What’s more important: Focusing on gathering a full suite of perfect, clean data and then exploring it to see what it tells you, or framing a difficult question and then going out to find the data that answers that question? I think the right choice is the latter, at least for fundraising shops new to analytics.
There is nothing wrong with pulling a bunch of donor data to audit for errors or play with in an unstructured way, without having a definite research question in mind. Exploration can help you get familiar with the data, which is never a bad thing. There is also nothing wrong with addressing imperfections and errors.
But if you’re looking to make a difference and advance the development of analytics in your organization, you should zero in on the biggest question or questions, and DO NOT wait until your data is perfect in order to do it.
That’s the substance of one of the recommendations in a research report published recently by MIT Sloan Management Review in collaboration with the IBM Institute for Business Value. (PDF: Analytics: The New Path to Value — How the Smartest Organizations Are Embedding Analytics to Transform Insights into Action.) The report is rather general, buzzword-laden, and focused on the private sector, but the observations are valid.
The first option — data-gathering and pure exploration for its own sake — is a valuable activity for shops where analytics is part of normal operations, i.e. the minority of shops. Such exploration can lead to additional insights for which you know there will be a receptive audience and, more importantly, can lead to framing the important questions of tomorrow.
For today, though, focus on your organization’s thorniest problems. Tackling the big unknowns might seem the risky way to go if you’re looking for a quick win — but imagine the response to making some headway on conundrums such as these:
- Engagement of young alumni is a deep concern. How can we identify young alumni who graduated in the last five years who are most likely to become volunteers and ambassadors for their class?
- Donor retention beyond a year or two isn’t what it ought to be. Which lapsed donors are most likely to be reactivated if we increase our efforts in their direction?
- Fulfillment on pledges is abyssmal. Which pledges are most at risk of defaulting and need early attention?
- (Insert your institution’s burning issue here.)
The second part of this message is to go ahead and use the data even if it isn’t perfect (within reason). Some people in our business cannot stomach basing decisions on partial or imperfect data. The result of their cautious fastidiousness is not perfection, it’s stasis!
I’m reminded of one of the most common objections to using contact information variables such as ‘business phone present’ as predictors: What if some of those numbers are the result of research or data appends, not a result of alumni/donor engagement? Well, sure, if you have additional information about where the data came from, then use it to split the variable to allow for separate correlation testing. But if you don’t, why should you assume that the entire project has somehow been invalidated? What a shame if analytics came to a halt based on someone’s notions about data purity.
Success in analytics is not an all-at-once deal; it’s iterative. It goes like this: “Let’s get some kind of answer or focus this year, and through that we’ll discover what the valuable data is that we need to augment, improve, or clean up for next time. Then we’ll make another, better model next year.”