The Model That Knows When To Shut Up
Most fundraising models rely on a simple set of assumptions:
-
Recency means readiness
-
Frequency means loyalty
-
Bigger gifts mean bigger love
And that all this adds up to a reliable way to pick who gets your next fundraising appeal. But here’s what most models don’t ask: What happens when you actually market to someone? Did they give because of that last email — or despite it?
The modeling every organization needs doesn’t just track donation behavior, it tracks what donors received so it can learn which donors respond to heavier contact, which prefer light touches, and which ones quietly stop giving when the volume gets turned up.
And crucially, it predicts how likely someone is to give in a particular month, based on when and how they’ve been contacted in the past.
That makes possible something most models can’t do: A donor-specific pulsing calendar. Here’s what that looks like in its simplified form with individual donors rolled up into groups for ease of visualization.
Green = solicit this month. Gray = hold off. Each row is a donor type with its own optimal cadence.
Some donors are classic year-end givers. Others are mid-year responders. Some give like clockwork once a year (but not all in December), these are anniversary donors, each with their own preferred giving month. Our model recognizes those patterns and builds a custom cadence around them.
This isn’t about who gave recently. It’s about who’s likely to give if we ask now — and who’s better off left alone.
Most “personalization” in fundraising is surface-level. Change the salutation. Swap the photo. Vary the copy block. But the cadence stays the same.
Real personalization means:
-
Knowing who responds to outreach
-
Knowing who gives after silence
-
Knowing who needs a nudge
-
And knowing who needs a break
That’s personalized pulsing — customizing the rhythm, not just the message. Is your agency doing this for you? Here’s your five-point gut check:
1. Does it include your marketing history — not just giving?
If the model only looks at who gave and when, and ignores how often you’ve asked, you’re modeling outcomes, not behavior.
2. Does it learn how each donor responds to being asked?
If it treats marketing as background noise or assumes everyone reacts the same way, you’re not capturing real-world dynamics.
3. Does it model irritation and memory?
If the model doesn’t learn that too much contact can backfire or that a well-timed appeal can have a delayed payoff it’s missing the behavioral nuance that drives real response.
4. Does it predict when not to ask?
A good model isn’t just about green-lighting solicitations, it should know when to pause to avoid burnout or tune-out.
5. Does it personalize cadence not just content?
Changing the message to match the person is mission critical. But if every donor is on the same calendar, you’re still treating them like a segment, not a person.
This approach, whether built internally or with a partner, respects your donors, saves you money, and gets better results. Because sometimes the smartest thing you can do isn’t ask more, it’s know when to shut up.
Kevin
4 responses to “The Model That Knows When To Shut Up”
Leave a Reply
Leave a Reply
Behavioral Science Q & A
Whether “help” is more engaging or not really depends on the framing and context. The word help can sometimes weaken the perceived agency of the supporter, making their role feel secondary rather than central (your point). On the other hand, help can also signal collaboration rather than implying full ownership of the outcome, which might […]
Read Full Answer
Great question! Here’s how behavioral science can help unpack what might be happening: Pain of Paying: Even a small extra charge can make giving feel more transactional than emotional, potentially reducing generosity. Fairness Concerns: Some donors might perceive donor cover as a surcharge rather than a contribution to the cause. If they feel the charity […]
Read Full Answer
The choice between “Your gift CAN…” and “Your gift WILL…” taps into the psychological framing of certainty vs. possibility. Currently, there is no academic research directly comparing these two framings in charitable appeals. However, I suspect no framing is universally better—the outcome likely depends on your target audience and the campaign’s goal. Here are some thoughts: Certainty Framing – […]
Read Full Answer
Integrating an individual giving appeal with other communications from a charity can have both positive and negative effects, and the outcome largely depends on how it’s executed. Advantages of Integration Brand Consistency: Maintaining a consistent appearance and messaging across all communications can reinforce the org’s brand identity and strengthen brand recognition and trust among your […]
Read Full Answer
I’m not aware of any in-market tests specifically comparing recurring vs. gift frequency language. I suspect the answer might not be the same with all gift frequencies, nor with all people. It sounds like a great opportunity for you to test and find out what works for your audience. Based on the literature, here’s a couple […]
Read Full Answer
Based on what we know from existing data, those renewal notices can actually be pretty effective in getting people to donate. They tap into our psychology – creating a sense of urgency, reminding us of past support, and using personalization to make the message hit home. They’re playing on our natural tendencies to feel obligated […]
Read Full Answer
THIS: does it include your marketing history — not just giving? AMEN!!!
