Wednesday, August 29, 2018

A Self-Fulfilling Prophecy

Even though, I’ve been in prospect development for more than 25 years, I continue to learn new things every day.  It amazes me.  Just when I think I have things figured out, this field can humble me.  Often, when I think I’ve found the right way to do something, something else comes along to challenge that best practice and show me another, better, more strategic approach.

To be honest with you, sometimes that just ticks me off.

I don’t mean that in a bad way (if that’s even possible).  That bit of an angry feeling comes more in the form of “Oh crap… I wish I had known that sooner.” It also often leads to “Great, now I have to learn that thing.”

That bit of anger usually fades away and is replaced by a genuine curiosity about how to actually get better, be more strategic, and focused.  That doesn’t necessarily happen overnight.  You see, it’s not my nature to be an early adopter, but when something new comes along, I definitely pay attention.

Unfortunately, while I’m paying attention – that new thing usually flies past me and I have to play catch-up.

That’s how I feel right now about Artificial Intelligence (AI).  I don’t completely understand it, but I see it coming like a title-wave.  Some people like the folks at City of Hope (Nathan Chappell and Nathan Fay) are riding that wave, hanging ten on their surfboards and waving at the people on the beach, who are wishing they were on that wave with the two Nathan’s.

Nathan Fay and I sat down to lunch several weeks back and I could see how committed he was to AI and I walked away having no doubt that he was going to help City of Hope raise more money.  I was envious.  Not jealous, because I love City of Hope – just envious that he had the support and resources to move forward.  Good for him.  Good for City of Hope.

AI is overwhelming to me. It’s huge. It’s mega-big. It’s a game-changer. 

I’m all about finding the story behind a donor’s gift and their connection to my organization and its mission.  It takes a lot of work to find that story and then turn around and tell it to the right people.

And when I say a lot of work, I mean it.  When I was at the Pancreatic Cancer Action Network (PanCAN), I came to realize I could spend every moment of every working day – finding and telling those stories.  Every moment.  Every day.

It’s a daunting task when you’re looking at a database of more than a million records.  Heck, it’s even a daunting task when you’re looking at a few hundred thousand records.  If AI can look at all that data and then go beyond to learn the story, I want in on that (By the way, it can).
Are you kidding me?

I don’t know about you, but I often feel a sense of urgency in the work I do.  I especially felt that working at PanCAN and City of Hope because quite frankly, people were dying every day.  Meeting volunteers, who lost someone to cancer will do things to you.  Being someone who has lost family to cancer does things to you.  In my case, it motivates me to do better.  Be faster.  More strategic. 

I have a strong work ethic, but if I can work smarter – I want that.  I know if I can combine the two, I’ll be a force to be reckoned with.  I’m just one person, but I can make a difference.  We all can.

Why am I writing all of this?

I recently listened to David Lawson’s podcast where he explains "How Big Data can translate into Big Good."

As always – David makes me think.  He has this way of presenting new ideas that really piques my interest.  Mind you, my interest was already piqued by conversations with Nathan Fay, but David’s podcast hit me in the face like a cold wave from the Pacific Ocean.

David addressed the bias we have in the work we do and the bias that often exists in our data.  He told the story of a visit with an Ivy League school a number of years ago where he was trying to convince someone that younger donors could have a major impact.  The response was “Our major donor’s average age is 72 years old.”  David replied with “As long as that is who you’re going to focus on, that is going to be true.”

Bam.  A self-fulfilling prophecy.

In my last blog post, I wrote about the predictive modeling results we just implemented.  I mentioned that I saw this trend where our highest scores were often represented by constituents who were alums, parents and either faculty or staff.  Those that had all three attributes “looked” like our best prospects.

I acknowledged that this made sense since they were the most engaged. It was our own self-fulfilling prophecy. I knew it was biased.  I also realized that if we continued to focus on these constituents, our model wouldn’t change.

David’s podcast reinforced the idea that we often do things that become self-fulfilling prophecies.

It’s one of the reasons I set aside those prospects/donors who had all three attributes and began my implementation of the results by looking at other alums who had high scores – knowing we didn’t have the best engagement with our existing alums. 

I wanted to change the focus of our major gift team (who has a history of focusing on new parents) and help them focus on a group of people that would have a more long term impact on our fundraising efforts.

As I listened to David’s podcast, my head began to spin.

I began to think about the fact that even though there is value in predictive modeling, there is even greater value in utilizing big data and more specifically, AI to do even more. 

I know there is bias in modeling. That's not necessarily a bad thing, if you know that in advance. It has taken me some time to realize that.  You see, it’s often a bit of a process for me to figure out what that bias actually is.  AI helps us avoid some of that bias, if not all of it.  At least, I think it does. 

