Nonprofits, Direct Mail, Strategy & Planning, Technology Evaluations

Nonprofits today are facing more significant pressure to raise more money and do so with increased efficiency and fewer resources than ever before. The pressure to “do more with less” has only intensified since the beginning of 2020 when the COVID-19 pandemic began, creating unprecedented demand for the services nonprofits provide and one of the most challenging fundraising environments in modern times. The ability of nonprofits to withstand – and even grow – in the face of these challenges depends on their ability to understand and respond to the unique needs of each individual in their donor base. AI is helping many nonprofits achieve this level of personalization and fundraising effectiveness at scale.

Slide Deck


Adam: The topic of today’s session is artificial intelligence, what it is and how can you use AI to actually increase your individual fundraising program and start to get some of that growth back to the revenue it can produce.

Tina: Hello, I’m Tina Wright, Annual Giving Director with Idaho Public Television.

We are the public television station and PBS affiliate serving the entire state of Idaho. As a nonprofit, our mission is to connect Idaho communities with inspiring, informed, and educational programming and services. We began working with Arjuna solutions in 2017 using their ExactAsk algorithm treatment for a segment of our direct mail campaign.

Our overall fundraising program was strong and our direct mail campaign, our largest revenue generator was the obvious channel to test. Arjuna solutions presented that through using exact desk, we could gain revenue with very little effort on our part using donors already on our file. They presented their case clearly, and I could easily understand the opportunity to realize a gain in ROI simply working smarter, not harder.

The actual process working with Arjuna solutions was simple and efficient, and most importantly, our donor information was encoded and private. Truthfully, I was pleasantly surprised that the first test was extremely successful with a net revenue increase of 59% and more importantly, an average increase in lifetime donor value of $4.87 cents.

I continue to work with Arjuna solutions on all of our direct mail campaigns and have consistently seen a higher return on investment and increase in lifetime donor value with each campaign. Given our success, it’s an easy decision for me to budget using Arjuna solutions. They have been efficient and professional to work with providing clear and necessary reports that I can present to my upper management and to track our progress.

These are extraordinary times in the world and in fundraising, I am confident working with Arjuna solutions will continue to be a profitable investment. And I believe this will be especially important moving forward. By using Arjuna solutions, Idaho Public Television is working smarter, and our ever-increasing bottom line is proof.

I wouldn’t hesitate to recommend Arjuna solutions to others that want to increase their fundraising revenue and donor value with very little effort and investment.

Adam: All right. So what I do want to start off by just laying bare what my goals for today’s session are. And what I’d really like to do is help everyone get a fundamental understanding of what artificial intelligence is and how this kind of technology was able to deliver that type of an impact for Tina Wright and her team over at the Idaho PBS station.

Hopefully everybody will walk out of this session today having a more concrete grasp and definition for what AI is and how it can help each one of your fundraising programs. So, what we can see here is that the definition of AI, or at least as I’ve gone ahead and defined it for purposes of what we’re doing here today, is that AI is the process of using machines that learn and acquire knowledge in order to make decisions.

And there’s a couple of components here that actually are pretty important. The first one is that you can see in all caps, the word process. Artificial intelligence may be a technology, as far as the fact that we are having machines do it, and that there is a lot of software code and other things that have gone into it, but the truth is that the use of artificial intelligence and having this actually work well, it is a process. And we will talk about what that process looks like in just a minute, but what’s important about it is to note, it’s not the kind of thing you use once and then just stop using it. To really use AI and to take advantage of it, there is this continual process of learning that the machine is going to actually be going through throughout your use of it. The next part of the definition here is that it has to be a machine that can learn, and we’ve all heard about machine learning previously, but what we’re talking about here is the type of machine learning that now is being made available through the cloud infrastructure provided by companies like Amazon, Microsoft, and Google.

And what this means is that they’ve already built in a lot of the algorithms that are necessary for us to have machines that have the ability to learn. What comes next though, is that once you have the ability to learn, you then need to make sure that you’re going to acquire and actually learn some particular piece of knowledge or insight.

