PegaWorld 2025: The Journey to Omni-channel: Navigating Paid Media Challenges with Navy Federal Credit Union

PegaWorld 2025: The Journey to Omni-Channel - Navigating Paid Media Challenges with Navy Federal Credit Union

Thank you for coming here to listen to our clients, Navy Federal Credit Union, talk about their omnichannel journey. I'm so excited to introduce Caitlin McMillan, Chase Tripi, and Arijita De. They have an amazing presentation, and you're in for a great show.

Caitlin McMillan: All right. Well, thank you everyone. We are really excited to be here. I'm Caitlin McMillan and I am a manager of our orchestration and analytics team in marketing at Navy Federal. My team's focus is Pega Customer Decision Hub channel integration and new capabilities, as well as our personalized Member Insights program.

Chase Tripi: Hey everyone, I'm Chase Tripi. I'm one of the assistant managers on the Marketing analytics team. My team covers credit cards, debit card, prepaid, and membership acquisition.

Arijita De: Hi everyone. My name is Arijita De. I am a Pega System Architect manager. My team focuses on solutioning and implementing Pega CDH solutions, which provide intelligent and real time decisioning for our marketing folks.

Caitlin McMillan: Great. So I will give a quick overview of Navy Federal. We are a credit union that serves the United States Armed Forces veterans and their families. We are member owned and not for profit with over 14 million members, and our members financial goals are truly our number one priority. We've received a lot of honors over the years for our ability to support their financial goals, and are really proud of that.

So we are highly member centric, and member experience or customer experience is very important to us as well as those financial goals. And one way we do that is by really planning out an omnichannel experience where our member insights and offers are all in a single unified decisioning engine. And that decisioning engine for us is Pega Customer Decision Hub.

We chose Pega Customer Decision Hub because it does two things really well. First is the adaptive modeling and second is the cross-channel orchestration. So the Pega brain that you see here in the middle of the slide, that is what is determining what the next best action for each of our members is, and where the content will present for all of the channels that you see in orange, which we have integrated to date.

And the key to having or the key to achieving omnichannel success is to ensure that you have your member experiences, insights, and data from all of those cross channels collected into Pega so that you're able to action off of them and make an informed interaction as the next interaction with the member.

And so one example of this is if I'm a member and I go into one of our branches and I'm speaking to a member service representative, and they happen to present an insight to me and I decline it. We don't want to send me an email with that same insight or offer the very next day. The opposite is also true, where if I accept an offer, sometimes you need to give me a little bit of time to actually complete that action, such as setting up a direct deposit with Navy Federal. And so we don't want to overcommunicate with a member in that case either.

So today our focus is really our paid social implementation journey, and it has been quite the journey. As many of you likely do for your companies, we have been advertising our products and services across many social media sites for many years. And today we're really talking about Instagram and Facebook in particular via meta.

So we have been building these base audience campaign lists out of our marketing tools for years. And what we'll do is we take those lists and upload them to meta, so that meta can match the target members that we have and any lookalike prospects with the associated paid campaign and the creative so that it can be presented in the target user's Instagram or Facebook feeds. And that has worked really well for us.

So when we migrated to Pega in 2021, we used that approach. That was a batch and blast capability that helped us do exactly that. We weren't really adding on any bells and whistles. We were just kind of lifting and shifting our previous process, and then we moved on to new integrations.

Well, that worked for us. But as we continued getting deeper into our maturity with Pega Customer Decision Hub, we realized that we had a lot of great lessons learned and success stories with the use of Pega's next best action approach, which is far different from the batch and Blast approach. And Chase will describe both of these shortly.

And so it was really the time for us to do a little bit more testing and learning and see if we could leverage Pega's paid media manager feature in coordination with the next best action approach to dynamically update the campaign list that we were providing to meta and get a lot of benefit out of that.

So we set out for a test and as with every test, you have multiple phases. So first we needed to prepare. This of course included needing to define our approach, defining our KPIs, as well as preparing our internal and external teams that were going to be involved in this change.

Then we had to build, test and optimize, which was quite the process, and Arijita will be speaking to that shortly. There were a few things that we needed to overcome in order to get this test fully live.

And then thankfully, it was a success and you'll see some of those results shortly. So now we are in this third phase of scaling, and that is where we are integrating additional channels so that we have even more power in this single unified decisioning engine and completing a migration to Pega Cloud, which will give us more performance power so that we can complete the migration of all of our paid media campaigns into this new way of doing things.

And now Chase will talk about the prepare stage.