At PanCAN – we took the modeling results and really analyzed it to the point where we started to develop our own model of who our best prospects were.  When I say “we” – I really mean our data analyst at the time – Victoria Merlo and I worked together.  She did the heavy analysis and really led the effort. 

We found very specific characteristics/data points in our donor records that really helped us focus on a specific group of people to target for major gift cultivation.  The modeling scores were a part of the profile, but now we were armed with additional data to help us be more strategic.

Like I said earlier, I’m always learning.  Every time I take on a project like a wealth screening or predictive modeling, I become more aware and more informed.  My goal is to find what is truly predictive.

That’s the thing.  That’s the secret sauce.

How can we know who is really most likely to give and how do we engage them?  The answers appear to be in AI through machine learning.

AI can learn things faster than I can.  It operates without bias and when operating in that realm, the data is never going to lie.  This is real science.  This is game changing.

I want in.

Right now, I have no choice but to do the work that needs to be done manually at times. It’s just my reality.  I will continue to move forward; armed with the self-awareness, that there is bias in what I do and that it’s not going to be perfect.  It can still be effective, but it’s far from perfect. My new objective isn't to just follow a predictive model - my goal is to change the model in a way that will make us more effective.  It's an idea that Lawrence Henze of Target Analytics actually put in my head when we took delivery of our predictive modeling results.

Models change.  We can impact that change.  Lawrence has helped me realize that.

I will continue to move us forward and continue to advocate for the use of AI.  I know I may only be able to take baby steps; when in reality, I want to sprint forward and dive into the wave where I can join the Nathan’s of the world.  
I’m trying to do that, by introducing companies like Gravyty to our organization.  Unfortunately, we’re not using their product just yet.  Fortunately, by introducing them to our team, we have opened the door for a conversation about AI.  It’s a start and I am hopeful. 

I know there may be resistance, but I can be like a dog on a bone when I think something is important. In the meantime, we are all doing the best we can with the resources we have.

It’s all about the big picture.  Non-profits want to change the world and make it a better place.  Those of us in prospect development can help drive that effort. If I’m going to have a self-fulfilling prophecy; let it be this.

Sunday, August 26, 2018

A Practical Approach to Implementing the Results of a Predictive Model

I can be such a geek.  I love evaluating and constructing major gift portfolios.  I love everything about the process; especially, when I have the data that allows me to lay a foundation for everything I do as it relates to assigning, un-assigning, reassigning and evaluating a portfolio of donors and perspective donors for a major gift officer.

I want my organization’s major gift officers to succeed.  I really want them to excel.  I want them to rocket past their goals and have a monumental impact on the fundraising efforts of the organization we serve.  I believe by working together, we can make a difference.  I’m committed to this partnership and I have high expectations for success.

It all starts with each major gift officer’s portfolio. 

It’s all about having the right constituents in their sights as they go through the development process with each individual.  I believe it is my responsibility to drive this effort.  It is my mission to do this well and to do so in partnership with my major gift officers.  I take great pride in this work and I’m determined for us to succeed together.  We absolutely have to do this together.

I am about to tell you a story.  It’s about how we attacked the process of constructing the best possible portfolios for our team of fundraisers.  It’s only the beginning of the story, but it is the first step in a journey of discovery and collaboration. 

We recently took delivery of our results from Blackbaud Target Analytics’ predictive modeling services. It is a product/tool I’m intimately familiar with.  I have partnered with Target Analytics at four organizations. I’ve learned a lot about predictive modeling along the way. My history with all of this includes mistakes made and lessons learned.  Every time I go through this process, I learn something new.  Every project is unique to each institution. 

Here’s the story…

For many years – those of us in prospect development primarily relied on wealth screening to help us uncover and discover who our best prospects for major gifts might be.  We essentially created lists of wealthy individuals and assigned them accordingly.  The process was flawed for a variety of reasons; most of all, because the process is void of addressing a constituent’s likeliness to give and/or their affinity for the mission. 

I won’t belabor the point.  Let’s just say that modeling makes our efforts more productive.  It allows us to be more strategic and a little more precise about the process of identifying the best possible prospects for our major gift efforts.

Our team embraced the idea of utilizing modeling as a means to make their portfolios more strategic.  They loved the idea that we could utilize a “target gift range” that was specific to our institution and couple that with a “likelihood to give” score.  That “buy in” was critical to our success.    
Our team had a history of looking at one’s overall giving capacity.  A number derived from a percentage of a person’s confirmed assets.  Even though some of the ranges were large – they were not specific to what was necessarily possible or even predictable.  Using predictive models changes all of that.

As we walked through the process – the team continued to embrace the modeling scores and see the strategic advantage it provided.  I’ll explain this more as we go along.