And once you know how to learn, what is it that you’re going to learn specifically? What kind of information are you going to put through the system? And what do you want it to be a subject matter or domain expert in? And all of this gets towards the third point here, which is it’s about exercising of judgment.

And when I say exercising judgment, what I really mean here is making decisions. And this really fits in line very nicely with our concept of intelligence, even with respect to how it pertains to a human. Meaning, one of the reasons why we all go to school, and we try to acquire as much intelligence as we can is so that we can make better choices, better decisions. And that goal, or that purpose remains the same, whether it is human intelligence or it is machine based or artificial. The goal is still the same. We want to have a system that can learn is going to have certain experiences and get certain information to acquire particular knowledge so that it can make better decisions or better choices.

Now, if we just take a step back and try to understand really what is the difference between a lot of the traditional data analysis, and more modern AI solutions, like the one my business provides, ExactAsk? And so what I’ve done here is I’ve put two very common options that our nonprofit customers frequently refer to when asking us, how does your technology compare to blank?

And the first one tends to be what’s called the RFM model. And I don’t know how many of you are familiar with what RFM stands for, but it’s recency frequency and monetary value. And the idea behind an RFM model is that using those three historical data points, you can actually start to score the quality of your donors in your database and get an understanding or ranking of which donors might be the best ones to solicit as part of a direct mail or other type of fundraising campaign.

One of the things that is limiting about an RFM model, however, is that it’s not very dynamic, it’s static. So you’re going to be performing the same calculation for every donor using the same formulas. And you’re going to simply be plugging in new information or the data points that pertain to a particular donor specifically. What all this means is that you have a formula and an equation and a model that is an overly simplistic view of donor behavior. It is looking backwards, meaning it only accounts for the data points and the behaviors of what the donor did historically. And it is also not providing any assessment as to what might happen in the future or what you should do in order to influence the donor’s future behaviors in a way that can help you better achieve your objectives.

When we look at the wealth screening tool, what we can oftentimes see is people want to know, what will help me understand how much to ask my donors is if I knew how much money each of these prospects or how much money each of my existing donors actually has to give. And believe it or not, what we have found to be the case here is that wealth screenings, that data does not tend to work extremely well when trying to understand or predict or even influence the behavior of a donor in the future. And this is particularly true at the smaller gift sizes. So when you think of donors who may be writing 50, a hundred, or even thousand-dollar checks, what we know to be true at this point in time, is that what makes one person give a hundred dollars and another give 150 bucks has nothing to do with how wealthy those two people actually are.

There are other factors about those donors and their relationship with each one of your groups that is the determining factor in how much cash they’re going to actually give. The other thing about wealth screenings is that data becomes outdated very quickly. Meaning that the information sources about how wealthy a person are, tend to reflect things like annual SEC filings and information that can be gleaned from other reporting sources that tend to not be very fluid, dynamic or really reflect the current state of affairs. So while wealth screenings may be helpful and while RFM can be helpful for scoring our donors and understanding who we may want to solicit at a given moment in time, what both of these things lack is to what they’re going to enable us to do.

Neither one of these traditional analytical tools allows us to really get a sense or understanding of if I take this action, I can influence the outcome that will occur in the future. And that’s really what AI is all about. Again, it’s all about focusing on the decision-making and making sure that the decisions that are being executed are the ones that lead to better outcomes.

The other component about AI that’s critical is just how scalable it actually is. A great example of the scalability of artificial intelligence comes with the personalization of thousands, tens of thousands, even hundreds of thousands or millions of solicitations or direct mail pieces that any one of your groups may be sending as part of a particular campaign.

And when you think about it, how can you create a unique model that would truly encompass each donor’s unique relationship and their behavioral patterns with respect to how they behave with your organization specifically, and do that for every single name in your file? That would be a pretty daunting task for any data analysts or in truth any human. Well for artificial intelligence, that’s actually rather easy. We can actually have our AI tool create separate unique models that describe the behavior as well as the expected future outcomes for every single unique donor in your file. And this means that we can actually have personalization at scale using AI that is truly personalized for each particular person.

