Most Canadian charities can tell you who gave the most money last year. Far fewer can tell you who is most likely to give again next month — and that gap is quietly costing organizations donors they could have kept, and gifts they could have asked for. Donor propensity scoring closes that gap. This guide explains what it actually is, how it works under the hood, and what it takes for a small or mid-size Canadian charity to start using it — without hiring a data scientist or ripping out your CRM.
Eighty percent of Canadian charities are already experimenting with AI in some form. But only nineteen percent send genuinely targeted, segmented appeals — the rest are still sending the same email, the same letter, and the same ask amount to everyone on the list, regardless of what that donor's own giving history says about them.
At the same time, nearly half of Canadian nonprofits are actively considering switching CRMs in the next while. Read that carefully: it means the tools charities already have aren't answering the one question every fundraising director actually needs answered day to day — who should I call this week?
That's the specific problem propensity scoring solves. It doesn't replace your relationship with your donors. It tells you, out of everyone in your database, which donors are statistically most likely to respond right now, so your limited time and your limited campaign budget go toward the people most likely to say yes.
Strip away the AI branding and a propensity score is a straightforward idea: a number, usually 0 to 100, that estimates how likely a specific donor is to take a specific action — give again in the next 30 days, respond to a specific campaign, or lapse and stop giving altogether — based on patterns in their own giving history.
It's the same math retailers use to guess which customers will buy again, and the same math email platforms use to guess which subscribers will open the next newsletter, applied instead to donation records. Dataro, the Australian fundraising-intelligence company that popularized this approach for nonprofits, puts it simply: past behaviour, analyzed consistently at scale, predicts future behaviour better than gut instinct does. Not perfectly — but reliably enough to change where a fundraising team spends its time.
A propensity model doesn't need anything exotic to work. It looks at a handful of signals that any experienced fundraiser would recognize instinctively — the model just applies them consistently, to every donor, every time, instead of relying on whoever on staff happens to remember that a specific donor always gives in June.
None of these signals are secret. What propensity scoring adds is consistency at scale — running the same disciplined analysis across a database of 3,000 donors that a skilled major-gifts officer already runs, informally, across the 40 donors she knows best.
Not every charity needs the same model, and a good propensity tool should adjust to how much donor history an organization actually has rather than forcing every charity through the same statistical process.
For a charity just starting to track identified donors — under a few hundred — simple rules already capture most of the value: gave in the last 90 days = warm, hasn't given in 18 months = at risk. No statistical model required.
Once a charity has a few hundred to a couple thousand donors, a calibrated statistical model (logistic regression, in plain terms) can estimate a real probability of giving again, rather than a rule-of-thumb bucket.
At larger donor volumes, a more sophisticated model can weigh dozens of signals at once and — critically — still explain, in a single plain-language sentence, why a given donor scored the way they did. That explainability is what separates a tool a fundraising team will actually trust from one they'll quietly stop using.
It's worth being direct about the limits, because overselling this technology is exactly how charities end up disappointed by it.
It is not mind-reading. A high score means a donor's pattern resembles other donors who gave — it's a probability, not a guarantee.
It does not replace relationship management. A propensity score tells you where to look first. It doesn't write the thank-you call, and it doesn't replace years of institutional knowledge about a major donor's family situation.
It should not be a black box. If a tool tells you a donor scored 87 without telling you why, you have no way to sanity-check that against what you already know about them — and no way to explain the recommendation to your team. A trustworthy propensity tool surfaces the actual signal behind the score ("consistent June giver, four years running") in plain language, not just a number.
Most propensity-scoring tools on the market were built for the US or Australian markets, and it shows. Canadian giving has its own shape: roughly 30% of annual giving is concentrated in December, driven by the same tax-receipt calendar every donor is working against; CanadaHelps is a giving platform with no real US equivalent; and Canada's Anti-Spam Legislation (CASL) — not CAN-SPAM — governs whether you're even allowed to email a given donor about a specific appeal.
