Years ago, I had the privilege of training analysts on how to report fundraising results. One of my biggest rules was simple: Don't rehash what's already in the results table.
If revenue was down 8%, costs were down 30%, and response was up 66%, I didn't need an analyst to write a paragraph telling me those things. I can read the data table.
The value of analysis was explaining why those things happened. That distinction matters even more today because AI is very good at being our eyes.
It can read thousands of rows of data. Calculate changes. Find outliers. Compare segments. Summarize a dashboard. It can do in seconds the work that once required hours of manipulating spreadsheets. And we should use it for all of those things.
But reading the data is not the same thing as understanding it.
Data Is Only the Beginning
Imagine you upload campaign results and discover that response rate declined 15%.
That's data.
AI might quickly identify that the decline was concentrated among recently acquired donors.
That's an observation.
Then you ask why, and AI suggests that perhaps those newly acquired donors are less engaged with the organization.
That's a hypothesis.
But is it an insight? Not yet. Because maybe…
- New donors are coming in through a new acquisition channel and are less responsive to the previous cultivation methods.
- The organization expanded into weaker donor segments to increase the engagement of newly acquired donors.
- The offer or messaging has changed from what they were acquired from originally.
- A matching gift, premium, or other incentive disappeared.
- Another marketing piece was sent out right on top of this one.
- The campaign moved on the calendar and suddenly there was less time for the donor to respond.
- The organization's brand changed and caused donor confusion.
- A major news event or emergency was competing for donors' attention.
The results table doesn't know any of that. And unless we've given AI that context, it doesn't know it either. But AI will still give us an explanation. That's what makes this moment both so exciting and so dangerous.
Before generative AI, bad analysis often clearly looked like bad analysis. Revenue declined because response declined.
AI can take that same shallow thinking and turn it into three polished paragraphs, an executive summary, five recommendations, and a beautifully formatted list of "key insights." The writing sounds authoritative, and the explanation sounds reasonable.
That's why the skill we need today is gathering the full context of the donor's experience and building that into the analysis of the data.
Context Is What Turns Analysis Into Insight
One of the mistakes we make with fundraising data is treating metrics as though they exist independently. As if the donor's experience is narrowed down to a number.
- Response rate may fall because our ask scared away lower-value participants, resulting in the average gift increasing.
- Return on investment may improve by cutting out lower-value donors, but that might also mean less net revenue to do the work.
- An acquisition campaign may generate fewer donors, but it might be attracting donors with greater long-term value, thus creating more net revenue long term.
In one example from our old results-reporting training, two fundraising offers produced roughly similar overall revenue performance. One generated a higher response rate. The other generated a higher average gift.
If our only question is “which one won?”, the answer—it was a tie—isn't particularly useful.
But once we ask what we learned, we suddenly have two tools.
The higher-response offer might be valuable for second-gift conversion or lapsed donor reactivation. The higher-average-gift offer might be useful when trying to upgrade donors.
A metric rarely means much by itself. Insight lives in the relationships between metrics—and in the context of what we want a donor to do.
That's where AI can become extraordinarily valuable to us if we use it wisely. Instead of asking it to tell us the answer, we can use it to help us interrogate the evidence so that we can start providing it, and our analysis, with more context.
- What changed in the performance?
- What donor groups is the change concentrated in?
- Which audiences behaved differently from others and how?
- What other metrics moved at the same time, and what relationship might they have with each other?
- What are several explanations that fit the evidence thus far?
- What evidence contradicts those explanations?
- What information would we need to distinguish between competing hypotheses?
- What should we test next?
That is a very different use of AI than asking it to summarize a dashboard.
The Goal of Insight Is Action
There's one more test I use when deciding whether we've actually reached an insight: Does this change what we're going to do?
If it doesn't affect our audience, message, offer, creative, channel, timing, investment, or next test, we may have found something interesting, but we probably haven't found something particularly useful.
Good analysis should create a loop:
What is the result?
Then:
What could explain it?
Then:
What do we believe was the cause, and what evidence supports that belief?
And finally:
What are we going to do to prove our hypothesis and ultimately improve performance?
Sometimes the honest answer is that we don't actually know why something happened, and that's okay. Turn the best explanation into a hypothesis. Design a test. Measure the result. Learn something. That's how insight compounds.
AI Raises the Value of Curiosity
AI is going to make reporting faster.
It will make sophisticated analysis available to people who previously didn't have the time, tools, or technical skills to do it themselves. That's good news for nonprofits.
But as the cost of producing analysis falls, I think something else becomes more valuable: the ability to ask good questions, bring the right context, challenge easy explanations, and turn evidence into decisions.
The organizations that benefit most from AI won't necessarily be the ones with the most sophisticated tools. They'll be the ones that know which questions are worth asking, which answers deserve skepticism, and when a plausible explanation still needs to be proven.
Because the goal is to learn something today that makes our work better tomorrow.