Once we took delivery – we imported the scores into our Raiser’s Edge (RE) database. From there we were able to run queries and export data into worksheets.

Once the data was in, I took a look at each fundraiser’s portfolio in conjunction with the scores of each of their constituents.  I then made recommendations.

Individuals with high major gift likelihood (MGL) scores (700 or higher on a scale from 1-1000) and target gift ranges (TGR) greater than or equal $5,001-$10,000 (single gift) were highlighted.  Constituents who had yet to make a gift in their target gift range were recommended for an upgrade.  Those already giving within their target gift range or higher were simply noted and recommended for further cultivation and solicitation.

In some cases – I looked at their confirmed assets (identified through Target Analytics Research Point).  If significant, it was noted.  I also looked at their philanthropy – if significant, it was also noted.  These two things didn’t drive our efforts.  They simply served as side notes.  More on that later.

Those with lower TGRs, but were giving above their TGR were set aside for evaluation and discussion with the fundraiser.  Example:  TGR was set at $501-$1,000; however, donor was giving at $5,000 level. Individuals like that deserved a closer look.  Those with low TGR and low giving were recommended for un-assignment; however, they were also set aside for discussion. 

Armed with this data – I sat down with each fundraiser and their supervisor and began to go through their portfolio name by name.  

In many cases the higher scores were reaffirming.  In some cases – the scores came as a pleasant surprise and the fundraiser welcomed the idea of working to upgrade the donor’s giving.

As we looked at donors who were already giving above their TGR’s – there wasn’t any angst in the fact that their total giving didn’t match their scores – even if they were giving well above their TGR.  We simply acknowledged the scores represented a very specific view at a person’s potential and not the only view. 

Our modeling scores are based on how much a donor looks like other major gift donors who are already giving at a major gift level to our organization (that’s a simplified definition).  If someone doesn’t look like our typical donor – but gives at that higher level, we considered it a good thing.  We didn’t lose any confidence in the rating. We simply noted it and moved on.

This was critical. 

I have worked with some fundraisers who focused on what “wasn’t there.”  In other words; some assume if a score and a person’s actual giving didn’t match for someone they knew – they assumed every score must be flawed.  That’s simply not true.  There are lots of variables that could impact a score – but it isn’t worth the time or resources to figure out why.  Every model has variables. It’s okay.

Next, we looked at those recommended for un-assignment.  In each case, if the fundraiser still believed the individual was a major gift prospect, we kept them in their portfolio.  I emphasize that the process looked at their donors through a very specific lens and I value their opinion more than the score, if the two didn’t match up. 

I trust fundraisers to make the right call and I believe they should be allowed to trust their instincts. Even if their prospect ends up not being a major gift donor – it doesn’t deter me from giving them the benefit of the doubt.  We all learn from the process.

In most cases – the fundraiser agreed to have prospects with low scores un-assigned.  The scores often affirmed their beliefs.  In some cases – it gave them the “freedom” to un-assign some of their constituents. 

The freedom and reassurance in being able to un-assign individuals was key in helping them embrace the process.  In some cases, they literally looked to me for reassurance.  In each case – I told them, “If this person wasn’t assigned and I was evaluating their potential, I wouldn’t assign them to you.”  It worked. The fundraisers were comfortable with the decision to un-assign.

There were some cases where the fundraiser decided one more effort was appropriate to try and upgrade a donor’s giving.  They agreed that if the effort was unsuccessful, they would have the individual un-assigned.  Each individual constituent was evaluated separately. We remained flexible at all times.

It’s interesting to note that some individuals with low scores actually had high confirmed assets.  That didn’t deter us from un-assigning them.  It may have caused us to pause in some cases – but we still un-assigned them in most instances.

I really loved that. 

Just because someone has wealth and the capacity to give, doesn’t mean they will.  Even if they made large gifts elsewhere, it still didn’t mean they would also give to us.  A low MGL scores gave us the confidence to not pursue the relationship for a larger gift.

That’s an incredibly valuable and important point.

Think about all the prospects you may have assigned in the past based solely on their wealth.  Think about the time, effort and resources it may have taken to qualify them further.  Think about how many of them ended up not being major gift donors.  Now, think of the savings in using a predictive model to help segment your database for a more strategic approach.

When all was said and done, we were able to pare down portfolios.  In some cases, we un-assigned as many as 40 or more constituents from a single portfolio.  That effort; in and of itself, was valuable.  We saved time and resources by removing a great many constituents, who aren’t likely to become major gift donors.

We didn’t judge the prior portfolios.  We didn’t dwell on the people being un-assigned.  We just made changes and moved on.  We focused on where we were going.

Now that we had smaller portfolios – we were ready to take the next step and make way for new assignments.