That ability is what makes AI such a potent and powerful tool for any fundraising program.

It’s also helpful to understand why artificial intelligence is such a good option now. And one of the things that really is important here is where we are in the life cycle stage with respect to donor management platforms, CRM systems, and other transactional type databases.

And so for at least 5, 10, 15, 20 years now, many nonprofits have been using these types of donor management systems and databases. What that means is that there is now a complete database, basically documenting all of these transactions between your group and each one of these donors specifically.

And the way that we like to think of it is that database is really like having a perfect log of a long running conversation that your organization has been having with each donor specifically. So the ability to have that information at hand and having it be easily accessible and also be down to the individual donor level.

I mean that we can see what each donor specifically engaged in, what they may have responded with, how big their gifts might’ve actually been. That allows us now to all of a sudden say, what if we took all of that donor-specific data, and that long running conversation, and what if we had a system that could now basically step in and effectively have a personalized communication in response on the basis of the data points that we have seen historically, and what happens in this conversation going forward?

So that really is a key aspect. The fact that the CRM and donor management systems have now been in place for a large number of years. One of the things that I would caveat here is a lot of groups are always concerned, what if my data is not very high quality or what if I don’t even know the true quality of the datasets that I currently have?

And what I will tell you is that when it comes to artificial intelligence, that is really not such an important question, depending on the type of technology the AI system can actually use. Now, one of the things I happened to specialize in both at Johns Hopkins and in the company I started here Arjuna, is using a type of AI that actually takes responsibility for having to create the dataset itself.

Meaning even if you were to start with no data, the machine would start to take actions. And there would be this combination of not only exploiting some amount of insight that may exist in an already established dataset, but also this idea of exploring what else might be possible. And the machine is going to see could we push donors a little bit higher? Can I push this type of a donor a little bit further? If I were to ask a little bit less from donor Y, would I be able to actually get a gift instead of not receiving any response, whatsoever? And so that ability to both use the datasets already in your CRM and also have the machine take responsibility for making decisions that will contribute to the growth of the most effective dataset means that all of a sudden, we have a very powerful tool and a combination of technologies that can really start to deliver substantial impact. The other components have to do with cloud computing as well as just how interconnected today’s systems are. So even if you had the ability historically to have an AI machine personalize a communication for every donor in your file, most often there was nothing you could actually do with that information. Meaning that AI system could not actually be connected to the other platforms or other tools or even other partners each one of your organizations may use or work with to actually execute on that mailing. Meaning, it’s not enough to have a machine that can know how much to ask each donor to give, you also need to make sure that, that information and each one of those decisions get sent to the front-line teams where they can actually print the right numbers and print each one of those personalized communications.

Because there’s been so much advancement in the past 10 years, in just how connected these systems are, all of a sudden that problem is no longer an issue. And so the combination of just how much data we now have in existing CRMs, the ability for machines to start to generate data on their own, the ability to do all of this from the cloud, and be able to have these systems be connected with your partners and other software tools means that the entire ecosystem is pretty much in place to now have machines, or artificial intelligence, start to work with your program and deliver the type of impact that Tina Wright was describing they realize it at the Idaho PBS station.

Now this is where we start to get into, I think some of the most important aspects of today’s presentation, which is how do we best use artificial intelligence. And as you may recall from the definition of AI, we have to always start off by understanding it is a continual process. So what you’re about to see is going to be me walking through what one cycle in this process looks like.

But at the end of the day, we want to repeat this cycle as many times as we possibly can. And we want to do that because that is how the machine is going to actually learn So, what we do first is we start off with all these data points here, and that’s what these multi-colored dots are on the left-hand side, and we take that data and we want to actually now make sure from those datasets, our machine is going to actually create or filter that dataset down to what it knows are the most relevant data points. And how does our machine know that? At this stage, we’ve done hundreds of millions of personalized solicitations across many years in many nonprofits.