Chase Tripi: Thanks, Caitlin. Good afternoon everyone. I'm excited to share how Navy Federal transformed our paid media approach by implementing Pega's Next Best Action. Our goal was to deliver highly relevant, personalized offers while keeping costs in check and ensuring timely outreach across every channel that we own.

However, we did face some significant hurdles. Our legacy model relied on monthly batch processes that too often delivered stale or irrelevant offers. Manual workflows increase the risk of errors and high volume concurrency across our digital channels like email, IVR, web contact center, and online banking pushed our infrastructure to its limits.

Firewalls and evolving API requirements, especially from meta, introduced some frequent disruptions, and shifting from batch and blast to real time decisioning required us to win over not only our internal teams, but agency partners to really embrace these new processes.

To tackle these challenges, we first centralized all decisioning in Pega Customer Decision Hub. Instead of having siloed per channel batched workflows, we now maintain one single source of truth that powers every outbound channel.

We rolled out the dynamic NBA across email, IVR, web contact center, and online banking to ensure we had consistent personalization across the board. And while direct mail still does follow a traditional batched cadence, its targeting is now informed by the insights that we can glean from our daily NBA runs. And finally, we established a new file delivery cadence for social media and agency teams, and conducted some cross-functional training to really align these processes and drive adoption across the organization.

And looking forward, our optimization strategy focuses on leveraging that first party member data to fuel the propensity driven experiences. By continuously recalculating the next best action in real time, we can respond to live member interactions with tailored offers.

Our strategic priorities do still remain clear to us, though. We want to become truly omnichannel with seamless member journeys, eliminate wasted ad spend through precise targeting, and drive higher conversion rates at every single touchpoint, while identifying and acquiring our high value prospects as efficiently as possible. All this while creating the quality experience that our membership has really come to expect from us.

Next, what sets the Next Best Action apart is its dynamic nature compared to our static batched approaches. We've demonstrated some notable lifts in both CPA and engagement metrics versus those traditional methods. And with Pega, the decisioning is unified within that one single platform, so providing a consistent foundation for every channel and our ability to segment those audiences, even at a granular level within Meta's ecosystem, has further enhanced that targeting precision and the scalability of automated daily updates has virtually eliminated the vast majority of our manual processes.

So ensuring we can do all this and maintain high performance without having to add additional headcount. And now Arijita is going to discuss our testing, optimization and some of our implementation challenges.

Arijita De: Thanks, Chase. All right. So first off, thank you guys for showing up to this session. I'm really excited to talk about what we did for our paid media implementation.

I do want to start by saying that we started our paid media implementation in 8.8.2 version. And when we started, Pega already had a new version out there. So the major issues that we faced was because we had not upgraded ourselves to the latest version, we just weren't ready yet to go to 8.4. So we started our paid media journey with 8.8.2 version.

So the first issue, that big issue that we faced was deadlock. So the way the paid media works is that let me give an example. So the first day you run the NBA and you say, okay, I want to target X members for paid media. Those are all brand new members right. It's the first time you're running it. So let's say the output is thousand members. You target all of those thousand members.

But the second time you run it and every subsequent run, now, some members may have fallen off from the initial list and some new members might have been added. We call them the deltas. So when we were on 8.8.2 version, our data set was actually residing in a physical database and it didn't know how to add and delete at the same time, causing a deadlock at the database level.

So again, Pega was great. They worked with us. They helped us understand. And then what ended up happening is that they already had this in their plan where in 8.8.4 version they were moving this physical database into Cassandra. So for 8.8.2 version, they did work with us, give us a hotfix, but they already had a fix in 8.8.4 version where by moving that to the table into Cassandra, which knew apparently whether to insert first or whether to delete first, we were able to resolve this issue and by the time we launched, we were on 8.8.4. So we didn't have to do any hotfixes or anything. It was just there.

The next one I'm going to be talking about is the filtering. So what does filtering mean. So essentially every organization that wants to market has to decide how many members it wants to market to. Now yes, we at Navy Federal Credit Union, we do have 14 million members, but not everybody is going to be eligible. We do have members, for example, who are under 18 years, and I can't legally market to those members. So those are all criterias, and it depends by what kind of marketing you are doing, what kind of organization you are, what kind of business it is. Those are all configured in Pega. You can configure them in the Pega system.

So industry wide there's two ways of filtering and Pega calls them. One is called the volume constraint which basically limits the total number of members in a day you're going to be targeting with any marketing campaign. And you can have several campaigns going on. You can have, for example, a credit card, hey, you're eligible, applied for this credit card. Or you can say you are also eligible to apply for a loan. So that would be two campaigns.

