldn't definitively tell us when does the economies of scale hit? What's the optimized number of dial per phone, etcetera? So the point is to

supplement data in the right way to then able to uncover those things. But that's the level set starts with what is the goal, what is the use case. So starting with What are you actually trying to accomplish? But as we think about what they're trying to accomplish, if we get a better understanding of the end state that they're trying to get to. How can you help guide them toward that experiment process and

this is gonna sound overly simplified, but one of the the tools that I've used in the past is an eighth grade science experiment document. Quite literally an eighth grade science experiment document and just writing it down because I find so often that just writing down what the objective is and clarifying it into words. at least now it's no longer the unspoken. It's gonna be better or worse. Like we're able to now start to use language to define what success would look like.

at the end of the experiment. What kind of tricks have you used to to kind of get everybody onto that same page and marching in the same direction? Yeah. So that's that's a good approach. I haven't used that exact approach, but I might add that to my my tool to my tool tricks. And the the the formula or the methodology that I tend to use, there is two two buckets of coals, if you will. And I agree with you completely that we have to crystallize

this this the whole undertaking, this exercise of running a pilot or a champion challenger has to mean something in the end. And for us to circle the wagons and say, well, okay, we did this experiment, what results were we hoping to get and and what is that we're trying to get out of this as an insight. The two buckets that I like to think of one are the tangible goals. And then one off sort of intangible or maybe semi tangible thing. So the semi tangible goals

tend to be around efficiencies in existing FTEs that are or resources that are being invested into a batch process. It could be just the overall cleaning up of the data can have an intangible effect or improving customer service because now we're reaching more a reduction in compliance, etcetera. Although there there are and there can be metrics associated with those, I still like to think of them as the intangible side effects

of having good data and good testing that's gonna come from that. The second pool is the tangible bucket. And this is where the use case, based on the use case, let's if we stick with within the third party and the first party servicing bucket. or umbrella that would be establishing contact rate accumulatively. So how how many what kind of influence, impact, or increase can we create by implementing a multi vendor strategy, data strategy into the contact rates. So that's one.

The second could be speed of liquidation, so that's another one. And then the third one is reduction in dialing effort to get to the an increase of contact rate and liquidation. So that's how I like to frame between the the two buckets, tangible and intangible, because what I wish if I could tell the the market, especially in our vertical, that data has

such multi threaded impact onto our processes. Again, there is currently such a heavy focus on digital strategies. I want to tell the market that emphasis on data is equally important because data is the the lifeblood of every piece of technology that you're implementing. So yeah, but back to your question.

Yeah, tangible and intangible is how I like to think. approach is I look at the marketplace and we think our way through the the data sets that we're starting to use, being able to measure out like what is the value here. But you're looking at two aspects. You're looking at what's my forward new value, and then what am I saving in terms of reduced value, effort, risk?

But whatever the case may be, you know, from a a reduction standpoint. So it is kind of that seesaw. How much am I going to improve, but also measuring how much am I going to be able to save, avoid, or mitigate over that same experiment. Yes. Yeah. And the the experimentation also includes interface with as I mentioned, technologically speaking, 120 different data products.

So we have a vast array of available data products. So the experimentation also includes in tying the right data product to the experiment or the test in question. So for example if it's a law firm that's following a very strict legal path to resolution that w and and they generally have greater margins.

then the discussion and the scope of the experimentation would include heavy use of verified data products versus a third party agency who's servicing the debt, the margins might not be there or for other reasons The emphasis might be on a phone scrub product or an address scrub. I also recently learned of a data product that. does two things. It's verified phone hits, which are quite unique, and warm transfers.

So this provider can dial on behalf or do a call dial so not disclosing who they're calling on behalf of, but establish a contact with the right party. and transfer that directly. So it's all about understanding the first part, which is the goal setting. What is the target? Is the target to improve the or is it to improve the liquidation or both? Or is the target to just agitate the pool and see what comes out, which is

probably not the most effective way of going about things, but different companies have different intentions. So it is not just yeah, my point being it's not just The experimentation of going from one vendor to a multi-vendor strategy, but also looking at the vast area of products that are available. And distinguishing between verified and unverified products. Excuse me, losing my voice a little bit. Yeah, it's all right.

It seems like the industry itself has been moving towards self service and digital communication technology. And as we think about those tools I always go back to these tools don't add value unless we're powering them with the right data. And then when you when we were preparing for the webinar that you and I had done a couple of months ago, we started talking about data decay. And if I remember the number correctly, it was thirty two or thirty three percent.

of data is going to decay on an annual basis, which means two years in, sixty six plus percent of your data is no longer valuable. When you're building out these data strategies, how do you look at the data decay portion? It's kind of the first part of it, but are you seeing more organizations that are actively experimenting as they're starting to realize just how quickly the data they have has lost?