Hi Lisa, yes, agreed – this is the key point and fatal flaw with 99% of models I’ve seen, including those from big agencies. They only use transaction data. This means, by definition, they are comparing Jack Donor to Jill Donor. It also means, by definition, that if Jill is a better donor than Jack on average, she’ll get selected for solicitation every time. And we will turn Jill into a “bad” donor as a consequence.
Why should we force Jill to make up for Jack’s donor deficiencies? I want to make decisions about Jill based only on her behavior in reaction to her promotion history – i.e. the behavior of the charity.
hi Kevin, in theory this all sounds great. Yes, most nonprofits’ databases do keep the donor’s giving history and appeal history (by channel) so that information is available. But sometimes that’s where it ends.
When you talk about a model, which model are you referring to and where is it available? There are so many modeling companies now that it would be helpful to be more specific. As far as I know, several companies are using both appeal history and giving history.
Also, I’ve seen this time and time again, what donors say and what they do are two very different things and the same apples to models.
So model with care and test and reverse test but most importantly, determine what you’re trying to test and why and determine what you want the model to do.
I’d be extremely curious about a model that can measure the donor’s irritation of getting yet another label package. How can you measure something like that just because the donor does not respond? Is there a camera built into the package? (yikes, scary)!
Also, determine what’s the goal of the model. Is it to save money by not mailing those who will never make another gift? Is it to raise more money?Is it to reactivate lapsed donors? Is it to find new prospects? Is it to find monthly donors within your donor base? I’ve seen hits and misses with models and always recommend testing a model versus your fundraisers gut and/or experience and see what happens. Models do cost money so make sure you use them where it makes the most sense.
There’s a lot here but let’s get past the generic platitudes.
This model is real, built by DonorVoice, and applied by us, every day, for our agency-of-record clients. It’s not a sales pitch however, this is only available from us for our AOR clients. It was written to note two things,
1) highlight how desperately the sector needs to replace what passes for modeling today. Because in most cases, it’s still just RFM with lipstick.
2) that this model is ‘stealable’, repeatable. There is nothing proprietary here, the only thing required is mindset shift and technical chops, latter can be rented, former is biggest barrier.
This model, using transaction and promotion history, accounts for concepts like adstock (positive memory) and irritation (negative response to overexposure). Those aren’t buzzwords. They’re functional parameters in a time-series framework. They’re not theoretical. They’re built to predict and change donor behavior, not simply describe it after the fact.
Irritation isn’t a vibe. It’s a measurable decay effect. When you see diminishing returns despite sustained or increased promotion, you’re watching irritation outpace positive memory associations. And when you don’t account for that, which almost no one does, you end up over-communicating, over-asking, and blaming the donor for “fatigue” when the fault lies entirely with the machine.
As for the “goal of the model,” I appreciate the list you rattled off but it misses the mark. This model is built with a singular purpose: to grow donor value by personalizing how and when to interact with an individual donor. It is the operational answer to the sector’s most important question: what’s the best way to interact with this person to retain them and grow their value?
Your point about “donors don’t do what they say” seems like a red herring in this context. This model doesn’t use survey data. But more broadly, that critique is as wrong as it is right. Dismissing donor-reported data outright ignores decades of rigorous academic work on validated scales, indirect measures, and behavioral diagnostics. There are many critical insights you can’t get from behavior alone – e.g., motivation, identity, satisfaction, emotional response. Not everything shows up in the transaction log. But again, that’s not what this model is using, so not sure of your intent in raising that point.
And finally, I’d suggest we stop romanticizing “gut” or “experience” as some noble counterbalance to data. It’s local optimization. It leads to overfitting strategy to the last bad test or best-performing label pack. And it’s a big part of why donor files are shrinking.