And that is just information or knowledge that our machine has simply learned over all those, solicitations and all those campaigns for all of our customer accounts. Once you have that data, you then want to put it into your analytics engine. And that’s what we at Arjuna a called the actual AI modeling platform.

What happens in this analytic stage is really important. The first thing that will actually occur is our AI will try to make sense of the data and of what’s happened historically to create a model of what it thinks will happen going forward. And it will do this for every donor in your file. What do I mean by that specifically?

I mean that there won’t be one model that our machine is using for every donor, rather it’s going to allow each model to actually change as most appropriate to truly describe each individual donor’s past behavior and what seems to influence it. So once an analytics engine has these models created, then it’s a matter of actually taking the data points for a particular donor running them through that model.

And that then gives us an output. And that output is going to almost always be a decision. And when I say almost always be, one of the examples I would use here is a common use case for artificial intelligence outside of the non-profit context happens to be autonomously driving cars. We all have heard a lot of news about these trucks that can drive themselves or Tesla’s autonomously driving car features.

The way that I like to explain this aspect is what’s really happening with that AI is that we are teaching a car how it can know when it’s time to apply brakes, when it’s time to hit on the accelerator, when it needs to turn its wheels left and when it needs to turn his wheels right.

Because the machine knows that, it can now make decisions as to when to apply the brakes, and in what amount of force. And those decisions then get executed by the car or by robotics or other components of that machine. At the end of the day, however, the single most important part of the AI was the systems that enable the machine to come to a decision as to what it actually needs to do next. And that is the decision that our AI systems are able to do for non-profits when it comes to their direct mail or direct to donor fundraising campaigns and knowing how much to ask of each particular person. In the final step, we’re going to then get these outcomes. And as part of these outcomes, there will be treatment and control groups so that our system, as well as us as humans looking back on what actually occurred, we’re all able to get a quantitative gauge and measurement of exactly what the impact of the systems decisions actually were. So by having a control group where the ask amounts would continue to be whatever ask amounts you guys would actually calculate for each donor without using artificial intelligence, we’re able to get a picture as to what would have probably occurred had the donors in the treatment group not received AI-based ask amounts as well. And when we can start to see the behavior of the treatment and the control groups start to differ, that tells us that we’re starting to now see the AI be able to influence the outcomes simply by making better choices.

All of that information and those outcomes gets put into this feedback loop. Now in a few minutes, I’m going to show a couple of case examples for nonprofits that we actually work with. And I’m going to go through three steps for each one of those cases. And what I want to point out here is that each one of those steps is basically one complete cycle of this loop here.

So when we start to look at what this actually means from your perspective, here’s a look at this life cycle stage, but with slightly different information here. It’s not just what’s happening analytically, but here’s, what’s happening in your campaigns specifically. And so what’s interesting to note here is this secure data ingestion, and I say that word secure, and put a lot of importance on it. There’s a lot of AI systems out there that will require you to have your data be aggregated with other nonprofit groups. It’s going to take your information, combine it with a lot of other data, and on that basis, it’s going to try to come up with AI that can provide what they’re going to call, more holistic insights and outcomes. One of the things that I’ve come to actually learn is that I don’t know if that’s always really the best case, both from an AI and modeling perspective, but also just from a practical perspective when it comes to the way nonprofits want to run their fundraising programs. We are continually getting a lot of feedback and we can hear just how important data privacy as well as data security is to each one of these groups.

One of the things that we’ve actually done to help accommodate that is we use a type of AI technology that does not require the aggregation or combination of datasets. In fact, we’ve actually seen that when you combine datasets, you may be drowning out some of the unique insights about your relationship with each particular donor.

So if you were to bring in other information about how that donor behaves when they give to other groups, all of a sudden you’re going to lose some of the unique characteristics and components of your relationship with that donor. So really this whole idea of security starts with making sure the datasets are actually split up, never combined, but also that it is using a type of AI where it’s able to come up with insights that are specific to your nonprofit and your relationship with your particular donors. And that’s why you’ll see nonprofit one as well as nonprofit two.