The other thing that Pega helps us do is action limits. So that credit card is one action. That loan is another action. Now what was happening with filtering is that in 8.8.2 version when we applied the volume constraint, which for example, let's say we say we only want to target 1000 members in today's marketing campaign. What that means is that it can mean that 1000 people get one campaign, or it's possible that 100 people get two campaigns. So that makes it 200 and then 800 get one campaign.

So what was happening is that this limit applies to the entire NBA, which has several channels in it. For example, the email, the IVR you can have call center and paid media was also part of that NBA. So at Navy Federal we didn't want to put any volume constraint on paid media, but we wanted volume constraint on the other channels. That was something that 8.8.2 did not have. It was later put in Infinity 23 is when they put out that feature.

So again, Pega was great. We worked with them. We put in a hotfix where they said, okay, if you see paid media as a channel, just exclude it and everybody else gets the volume constraint versus paid media doesn't get it.

The other thing with filtering is the action limit that I was talking about earlier, where it's like, well, yeah, you want to send out 1000 total, but of that 1000, only the credit card can only account for 100. So that means that even though somebody is eligible for a credit card, you can't market to them because my action limit is 100. So that was something again, we couldn't configure for paid media. Everything else was working as expected, but we did not want that limit to apply to the paid media action.

So again, we worked with Pega where the same hotfix that applied to volume constraint also applied here. And we were able to say, okay, this is a paid media campaign, and this action can be part of paid media or email. But when it's part of paid media, don't apply the volume constraint or action limit, but when it's part of the other channels, please apply the action limit. So that was filtering.

These were the two major ones that we faced. The other two are actually teeny tiny, not as big, but the deployment one. What happens is that when you're trying to target, this is specifically for meta, by the way. So the way it works is that because, as Chase said, it was all automatic. You run the NBA, the output is generated is pushed to Facebook or Meta directly. How does that happen? You have an ID that it takes with password. Of course, you know that you are going to be using to say, hey, this is Navy Federal's meta ID, please use this and target our members.

What was happening is that we would configure the ID in dev and prod both, but when we deploy the entire ruleset from dev, what happened is that the prod ID got overwritten. So that was something. It was, again, not a big deal. It's a teeny tiny one. So every deployment would end up happening is that we would as part of post- deployment checks, we would go in and then change the production ID again. This was again fixed in Infinity 23.

The next one is the firewall. And this was a completely internal Navy Federal challenge that we faced. Pega was instrumental in actually helping us identify that this was an issue. So for our firewall, we had used a specific meta URL to clear it through our firewall. What we didn't realize is that there are several other such URLs that needed to be whitelisted and go through our firewalls, so Pega helped us identify the fact that we had to generalize it. So when we generalize it, it just works. So again, not a big one, but Pega was instrumental in letting us know that's what was happening. And the last one I'm going to talk about is LiveRamp customization. So LiveRamp is actually a third party tool that helps us target our customers better. What does that mean? Okay, so when you go into your banking partner, you give your name, you give your social, you give all of that. You don't give out your phone numbers. Or maybe you do, you don't. But you definitely don't give up your Instagram handle or your Facebook IDs. So how do we target you on Facebook or Instagram or any other social media? Because that's something we don't have. So we do it with a series of matching with the information that we have. And we do a lot of permutation combination.

Now, what LiveRamp does is it is really good at these matching process. So yes, as Navy Federal we can do the matching ourselves. But we may not be as efficient as LiveRamp is because that's all LiveRamp does.

So the way we were doing LiveRamp initially was actually using a batch and blast, and we would manually run the batch and blast, get the output out, we would manually send the file in, which was a lot of manual intervention, and we wanted to automate all of that.

So we worked with Pega. Again, there's something called segments within Pega CDH that you can actually configure. So initially we couldn't do LiveRamp over segments, but we worked with Pega. Pega helped us set up those segments and segments. Once you set up, can actually be put on a job scheduler and a schedule run.

So once it was set up at this point, I think we have 20 paid media actions that go out to LiveRamp on a scheduled run. And everything's automated. They run on their own schedule. They spit out a file, one single file, and then that action and that file is sent out to LiveRamp.

So again, we work through all of those challenges, issues with Pega. And as Chase will talk about in the next slide, we were able to launch our MVP. And we did go through multiple iterations of that. So over to you, Chase.

