I get it. You are a people person. It’s why you are so good at cultivating and maintaining relationships with your donors.
Then, one day, you took a promotion with a new organization to head up their fundraising team. And in your first meeting with your new Executive Director, you’re asked to take a look at the recent fundraising data to “do some analysis.”
This is not a people person sort of assignment.
Undeterred, you wade into your donor database and start running some reports provided by the software. Maybe there’s not much data there. Even worse: maybe there’s a whole lot of data there.
You start pulling everything you can get your hands on into an ever-growing spreadsheet. You look for more numbers — maybe some charts (“Yeah, charts! I need charts!” you say to yourself) — and then finally you start to wonder just what exactly you’re looking for, and it sets in: the dreaded analysis paralysis.
If this sounds familiar, you’re not alone. As we hear more and more about becoming #DataDriven, nonprofits are struggling with where to begin.
For some organizations, the problem begins with data collection; if you’re going to analyze data, you need… well… data. Others may have reams of data, thanks to using modern tech tools that make data collection easy.
It may sound like equal parts complex mathematics and arcane wizardry. And in some cases, it is — talented data scientists can do amazing things, using vast arrays of data to build complex predictive models that can determine the likelihood of someone taking a specific action with incredible accuracy. But whether you are building just such a model, or you’ve just been tasked with “doing a little analysis” for the first time, successful analysis comes down to one crucial factor: asking the right question.
How to Ask the Right Question
At its heart, that’s what a real data analysis project is designed to: use data to find the answer to a particular question. The trick is, you have to make sure you are asking the right question.
In a predictive modeling scenario, that question might be “Which of my constituents is most likely to give a single gift of at least $10,000 in response to a personal solicitation?” This question is very specific, and lays out the exact parameters to be tested.
Chances are, you have a friend or acquaintance who enjoys being a language cop whenever you ask a question. Most of us probably first encounter this as schoolchildren:
You: Can I go to the bathroom?
Joyless Teacher: I don’t know, can you?
You: May I please go to the bathroom?
Joyless Teacher: You may.
Well, data analysis is very much like that joyless teacher. So before diving in to find an answer, take the time to ask the right question. For an example, let’s look at our modeling question above:
“Which of my constituents is most likely to give a single gift of at least $10,000 in response to a personal solicitation?”
Again, this is a great question, with absolute specificity. To get to this, though, you might start with a less-than-ideal version of the question:
“Which constituents should I ask for a large gift?”
So what makes this question less-than-ideal?
For starters, “large gift” is pretty vague. How exactly are you going to measure “large”? Some emerging nonprofits might look at anything over $500 as a large gift, while others might define large gifts as those $500,000 and up. There’s no single definition of “large” for every nonprofit. So we need some parameters. In our ideal version of this question, we define large as $10,000 and up.
There are a lot of ways to ask for a gift. Sure, not all of them are ideal — I wouldn’t recommend contacting a donor via their Twitter account to ask for $10,000 — but remember, good analysis requires specificity. “Ask” is a little too vague, which is why, in our ideal question, we specify a “personal solicitation.”
Finally, we come to the most nit-picky critique. Saying, “Which constituents should I ask” seems OK, but it doesn’t imply any inherent value that makes someone better or worse to ask for a particular gift. The snarky response (which is always the most fun way to respond) would be, “whoever will give you a gift.” And that’s really what we want to identify: who will give you a gift? Well, analytics can be amazing, but they fall a wee bit short of precognition. So, the best way to phrase this is in our ideal question, where we ask who is most likely to give a particular gift.
Asking good questions is not limited to statistical modeling projects. For example, you might have simpler questions like:
Which event provided the greatest lift in donations from event attendees in the six months following the event?
What are the top three direct mail appeal topics in the past year have yielded the response rate?
Which social media platform has brought in the highest number of new donors over the last three years?
Obviously, the right question will vary with the needs and activities of your organization. But finding the right question is the key to starting any data analysis project.
Focusing Your Attention on the Right Data
Once you have a good, specific question, it becomes much easier to look at the pool of data available to you and make decisions about which data points will be most useful in answering your question. Instead of drowning in a sea of “big data,” unsure where to start or how to determine what’s most important, you can look at the particulars of your questions and then identify the key data points most likely to lead to answers.
For example, let’s look at the events question in our list: “What are the top three direct mail appeals in the past year have yielded the response rate?” To find an answer to this question, only need to look for:
Total number of recipients of all direct mail appeals within the last 12 months
Total number of financial responses to all direct mail appeals within the last 12 months
Topics for each direct mail appeal within the last 12 months
And that will give us our answer.
Of course, this is a simple example. More complex questions, like the original ideal question in this post, will require many more data points. And if you’re planning to build a predictive statistical model, you’ll need a little advanced mathematical know-how.
The next time someone asks you to “do a little analysis,” don’t just jump in and start wading through reams of data hoping insight will magically spring out from between the columns of an enormous spreadsheet. Instead, take some time to ask some very specific questions, and then look for the data points that are best suited to help you find the answers.
Who knows? A little focused data analysis time could help you ensure you’re using those incredible people skills of yours on the people who are best suited to help you raise more money and do more good.