Can Machine Learning Thaw Out The Returns Of Cold Calling In Fundraising?

How many cold calls to prospects have you had to make during your fundraising career? How many do you still have to make? Do you still find a reason to not make them, letting them fall off your to do list despite your best intentions?

When you make the call are you super prepared should they pick up the phone or do you just wing it on some loose research, past engagement and knowledge of your organizations’ needs?

Look, we are all friends here. I know it’s a little bit of everything, because the reality is that cold calling is rough. It’s awkward. It can be a nightmare for those going in under prepared and compounded even more so if you are new to the fundraising game and have simply been handed a call list by a delegation happy Director of Development saying “Can you go through this list in the next 48 hours?”

A harrowing part of structuring out this piece was trawling through countless cold call statistics. I’ll pick 5 to make my point and move swiftly along…

  • 87% of prospects say salespeople don’t understand their needs. (TOPO)
  • 63% of sales representatives identify cold calls as the worst part of the job, according to sales agent stats. (No More Cold Calling)
  • 58% of prospects say they currently find cold calls useless. (RAIN Group)
  • 60% of cold calls go to voicemail. (XANT)
  • A reasonable cold call success rate is 1-2%, according to sales statistics on cold calling. This might lead to the conclusion that cold calling is dead, but the truth is that this technique is still valid in an overall marketing strategy. 

So if it’s still a valid strategy and still sneaks itself into fundraising plans the world over, how do we find the right language, make it authentic, and make it more effective? How do we build confidence in our frontline fundraisers and understand that emailing is not the substitute here, it is just throwing our requests to meet with potential donors into the abyss where everyone is a/b testing subject headings to break through the clutter. Here’s one more stat for you. More than 80% of cold emails stay unopened forever (Campaign Monitor). 

It’s enough to bring a tear to the most optimistic of gift officer’s eye.

Last year, in one of the last in person events I went to, I saw Babylon Health CEO, Dr. Ali Parsa talk about computational health and in particular how artificial intelligence is revolutionizing the health sector. Dr. Parsa shared a video of how the AI of Babylon Health can help detect your disease. I have watched it countless times since thinking about it’s application across a number of sectors. Take a quick look at the embedded video below and rejoin me on the other side…I have an idea.

Welcome back…cool platform, right? So where are we trending here? It’s not like a donor is going to call up your organization, press ‘7’ to talk to an automated bot and then systematically fall on a gift based on a predictive model of what you might be passionate about. This isn’t about organizations bombarding donors with robocalls either.

Donor conversations are always going to require building trust, active listening and authenticity. However that’s not to say you can’t weave in some automated support.

There are some great companies out there for campaigns and advocacy that build out scripts and forward them via email, calls, letters to the editor etc. Many with just one click, but many with an accompanying ‘coldness’ of its own in the form of its delivery.

A great blog post that talks about the need for authenticity when working with fundraising scripts (albeit from the angle of political fundraising) comes from Calltime.ai who state that “fundraising should be relational, rather than transactional. This is why a conversation guide is far more preferable than a script. Most of us can tell when you’re reading off a script, and it instantly feels impersonal and disconnected. Those are not the feelings you want someone to have before asking them to give you money. It’s important to stay present in the conversation, really listening to the person on the other side of the phone so you can offer a genuine response and let the conversation have a natural flow to it. A script often gets in the way of this.”

What if there were platforms that existed that were in essence fundraising scripts, but instead of being static, had more of a conversational tone, could still help move forward qualification to cultivation and eventually to a gift. What if this platform tailored the ebbs and flows of donor meetings in a more intentional and relational way and at the same time ascertain with higher degrees of certainty what that prospect might be interested in. 

I could see a platform that syncs with your CRM that builds individual scripts in real time, pulling previous data points in a dynamic way. Imaging talking to a recurring donor who has been giving unrestricted gifts for a number of years, and during a call suggesting possible named scholarships that might be of interest and then scouring college databases to see what matching gifts are available, ultimately presenting proposals with the donor math calculated and spelled out in easily understandable terms.

Imagine having a conversation with a donor who shares information about their spouse of which you didn’t have, sharing that they do their philanthropy through her IRA. The system could then in real time identify her biographical data, see that she is over 70 & ½ and that the IRA Charitable Rollover exists. Within the next 5 minutes you might be able to walk them through a Charitable Gift Annuity, the benefits and what returns they might receive across a sliding gift scale.

Couple this high touch, highly personalized approach with automatic thank you emails and text messages, syncing back to your CRM with recommended follow up that will ping you when an algorithmic model has pinged you to do so, then you have all the hallmarks of what could be deemed fundraising optimization.

But let’s not get ahead of ourselves here. This would be restricted to calls and emails and not in person meetings. It could however provide a briefing pack that would recommend certain points you might want to cover.

The tech to make this happen exists and could be optimized by uploading countless scripts from nonprofits nationwide, together with conversion rates across a range of different qualitative and quantitative data points. The numbers to make this effective certainly exist with about 2,610,000 results being shown via a simple Google search for fundraising script templates.

Making this work would be a mix of text mining and analytics to derive qualitative insights from unstructured text and identifying patterns or trends from it. What words work, what words evoke stronger levels of empathy, what words move a prospect closer to a gift and which words whether through a direct ask or proposal converts a prospect to an actual donor.

Building out ‘conversation guides’ using predictive scripts will never be perfect but with a mix of both reinforcement and supervised learning as an ongoing process (and commitment) we will make the modelling stronger and help build out more effective scripts across a wide range of disciplines not just ‘hey here is a good generic fundraising script I wanted to share’.

Supervised learning (SL) is the machine learning task of learning a function that maps an input to an output based on example input-output pairs for example if they say this, then this desired outcome is the more likely next step. It can also map out new examples.

Reinforcement learning (RL) is an area of machine learning concerned with how intelligent agents ought to take actions. This means rewarding the machine for the ‘correct’ answer. 

The main difference here is around exploration (suggesting patterns) and exploitation (widely understood & accepted knowledge).

You won’t have to worry too much about what is happening in the background as I believe most of the AI modelling that is used commercially in the social sector will largely have explainable functions. For example predictive scripts will show the rationale for choosing those words and why in that order. The models will also update daily at a minimum which will mean stronger lead generation and identification.

My only concern is that predictive text might see a renaissance in headsets to ensure hands free calling. So if you are happy to look ridiculous on the other end of the phone we might be able to confine cold calling to being a simple descriptor and not a critical assumption of an end result.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s