a single file or or or a handful of accounts, whatever the the deeming sample size is, is sent to one single source of data and then what's measured at the number of hits. So how many hits were returned. Which is often misleading because data vendors are very good at returning hits. But in my experience when we work with setting up a test, we work with the data vendor and their inherit intelligence to refine the configuration and refine the criteria for a qualified return so that we're getting

hits that are going to be high caliber and high quality. And the data vendors are very gracious in because they want the the best possible result from the test as well. So so that's one shift. And the second again fum if we think about how historically tests and experiments are run, the emphasis is very high on the hit. The second portion is then those hits are worked in some capacity, which is often undefined. Again, we don't take the

approach of defining that we're going to for example in in in the case of a phone test, that we're going to dial each phone X amount of times. It's not predefined. And then the third thing is rarely is the impact measured in terms of liquidation or the overall impact in the contact rate. So what we do when we run a test, first of all, it every test we run is across multiple vendors. So we're removing that

no single source. And I've personally consulted for many data vendors and I feel confident in saying that no single data vendor can provide optimal coverage. So the very first thing that we do in our methodology for testing is we implement a multi vendor strategy. And the way we do so that we're doing We're being fair across the board.

is we would split and create strategies where each vendor gets a placement in in inventory and volume in first position, meaning they're getting the first look at the account. and then every other position thereafter. So it's a very comprehensive way of testing and at the end of it the insight you end up with is you can confidently tell from a statistical standpoint then when vendor A was implemented in given accounts in the first position.

without any interruption or any other disruptions, what was their coverage, i.e. how many hits they could return. And then we enforced probably is a harsher word, but we work with our clients to make sure that data and inventory is worked equally across the board so that every vent data gets worked, because that's often a missed point that, you know, vendor B provides equal amounts of hit.

But somehow something happened in the dialogue campaign where we ran one in predictive, the other one in preview. So there's a lot of variables and variances that can impact the performance and the end result. And the third point is that we make sure that we're capturing some sort of leading metric and based on the use case and it's based around

what we're what we're driving at in the end of the exercise. What's one takeaway that we want to walk away with? It could be as simple as increase in liquidation, but across the board we're looking for one key defining metric. It could be in increase in in in your contact rate. It could be speed at which the liquidation happened. It could be return in return mail if we're if we're discussing addresses, etcetera. So That's how we approach it. We begin with a multi vendor strategy.

we would give equal opportunity to each vendor in every position. So starting with first position and uh Probably an easier way to think about this if we had hundred accounts.

we would create and if we had four vendors, we would create four different sequences where vendor A, B, and C and D get the first position in each one of those sequences. So each vendor can at the end of the day get first look at accounts, show us their coverage, show us the capacity to to provide qualified hits and results, and then we work with our clients to make sure that the effort is uniform across the board.

and then we measure that one one defining value, that one defining metric. Interesting. Now the structure I'm assuming is feeding pretty well into these digital strategies because now you're identifying who's going to be able to provide what at which point. You're documenting that process. You're determining what success looks like.

in an early stage and then you're starting to feed these tools. Have you seen an increase in Organizations' willingness to experiment with new data as you've seen the adoption of these digital channels rise. The phone numbers are are important, but now we're looking at so many different rate email addresses and so many other things that decay even faster. I think my experience has been mixed. There's there is ton of excitement. Definitely I'm getting a lot more curiosity.

around digital strategies coming to coming through me. But we are also implementing a lot of clients on on emails and sell only scrubs. which are probably a better choice for texting strategies or phone data that is coming up with some sort of line type indicator so that we can separate the voice over IP and the cell lines, etcetera.

And create digital strategies, especially texting strategies. So I think my definitely the market is much more curious. There is some hesitation still around emails and again I'm not a lawyer. I don't to p P one online but but um there is there is still some confusion around can email be used as an alternative to pre to you know, historic methods of contacting the client and correspondences. So again, we we've had and I've been part of certain discussions where it's well

clarified that they can be used, but there is still some hesitation. But but overall I think there's a lot of excitement and curiosity in the market about this and where where we've implemented email we've seen some stunning results in terms of the delivery rates, the open rates, very, very healthy. And you can see why that is the future. It has to be quickly if it's not already on its way to becoming table stakes where it would become very fundamental to the business.

to run digital strategy because they're so efficient. They're the speed at which correspondences can happen, contact can be generated, et cetera. So I'm I am along with the market. I'm very excited and keeping an open mind and testing various, various sources to find

data sources for our clients. Do you think that the increase in adoption for artificial intelligence is also going to drive more people to look at the data? Because again, just like the digital strategies, we can buy the fanciest tool. But if we are not fueling it correctly, like we can go buy a race car, but if we're putting bad gas in it, we're not gonna go very far, very fast.

Yes. How do you think the artificial intelligence deployments and adoption are gonna impact what you're doing from a data perspective? I think the first thing I would say is again, around with AI, there is a lot of curiosity and excitement in the market and might m first urge the the audience

or any anybody entertaining the idea of implementing AI to look for truly agentic sources of AI. Because AI can be misleading if we just stick with general terms, right? So what I've seen when implemented what performs is true agentic AI. that that is not just an LLM language model feeding scripted answers back, but it can it it can have some rationale behind it. So that would be my first comment.