Chase Tripi: Thanks, Arijita. So after deploying the dynamic NBA, we immediately saw positive changes. Our social listening team observed that members began reporting that offers felt more timely and relevant. And our teams experienced a much smoother workflow with less manual intervention and even fewer data errors.

Quantitatively, link clicks increased by 12.8%. Credit card applications surged by 26.3%. And on our cost side, we actually reduced cost per click by 6.6% and cost per application by 14.4%.

But perhaps most importantly, gaining these real time eligibility updates enabled sharper targeting. And just like we heard from Frances this morning, using Pega to free up our staff allowed them to focus on strategic initiatives instead of just manual batch processing.

And you can really see the manifestation of that sharper targeting here in these charts, reflecting that each day the NBA ran, there was a small volume of members that were either removed or added to the live offers. Targeted audience. So although the delta might seem small, our targeted populations might range from the hundreds of thousands to multiple millions. So a decrease of 0.3% could be 1500 members to 12,000 members on any given day.

So it's here that we really realize the benefits of dynamic targeting, right? So not only cost saved, but we've created a more efficient, engaging and product increasing environment for members to interact with. And next, I'll hand things over to Caitlin to discuss our future plans for channel scalability.

Caitlin McMillan: Great. Thank you. So again, those were just the results from that one initial test. Since then, we have completed additional iterations, as we mentioned, by just rotating in and out different offers and looking at the effectiveness for different product categories.

I'll skip to the middle one first. We do plan to migrate the rest of our paid meta campaigns into this approach of using the Pega paid media manager and next best action. We just need to get through this kind of hurdle that we have of having the appropriate performance juice to make all of that happen. And so once we complete our migration to Pega Cloud at the end of this month, like I said earlier, we're really optimistic that we will be able to really ramp up the volume and the quantity that we'll have in the NBA for paid media without negatively impacting any of the other associated channels. So that will be really exciting.

We also have our mobile integration underway, hoping to launch that by the end of the year. And this will just give us one more channel that is kind of part of this overall puzzle and piece to that omnichannel experience. This is a really important one to us due to the high number of mobile users that we have every day. So it gives us a lot of interaction opportunities to not only present new insights and offers, but to also get that feedback and kind of continuously try again. So we're really excited about that one.

And then as I was mentioning, you know, we have LiveRamp we have a lot of other components for this overall paid media ecosystem that we haven't quite fully optimized. We have a lot that's working for us today, but there could be more efficient ways of doing things, you know, more data points that we could get back. And so that would all be on our roadmap for future years.

And so in summary, you know, we had these three core phases where we prepared for the change. We built, tested and optimized. And now we plan to scale.

In preparing for the change, you know, of course you want to set out what your objectives are. Really define your approach and then make sure that you have that proper alignment. In our case, you know, it came down to things like establishing a new delivery cadence and also making some compromises. You know, it's not in real time, but does it need to be for this channel? Are we still going to get enough value out of it? And we definitely did.

For testing and optimization. We went through quite a few hurdles, but thankfully it was all worth it and we had great support from our Pega team as well as our Coforge partners and many groups within Navy Federal. And so we did that initial launch and subsequent iterations.

And then again, we are now in that planning to scale phase, which we hope to have really underway, I believe, by the end of this year.

So that is all that we had for you today. If you have any questions, please come up to the mics and we're happy to field those.

Q&A Session Audience Member 1: Hi rookie question. You guys pay to advertise to your own members?

Caitlin McMillan: Yes. So there's always an opportunity for some cross-selling, upselling. I believe it's a little bit more of a focus on our prospects, but we do have some engagement or cross-sell campaigns that we're doing. Chase, do you want to speak to any of those? I know your team's more involved.

Chase Tripi: Yeah. So we have, you know, we have organic channels where we put content on for members to see passively. But when it comes to trying to getting a specific offer into somebody's timeline, you know, there is a paid approach there. So we have our paid channels like social media, paid media.

And like we said with LiveRamp, we give them these social media lists on a somewhat regular cadence, and we're actually increasing that cadence. You know, LiveRamp has asked for more frequent. It's been such a success that they've asked for more frequent lists from us. And those go to campaigns like Uber, Lyft partnerships where we might not own that channel, but we can provide them a list of individuals to give them a benefit. So cool. Audience Member 1: And I guess you're probably filtering those lists on products those customers don't have yet, right? So, yes, customer no doesn't have the product show up in the feed.