Once that data comes in, we then go through the modeling process. And at the very end of that, you get what’s called the personalized ask amounts or the personalized gift arrays. And that’s where we come right here. So that’s where we see in this type of an AI application where you can receive something concrete and tangible, meaning this is where the decisions are being made.

If you could have a machine go through your file and tell you exactly how much to ask of every donor for each particular campaign you might be sending them over the coming 12 month period, all of a sudden, we can give you a CSV file or we can work with your agency, a group just like Allegiance, and we can give them a file that has in one column, each donor record, and the next column, how much each donor should be asked to actually give, and that can get uploaded directly to your printing file.

We can send it directly to your printer, or when it comes to email and other types of solicitations, we can take that file and put it straight into whatever platform or technology it is that you guys are actually using to manage those electronic campaigns. We then get the outcomes again, and we start this process again, where basically data comes in, we model it, give you personalized gift arrays, outcomes, and we keep doing this process.

Again, one of the common themes that you’re going to hear from me today is how it is a commitment to this process and repeating these iterations over and over again, that leads to those type of fantastic outcomes that Idaho TV was able to achieve.

So I’d like to take a few minutes here to walk through a couple of case studies so you guys can see what the outcomes could actually look like or what the benefits and impact can look like for groups just like yours. And this first case study is for a group called MAP International, and they are a religiously affiliated group that helps to distribute and collect medical supplies.

What you can see here is that we ran through three of those loops. Meaning we ran through the AI process three complete times, starting from when we collected data, modeling occurred, personalized gift arrays went out and that would be day one. And then we give 60 days for the outcomes or for the results to basically pour in.

And at the end of the 60-day period, we close that reporting cycle and then we start the next loop. What you can see here is that in blue, you can see how much money or the revenue that came in from donors who were solicited with ask amounts determined by ExactAsk or artificial intelligence.

And you can see that they start off nearly identical. And then you can see that the treatment group starts to actually produce higher amounts of revenue. And as time goes on and we get from learning cycle one to then we start the process a second time, you can see that all of a sudden the gain starts to increase. Meaning at the separation between the blue and the white lines becomes a little bit larger. By the third one, it becomes even larger. On top of that, you can see that the giving patterns, meaning exactly the timing as to when the shape of these lines and when some of these things like this immediate jump up or steep rise occur, are nearly identical for both the treatment and control group. Meaning that the nature of these donors was very similar, and the patterns that we actually saw between each group was very similar. But simply by knowing how much to ask and changing just that one variable and making a better decision with respect to that single variable, you’re able to start to create substantial improvements in revenue that can really fuel each one of your non-profits into the future.

And while we are showing three cycles here, one of the things to note is that this process continues on continuously. So at the end of the day, once you start using AI, you don’t just stop after three cycles, this process continues for every campaign that you guys decide to actually send. The improvements then become much larger in scale, as well as persistent over time.

And we can see how this can really result in a substantial impact to your bottom line.

If we look at the results campaign by campaign, meaning that this group, they’ve had campaigns for their active donors or renewal file, they use the AI to go after some of their lapsed donors. They even use it to do new donor acquisition. What we did here as we put in the total cost of using a solution like ExactAsk right here. And that’s what’s in the red parentheses here. We then have how much money was brought in by the control group versus the treatment group. And then the two key metrics we use to measure the impact is we look at was there any lift in revenue from the treatment group over the controlled group. And not just was there a lift in revenue, but what was the ROI that you guys were able to actually achieve when you use this type of artificial intelligence. And what you can ultimately end up seeing is that across all these different campaign types, they spent a total of $3,500 and were able to raise an incremental 74,164 bucks.