My next step was to look at our results for our un-assigned prospects.  I started with our alumni.  Like many universities, our alumni engagement could be better.  More specifically, our alumni giving could be higher.  We have been heavily depended on new parents for our fundraising efforts and one of the goals I set when I arrived was to identify more alumni as potential donors.

Our modeling project allowed us to pivot and focus on our alums.  We are still going to engage parents and other constituents, but beefing up our efforts with alumni is going to be critical to our long-term success.

As I analyzed our results, it became clear to me that a very specific segment of our database was scoring the highest.  Of our alumni, those who scored the highest also happened to be parents and either faculty or staff.  They represented three constituent types, all rolled into one.

This made sense.  Who were our most engaged alumni?  It’s those who not only went to school here, but also worked here and sent their kids here.  It made total sense that they would have the highest modeling scores.

That being said – we decided to set this group aside.  They would be worthy of a separate campaign and effort (Perhaps a topic for a future blog post).

My immediate goal was to find other alumni – who only had the constituent type of “alum,” but had scores indicating they could be great major gift prospects.  I looked at two groups – those previously assigned and those who had never been assigned.  I also factored in their giving – and if they had recent giving, they became a higher priority for assignment.

I always operate with the foundational belief that our best prospects are the people who are already giving to our organization.  That’s why giving is always a factor.  It’s not the only factor, but it is important.  The exception within higher education is new parents.  In a hospital setting, the exception would be new patients or the family of new patients.  Those constituencies obviously don’t have a giving history – but because of their immediate connection to our organization(s) make them viable.

If an alum had been previously assigned – I checked our database to see if it made sense to reassign them this time around.  If I couldn’t tell – I went ahead and recommended assignment. Our executive team also reviewed my recommendations.  They did this quickly.

Together, we set parameters for each fundraiser’s assignment.  We based assignments on geography, their giving, their backgrounds, etc.  This gave me the freedom to make recommendations on my own and assignments were made in a short period of time. We didn’t have long meetings to go over each and every name.

It’s important to note our university’s development team is unique in that our officers are not based in specific schools or departments. Our officers are physically located in one office building.  They are part of a central team.  We believe in open cultivation and this allows us to truly be donor centric.  Officers can have alumni from any of our schools within their portfolio. 

It’s really a beautiful thing (perhaps yet another topic for a future blog post). Back to the process…

It didn’t take long.  It’s amazing how fast you can move when you have great collaboration.
We are now ready to move forward.  Our fundraisers’ portfolios have a higher percentage of alumni than previously.  More importantly, each portfolios is now more strategic and full of potential.

Now the work really begins.  The fundraisers will be doing their part to engage constituents and the prospect development team will be doing our part to ensure we have a pipeline of prospects at the ready.

This is also where the fun activates and the magic starts to take shape.  I’m excited. We’re all excited.

A final few words of advice…

Trust the models.  Don’t focus on assets. Don’t spend hours and hours confirming assets.  It’s not worth the time when you’re simply constructing portfolios and making assignments.  You can confirm assets later. Do that once your fundraiser has engaged with the prospect, qualified them as major gift prospects and begin to move them towards an ask.

I believe this is where many organizations get bogged down in the process. Don’t do it. That process can take weeks even months.  You can (and should) confirm assets when the donor is further along in the cultivation process and as part of a more comprehensive profile, but not at this point.

Speaking of which…

You don’t need to take the time to develop comprehensive bios for each name either.  It’s just not necessary.  There is a time when this may be appropriate – but most definitely not at this stage.  I thought about typing this paragraph in all CAPS to emphasize the point, but I think you get the picture, right?

You can put a lot of time and resources into developing a profile for someone your major gift officer may never get an appointment with.  They don’t need a comprehensive profile to get a meeting. 

There are always exceptions to every rule – even this one; but again, as a general rule and as a foundation for your success – don’t go down this road.

Besides, great fundraisers don’t need a lot of information going into a meeting.  They’ll learn what they need to know organically.

With modeling, you have enough information to get your fundraisers on the phone and out the door.  Be efficient with your time and theirs.  Focus on the volume of names required for success.  Your fundraisers will need to make a lot of calls to secure a proper number of meetings.  From there – they will find an even smaller group who actually become donors. It is a monumental effort and the process I described allows you to keep pace with the effort.

Again, get your fundraisers out the door.  They’re not going to raise money sitting in their office. 

Modeling improves process.  It expedites it.  Just keep it simple.

This is all very doable; even if you are a one-person shop.  You just need the data. If you can produce a simple query and navigate an excel spreadsheet, you can accomplish exactly what you need to do. If you have other tools to help you visualize the data, even better.

Again, keep it simple.  You can do this in a practical and achievable way.