Chase Tripi: Yeah. You know, it depends. You know, we have a lot of our, you know, the pre-approval. The invitation to apply is only go to individuals who don't currently have credit cards. And that's kind of like my team's bread and butter. But, you know, we've entered into the space of, you know, second card marketing. So it depends. Sometimes you might want to see if somebody could benefit. And that's the benefit of Pega, right, is if we know someone might have a propensity to engage with a product they might already have, perhaps they're going to be in the market to get another one.

Audience Member 1: That was going to be my next and last question, which is, since they're members, you probably maybe have even more information about them than a regular bank does. I don't know if that's true or not, but do you guys have adverse selection kind of things? When I hear, like we drove Apple Card applications up 26%, I think God, were those the people we wanted to apply or were those people who, you know, I mean, obviously we want everybody to have credit, but some people could be an adverse selection play right where you get the wrong people in because you guys do. You look at conversions, do you use more data because you're a co-op or how does that work?

Chase Tripi: So I think the lift that you saw in our platform, it's like, yes, these are people that we wanted to bring in, you know, we're very selective about, being a credit union and especially being military focused. We you have to get product acquisitions right to keep the business running. But we feel like we have a responsibility to our membership to not put credit cards in everybody's hands. You know, we try to take ownership of your financial well-being.

So if we believe a credit card is the right product for you, we'll solicit it to you. But if we can see all the data that we own of you, and you might not be financially in a situation where credit might hurt you, we're not going to offer you that card.

Audience Member 1: A threshold propensity for that or your other underwriting.

Chase Tripi: Yeah. So we yeah. So in terms of like financial wealth and financial health, we actually have a modeling team here that gives us financial distress models and different models that help us determine someone's financial well-being. They kind of holistically put that together. And then we plug that into our credit card applications and Pega to make sure that, like if we're showing you a credit card offer on Facebook, that one, you're actually eligible to receive that credit card. And then two, is it the right thing for you? Is your final financial situation in a place where you could open a credit line?

Audience Member 2 (Rakesh from Citizens Bank): Hello, my name is Rakesh from Citizens Bank. So we're kind of in telling you guys, but in the same boat. I just wanted you to understand. So I understand you mentioned that, you know, you started your paid media journey with 8.8.2. And you mentioned about LiveRamp, right? So curious to understand. So CDH has something called identity matching, right. Is that something that was not available in that version as an option to basically identify and tag them to your existing customers whom you are marketing to instead of, let's say prospects, on the paid media channel. And in that case, if that's the case, right. How do you how did you work to tie those the dispositions back to the real customers through paid media.

Chase Tripi: So our paid media approach right now is through meta. So Instagram and Facebook. So when it comes to identity matching, we have to be able to upload a list of email addresses into meta for them to make those matches. So from an identity match standpoint, I think with us on the relationship with meta, I'm not sure if that would be beneficial for us right now, because we might not be able to know if the email address that we have on file for you is the one that you have on file with Facebook.

So that's probably answers that question. The latter half of your question was around, was it LiveRamp? Sorry. The dispositions, the dispositions. Oh, the returning dispositions from from Facebook. So the actual take up data is fed back to us from our social media listing team. In terms of the dispositions. I can't recall the actual feedback. We're not getting one back today from meta. So that's that walled garden. I think there are some solutions out there. It's just not one that we've prioritized yet.

Caitlin McMillan: And so it is, other than direct mail, it's the only other channel where we don't close the loop on direct feedback into Pega interaction history. But to Chase's point, we are getting that data from meta. It just is not at the individual level. So it can't be used in Pega, at least with our current setup. So we're kind of having to connect some of those dots outside of the tool a little bit more downstream. So still able to get to results, but we can't quite action off it. We just know who was decisioned for which campaigns.

Audience Member 3 (Kumar from NAB): Hi my name is Kumar and I'm from NAB. Keen to understand, you know, about web personalization or on site media. So are you using Pega CDH for that as well? For web personalization? And what kind of offers you know, you are putting in front of a customer in an authenticated space as well as in an authenticated space.

Caitlin McMillan: Great question. It's on our roadmap, but we're not there yet. We are currently only serving up what we call insights and offers on our authenticated web. So our online banking and then we'll be extending into our mobile app. I think doing that on our website, pre-authentication, would probably be in the next year or two. It's something that we've explored.

We are getting some data from our pre authentic website and kind of funneling that into Pega so that we can action on it with trigger based emails, but not doing any direct Pega powered personalization on our websites today.

Moderator: Any other questions? Great. Well, thank you so much for your time. Really appreciate it.

Speakers: Thank you. Thank you.