Obviously that return on investment is substantial, but equally as exciting is just what that raw number is. This is across three campaigns, none of which were, let’s say incredibly large, and we can see that already, you’re able to start to have a very meaningful impact. And with an ROI of nearly 2,200% here, but also the lift in revenue was 18.3%. I also think it’s important to note this was achieved in a matter of months and not years. So when we talk about that learning cycle, really, when you start to understand that you can do five, six more of these learning cycles in a single year, all of a sudden the value you can realize becomes even greater.

One of the things that we also like to do here is, we like to not include any of what we call the outlier gifts in the results that we actually show. So if any donor made a particularly large gift, they were not included in these outcomes here, but we do still provide that data so you can get a sense as to, it’s not just a matter of increasing revenue from that particular campaign, but the AI and the ask amounts, they do also help each organization start to get donors to make those much larger gifts. And you can see that by the fact that the treatment group produced substantially larger amount of revenue from those large gifts than the control group did.

So now I want to look at the same exact slides, but for a second group. And this group is called Partners in Health (PIH). They’re based out of the Boston or New England area. And they do very similar type work to what MAP International does, they distribute medical supplies. And we can see that the outcomes are very similar.

Again, you have the treatment and control group that start off very similar. As time continues to pass, you can see that the Delta between them or the jump from the blue line to where the control group was, the separation becomes larger and larger. After the first run or after that first loop, it’s about a 25% lift.

After the second one, it goes to 51%. After the third one, you’re already up to 62% lift. again, this is after only three learning cycles. So imagine what this can look like when you’re doing it many more times, even in just a single year.

And when we look at what happened campaign by campaign, you can see again, a not huge financial investment being nearly 6,000 bucks here, but they were able to get nearly $23,000 more. It ended up being a lift of 90% as well as an ROI of 383%. All very meaningful numbers here, especially that this is after three campaigns and only three learning cycles.

So, what we like to do is use these results and this understanding of what AI is and what it takes to be successful with it as a framework for starting the discussion with each group and helping them start to understand whether or not AI is the right kind of a tool for your fundraising program at this moment in time.

And really what that question comes down to is: Do you have a CRM? Have you collected data for a year, two years, however many years now? Are you able to commit to using the AI over a series of iterations so that you can actually see what the outcomes are after each learning cycle?

And if you’re able to do those two things, it gets even easier when you start to work with a group like Allegiance. We have a good amount of experience working with these guys, and I can tell you working with them is a true pleasure. And they tend to make it very easy and almost effortless on the part of both us and you guys to get all this stuff done.

So in truth, many of our customers don’t even have to do any additional work. It’s just something that between us and Allegiance, we’re able to provide and incorporate into everyone’s campaigns and solicitations. The PIH and the MAP International case studies obviously have great numbers.

But what I think is really important is the idea that these two groups were chosen because they actually reflect a very typical outcome that we see for groups of that size. So between what you guys saw Tina Wright and the Idaho TV station discussing, and now these two other groups, you start to get a sense that these outcomes, they’re not unique, they’re not that rare. This is the type of lift and the type of ROI that AI is capable of delivering for each and every group.

So again, what are the best practices for how to use this type of a technology? It comes down with, I’d say first and foremost, using a provider like Allegiance and Arjuna that can work together in order to make this the most successful endeavor for your group.

It really allows us to streamline operations. We can ensure everything works and connects with your fundraising strategy, your fundraising program, and the other fundraising tools that your team uses. And the last thing is having a strong commitment to those iterations, to those cycles, to that process.

I’m just going to finish up here with some of the comments we get on a regular basis as to why groups do choose to go with an AI solution. And in this case, I can tell you what we hear when customers tell us why they’re going with our solution specifically. And what I can say here is that right now for every dollar a nonprofit spends on our AI tool, they are receiving $3.95 back.

That obviously is a substantial ROI. In truth, I’m not sure if there’s a lot of other things you could do in your program that could deliver that type of an impact simply by changing one variable or one number, the ask amount. We can see that the average lift in revenue is about 18%. Even within that first year, we frequently are told how it’s amazing that lift is able to occur while maintaining response rates equal between the treatment and the control groups.