But secondly, yes, absolutely. The age of AI is upon us. So once we implement agentic AI with with rationale, then absolutely the data becomes it is sort of that one two punch. If we want to assure success, then we have to emphasize the data. And thankfully data vendors are coming along for the ride. You we have to understand that they are investing

large sums of money annually into their research and development and building strong algorithms, et cetera. So Thankfully the data vendors are also responding to the the upcoming trends of AI and digital strategies and now creating products that are going to feed into those strategies. all the way t from unverified to verified sources of information. We've got vendors that can track individual consumer movements and that sources can be fed in some instances. So

again, it's it's it's a great time to be looking into and and uncovering'cause the possibilities are becoming quite exciting. New and interesting data seems to be arriving. to our states. You know, we saw license plate data get added into the mix a few years ago and what did that mean both from a

repossession perspective and from a location perspective. And then we saw a lot of changes and people working remote and the challenges that came with that world. But I think is a lot of the world has started moving back into the office for the employment verifications and those types of data sets that life's have become more stable since twenty twenty. And now we're gonna start to see these data patterns start to play a larger impact in our ability to predict the future.

And that's kind of what we we what we have to do as debt collectors is try to predict When or why would they answer the phone, right? Like where are they? It's all it's a lot of prediction in terms of is this account collectible? But Yeah, I think you're right. It's a great time to to be in the business with all of this new information. um and data points that we can start turning into actionable intelligence.

It's got a lot of signals. But drive in that. Satisfaction is our objective with a lot of these data twists. Now, for anybody who's kind of still holding back in terms of my data works just. fine and they haven't rebuilt their your waterfall they're really taken a look at in the last five years or so. What advice do you have to those that have not actively partaken in improving their data waterfalls over the last half?

my advice and request would be that if it's a very high risk strategy to keep your data s data at the back end or not emphasize it. I think it's data is the new competitive advantage. If you have a multi threaded, multi vendor strategy that is dynamic enough that gives you reporting to see which data source is performing for you and which isn't, unless you're doing that.

you're you're running a high risk strategy, in my opinion, because you don't have that competitive advantage of accelerating liquidation, accelerating contacts, accelerating customer service, or reduction in compliance. events, etc. So yeah, my I urge everybody listening to if you haven't evaluated your data, especially in five years, because the law has changed in five years.

than to be curious, to cu to curiously ask questions as to reach out to your data vendor, reach out to T E C and w we'll be happy to have a have that dialogue. to really take a deeper dive into your consumer treatment as to how you're dealing with consumer today from a touch point standpoint, either it's an established contact or correspondence and how that can be accelerated or efficiencies that can come from deploying multi multiple sources of data. Increasing your coverage.

increasing the footprint that you can reach out to at an accelerated pace through digital strategies and then all the end metrics that matter, which is the the revenue, the liquidation, the reduction in compliance, the efficiency from an FTE standpoint, all of that is associated around data. There is no doubt in my mind. That's the data plays a pivotal role in all those different segments of your business. When you say that data's a competitive advantage in the future

I think you're right on point. And it's not just about the raw data itself because there's so many data vendors and even being able to go out and buy the data is not the end solution. It's about being able to interpret those signals, turn them into actionable intelligence, and then take the next

action with it. And that is where I really see that competitive advantage solidifying itself and being something that becomes eventually Easier and easier for an organization to defend and harder for a new organization to penetrate the space because they're not gonna have that historic context. Yep, absolutely. Yeah. Most agencies, for example, compete on a scorecard. So we're already cognizant by design to pay attention to metrics that we discussed. It's

the overall aggregation of contact rates, the speed and acceleration of liquidation, etcetera. So those are already points of competition. I think it's us tying the connection that Data is that one solution or could be that one missing link that can have impact across the board and make you successful in so many different ways.

Prince, this has been an absolutely fantastic conversation. Every time that I sit down and talk with you, I learn a little something about the data waterfalls and how I could start looking at attacking data across the debt collection industry because there's so many tools, there's so many new ways of doing things. And if we're not going to feed it with the best available data, we're never going to get the most out of these platforms.

Great. Thank you for having me on. Hopefully, for our listeners, this was insightful. Yeah. I I urge everyone to to be curious, keep experimenting and There is a lot lot of fun. Again, I'm I'm a standard data nerd, so maybe maybe this is too exciting for me personally. But um hopefully we can infect with with some of the curiosity around data and and

have some sort of insight around that. For those of you that are watching, if you have additional questions you'd like to ask Prince or myself, you can leave those in the comments on LinkedIn and YouTube and we'll be responding to those.

Or if you have additional topics you'd like to see us discuss, you can leave those in the comments below as well. And hopefully I can get Prince back at least one more time to help me continue to create great content for a great industry. But until next time, Prince, thank you so much for your time. I really appreciate all your insights.

Thank you, Adam. Thanks for having me again. Absolutely. And thank you, everybody, for watching. We appreciate your time and attention. We'll see you all again soon.