That last point matters more than it sounds like it should. CASL requires charities to track express and implied consent per contact, per channel, and implied consent expires after 24 months. A propensity model that hands you a perfectly ranked donor list is only useful if you also know which of those donors you're legally allowed to contact — which is why a Canadian-specific tool worth using treats consent tracking as part of the output, not an afterthought.
This is the part that surprises most fundraising directors: you do not need a data team, a CRM migration, or months of setup.
What you actually need, at minimum:
That's the floor. Everything else — campaign codes, cause tags, channel data, consent status — makes the model sharper, but isn't required to start. If your donor history lives in Raiser's Edge, Bloomerang, CanadaHelps, or a spreadsheet nobody's touched since last year, you can export it as a CSV and get a first look at scored data the same day.
As a rule of thumb, propensity scoring gets meaningfully more predictive somewhere around 500 identified donors. Below that, simpler rule-based segmentation still gets you most of the value without needing a statistical model at all — which is exactly why a well-designed tool should offer both, and move between them automatically as a charity's donor base grows.
In practice, a propensity score isn't something you stare at — it's something you act on, in a handful of recurring situations:
Most CRM-native scores are useful for internal segmentation, but they typically don't predict a next likely giving date, don't flag lapse risk before it happens, and don't track consent windows. They're a starting point, not a substitute for a purpose-built propensity model.
If the tool requires you to understand the statistics behind it, it's the wrong tool for a two-person fundraising team. A good propensity platform should do the modeling and hand back a plain-language reason, not a spreadsheet of coefficients.
For Canadian charities, this is a legitimate and specific concern, not a generic one — ask directly where the data is processed and stored, whether it's pooled across organizations, and how CASL consent is tracked, before evaluating anything else about a tool.
The opposite is usually true in practice. Large nonprofits with dedicated analysts have already built informal versions of this. It's the charity with one or two fundraisers and no analyst who gets the most out of a tool that does this work automatically.
Take a mid-size Canadian charity with roughly 2,000 identified donors and a two-person fundraising team planning a spring appeal. Instead of sending one letter to all 2,000, a propensity model returns a ranked list: the top 150 donors score above 80, each with a one-line reason attached — "gave to a similar appeal in each of the last three springs" or "gift size has grown 20% year over year, no missed cycles." The team calls the top 30 personally, sends a slightly more personalized email to the next 120, and sends a lighter-touch, lower-cost version to the rest of the file. The result isn't a bigger list — it's the same list, worked in the right order, with the two staff members' limited time spent on the names most likely to convert it into a conversation.
That's the entire value proposition in one example: not more donors, not more emails — better sequencing of the same finite time and the same finite donor file.
It's a fair question, and the honest answer is: accurate enough to change behaviour, not accurate enough to replace judgment. A well-built propensity model, validated against a charity's own historical giving data rather than a generic industry benchmark, typically identifies a top segment that converts at a meaningfully higher rate than the file average — often two to three times higher for the highest-scoring decile compared to a random or alphabetical send. It will still be wrong about individual donors sometimes. The value isn't perfect prediction for any one name; it's a consistently better hit rate across hundreds or thousands of names than guessing, remembering, or sending to everyone equally.
If you're considering this for the first time, the honest timeline looks like this: export your donor CSV in week one and get a first scored view back the same day. Spend the rest of week one sanity-checking the top and bottom of the list against what your team already knows — this is the step most charities skip, and it's the one that actually builds trust in the tool. By week two, run a real campaign, ideally a small one, against a scored segment instead of the full list, and compare the response rate to your usual broadcast approach. By week four, you'll have enough of your own results to decide whether this belongs in your regular fundraising cycle — not someone else's case study.
Donor propensity scoring isn't a replacement for good fundraising judgment — it's a way of giving that judgment better information, faster, across a donor list too large for any one person to hold in their head. For Canadian charities specifically, the tools that matter are the ones built around CASL consent tracking and Canadian giving patterns from day one, not retrofitted from a US or Australian product afterward.
See what your own donor data says.
GivingSignal's Donation Likelihood tool scores your database directly from a CSV export — no CRM migration required. And if you're not ready to run a full campaign yet, the free CASL Consent Tracker is a good first step: it shows you where your consent records actually stand before you send anything.