Meaning that it’s not as if you’re able to raise more revenue by getting much larger gifts from a handful of people, but you will have lost a lot of donors who you might’ve turned off because you simply asked for too high an amount, but you’re able to actually use the AI to identify which people can give more, ask them, and then for the donors who do not appear ready to give a larger amount, you can still ask a smaller sum. And that means that you can keep participation or response rates completely level while still getting that 18% lift in revenue.

So, if I were to summarize everything here, it would be that look, AI is a process, but it is a process that is highly scalable. It’s forward-looking in the sense that it focuses on taking the data and learning actual knowledge, and then be able to apply that knowledge, to make better decisions that will impact future outcomes.

And it also removes the limit on how much value we can truly create. And it does that because of just how scalable the system is. One of the things I like to frequently say is if we could take a fundraising director and clone them a million times, so that way there was only one fundraising director managing each donor relationship, no matter how big or small the gifts were.

Under that scenario. Yeah. We would expect to actually see the outcomes improve. But that’s obviously not possible. We can’t hire enough people to actually do that type of mass personalization. Artificial intelligence can. A machine can. And that really is what’s key because that means that all of a sudden how much additional value can be created has no ceiling on it. You can always be getting donors to give more, to be growing, to be participating in the support of your group for longer periods of time. And you’re able to influence the outcomes that will happen in the future, by what you might do today. And so the combination of these three things really sets everybody up for potentially using an artificial intelligence tool to generate the same type of outcomes we just discussed.

So on that note, I want to thank everybody for their time, and I want to make sure that we provide enough time for everybody to ask questions and just to see what’s on everybody’s mind. So I will turn it over and see what questions may actually exist here.

Brett: I do have a question. So I am curious in the MAP case study, does that show all the segments they had mailed in that campaign? So that was a combination of their active, lapsed and acquisition donors?

Adam: So what we can see here is that they actually ran different campaigns. They did do their active or renewal campaigns. They did a separate lapsed one and a separate acquisition one. What you’re seeing here is that they initially tested the AI in a smaller portion of their overall campaigns.

But when they saw these outcomes, obviously they increased their volume across the entire campaign size.

Brett: Yeah, I was curious on the graph before this, if that was the aggregated data. I think what we’re seeing is that it was stage one was maybe just their actives, and then they add in the other segments as they saw the success over time.

Adam: Exactly. And this graph here does show the overall, meaning this graph is looking at across all of the campaigns being done. And here we’re showing you how that gets broken out by each particular campaign type. So you can see both the aggregate impact as well as the campaign-specific impact.

Brett: So other questions from the audience?

Adam: I think they just started to see a couple come in on the question box. So the question is, how does this technology work on acquisition lists when there’s no prior donor behavior? And I would say that there’s really two ways it works, both of which are extremely important for producing the long-term type of benefits here. The first is when you rent an acquisition list, oftentimes they will give you a handful of other data points about those people or about those names. What we’re able to then do is start to look at what donors in your system currently actually may appear to be similar to each one of those names on that acquisition list. And we can use that as a very well-informed starting point. What we expect to happen once you do that first solicitation, and we see the way each one of these donors respond and what their behavior is going forward, the machine is then able to take that information and really start to hone its understanding of each particular donor’s behavior far more substantially going forward. So really the ability to take some of your existing donors and try to identify similar ones on that acquisition list allows us to not start from scratch, but really a lot of that insight comes from what happens after that initial gift. So we can get a much earlier and faster read on each donor’s behavior, what drives it and can now really increase the value of each one of those newly acquired donors going forward.

Other questions? All right, I don’t see much else coming in.

Brett: Great. Thank you, Adam. AI is certainly an intriguing topic and holds great potential for nonprofits to raise more money and help them grow and support their missions. So we really appreciate you sharing your insight with us today.

And thank you everybody for joining us. We hope you enjoyed the presentation as much as we enjoyed bringing it to you. If you have any follow-up questions, please feel free to reach out. And we hope you have a great afternoon.

Adam: Thank you, guys.